Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Bottom-up energy supply optimization of a national building stock

Bottom-up energy supply optimization of a national building stock The installation and operation distributed energy resources (DER) and the electri cation of the heat supply signi cantly changes the interaction of the residential building stock with the grid infrastructure. Evaluating the mass deployment of DER at the national level would require analyzing millions of individual buildings, entailing signi cant computational burden. To overcome this, this work proposes a novel bottom-up model that consists of an aggregation algorithm to create a spatially distributed set of typical residential buildings from census data. Each typical building is then optimized with a Mixed-Integer Linear Program to derive its cost optimal technology adoption and operation, determining its changing grid load in future scenarios. The model is validated for Germany, with 200 typical buildings considered to suciently represent the diversity of the residential building stock. In a future scenario for 2050, photovoltaic and heat pumps are predicted to be the most economically and ecologically robust supply solutions for the di erent building types. Nevertheless, their electricity generation and demand temporally do not match, resulting in a doubling of the peak electricity grid load in the rural areas during the winter. The urban areas can compensate this with ecient co-generation units, which are not cost-ecient in the rural areas. Keywords: energy systems, typical buildings, Mixed Integer Linear Programming (MILP), Zero Energy Buildings (ZEB), building stock, aggregation 1. Introduction well as governments [13]. Therefore, analyses are needed that predict technological development and their system 1.1. Background integration: Many works for the building sector [14{18] The residential building sector is responsible for 17% solely focus on GHG reduction strategies for heat demand. of worldwide CO emissions [1]. In Germany, it was the They conclude that signi cant energy saving potentials source of 10% of the total Greenhouse Gases (GHG) through can be accessed by increased refurbishment rates and that fossil fuel combustion in the year 2015. Moreover, it was residual heat can be supplied with renewable energy. responsible for 12% of total emissions due to the GHG Nevertheless, in relation the heat demand of the build- footprint of its energy imports [2{4]. These emissions must ing sector can no longer be regarded as being more isolated be cut in order to reach the overall goal of net zero GHG from the other energy system: Heat pumps are seen as a emissions in the second half of this century [5] with the key option to eciently provide space heat [19, 20], while goal of minimizing the impact of anthropogenic climate Combined Heat and Power (CHP) generation allows an change [6]. Therefore, the European Union introduced the ecient usage of chemical energy carriers while providing concept of "Zero Energy Buildings" (ZEB) in the context exibility to the grid [21]. Furthermore, a trend towards of its energy performance of buildings directive [7, 8] with an increased self-supply of residential buildings is appar- the goal of deploying GHG-neutral buildings that com- ent with the rapidly falling prices of photovoltaics [22] and pensate for their emissions by exporting on-site generated batteries [23] constituting grid parity [24], meaning that renewable energy [9, 10]. the levelized cost of self-generated electricity is below the While the objectives are clear, the pathway to GHG- retail electricity grid price. neutral building stock is uncertain and the integration of Both trends, i.e., the changing heat supply and the in- new technological solutions unsettles utilities [11, 12] as creasing self-suciency of the buildings, will signi cantly change the future grid demand and challenge the feasibil- ity of current electric grid design. Therefore, new analyses Corresponding author. Email: l.kotzur@fz-juelich.de Preprint submitted to Energy & Buildings January 7, 2020 arXiv:2001.01554v1 [physics.soc-ph] 2 Jan 2020 Nomenclature CAPEX CAPital EXpenditure OPEX OPerational EXpenditure CHP Combined Heat and Power QIP Quadratic Integer Program CRF Capital Recovery Factor RMSE Root Mean Squared Error DER Distributed Energy Resources SFH Single-Family House GHG Greenhouse Gas UB Upper Bound LB Lower Bound WACC Weighted Average Cost of Capital MILP Mixed-Integer Linear Program MFH Multi-Family House ZEB Zero Energy Building are required that consider the adoption and operation of is guaranteed. Some of the models even include the in- new supply technologies and eciency measures, predict- vestment decision into eciency measures by changing the ing the spatially- and temporally-varying impacts on the buildings envelope [41, 45{47]. grid infrastructure. All models would enable the analysis of the impact of technology adoption and operation on the local infrastruc- 1.2. Related works tures: for instance, Lindberg et al. [37] apply a MILP to design the supply system of a Multi-Family-House (MFH) Various works have already perforemed top-down anal- and analyze the resulting electricity grid load for cost- yses of the load change due to single Distributed Energy optimal system operation under current German regula- Resources (DER), like photovoltaic [25], heat pumps and tions. Schutz  et al. [39] use a model to evaluate the opti- battery electric vehicles [26], or Combined Heat and Power mal technology adoption with currently considered incen- (CHP) [27]. Seljom et al. [28] analyze the impact of pho- tives and market conditions for the case of three reference tovoltaics deployed in ZEB in for the future Scandinavien buildings. energy system. Nevertheless, as demand, exibility op- tions and generation are closely connected to the building 1.2.2. Archetype buildings supply systems, these technologies can not be evaluated Although these analyses and models already provide independently and must be get holistically regarded and many insights for the application cases, a further gener- modeled as system entities. This can primarily be done by alization would be required to upscale these results to an means of detailed bottom-up models that simultaneously aggregated nationwide perspective. Furthermore, the spa- consider investment decisions and the operation of DER as tial variation due to regionally di ering building topologies well as eciency measures and demand side management. would be required to integrate the results to grid mod- els. Therefore, a set of representative buildings is required 1.2.1. Building optimization that characterizes the spatial diversity of building stock Thereby, the models must account for cost optimality, and that can be used for the previously described building as the main motivations of building owners to adopt di er- models. ent supply technologies are savings or earnings emanating In general, such typical buildings are often referred to from their installation [29]. This also applies for the ef- as archetype buildings and are commonly used for model- ciency measures or energy retro ts, where the need of ing GHG reduction strategies in the building sector [48], replacement or nancial pro tability are the main activa- as described in the Energy Performance of Buildings Di- tors for the adoption [30]. rective [7, 8] of the European Union. Therefore, many di erent optimization models have In this context, Corgnati et al. [49] introduced di erent been proposed for determining the cost optimal invest- pathways to determine representative reference buildings ment decisions and operation of building supply systems: for the analysis of cost optimal refurbishment measures, either as Linear Programs (LP) [31, 32] with the advan- but conclude that in reality most often a mixture is used tage of good computational tractability but the disadvan- due to the di erent available data for di erent buildings tage of not being able to account for economies of scale; stocks. or as a Mixed Integer Linear Program (MILP) model [33{ Mata et al. [50] proposed an analytical methodology 39]. Furthermore, two-level approaches that determine at to aggregate archetype building stocks based on publicly least a part of investment decisions with a meta-heuristic available data. The steps include a segmentation based solver and operation with a simulation or optimization are on categories such as construction year, a technical char- popular [40{45]. The last approach can account for very acterization such as the thermal transmittance as input detailed physical models but no global optimal solution values for energy performance models, as well as a quan- 2 ti cation to scale the buildings up to a nationwide level. 1. aggregate a spatially resolved building stock with The methodology is applied to France, Germany, Spain, a limited number of archetype buildings which and the UK, with the resulting nal energy demand show- are described by a set of attribtues related to their ing a deviation of less than -4% to + 2% to the aggregated energy supply and demand statistical values. 2. and to optimize those di erent buildings in par- Various nation-speci c works exist to quantify the en- allel, considering a superstructure of supply tech- ergy consumption of the building sector with the help of nologies and potential eciency measures. archetype buildings: the German residential building stock Thereby, di erent regulatory regimes can be considered, is described by a framework developed by the Institute fr resulting in di erent future technology installations and Wohnen und Umwelt (IWU) [51, 52]. This schema cat- operations, and grid demands. The spatial assignment of egorizes the stock into classes that di er by construction the archetype buildings allows then for the local evalua- year and building size that are represented with detailed tion of the changing energy demand. The general idea is technical parameters. This stock description has been ex- visualized Figure 1. tended to other European countries in the framework of the EPISCOPE project [53]. The US Department of En- ergy (DOE) introduced archetype buildings for the resi- dential sector [54] and the service sector [55] in the USA, referred to as benchmark or prototype buildings. An ad- vantage over the European database [53] is the fact that additional time series data are provided for the di erent building demands, including electricity, hot water, cool- ing or heating demand for typical days in di erent climate zones. The aggregation to archetype buildings is also widespread in the context of urban energy models: DallO et al. [56] present a work ow to derive archetype buildings with Figure 1: Structure of the bottom-up model to optimize the a spa- a combination of statistical data and a survey applied tially resolved building stock and determine the changing infrastruc- to sample buildings. Meanwhile. Cerezo et al. [57] and ture usage. Sokol et al. [58] introduce methods to estimate unknown attributes of the proposed archetype buildings, such as With the help of this modeling approach, the poten- comfort temperature levels, based on a probability distri- tial for self-sucient energy supply systems in residential bution. This approach can make use of measured energy buildings can be eciently evaluated and its large-scale data in di erent buildings, such as annual or monthly gas techno-economic impact on the grid demand can be de- demand, and ts the uncertain attributes to it. More- rived. To do so, Section 2 introduces the used aggregation over, Fazlollahi et al. [59] and Fonseca and Schlueter [60] tool, the used data sets and the parallel building optimiza- use k-means clustering methods to group similar buildings tion for this model. The approach is applied for a reference in urban areas by using the location of the buildings and scenario in Section 3 and validated to the available aggre- spatially resolved statistics. The advantage of clustering is gated energy demand data in Germany. Thereby, a trade- that the simulation models or optimization models can be o is made regarding the number of archetype buildings applied to the zones instead of the single buildings, which to describe the diversity of building stock on one hand, reduces the number of variables and the computational but limiting the computational by the number of Mixed- load of the related models. Integer Linear Programs to solve. In order to derive the changing load, a future supply scenario is introduced in 1.3. Own approach and structure Section 4 for the year 2050. Section 6 critically recaps the In summary, many works exist that consider detailed work and draws the main conclusions. integrated building optimization models to determine the cost optimal technology adoption and operation. There also exist many di erent approaches for the aggregation 2. Methods and data and segmentation of archetype buildings. Nevertheless, to the knowledge of the authors the combination of both The model consists of two main components: for the purpose of analyzing the spatially and temporally 1. The building optimization model, described in Sec- changing demand of the infrastructure, i.e. gas grid and tion 2.1, represents the decision making of the build- electricity grid, does not exist. ing owner regarding the design and operation of the energy supply and energy demand. 1.3.1. Therefore, this work proposes a two stage framework to 3 2. The aggregation and distribution of archetype build- 2.1.2. Optimizing structure, scale and operation ings, introduced in Section 2.2, makes the build- The building optimization is based on a typical Mixed- ing optimization results generalizable to the perspec- Integer Linear Program (MILP) with the objective of min- tive of regulators and grid operators on a nationwide imizing the annual energy cost of a single building as pro- scale. posed in the vast literature. The operation and design of the supply system is modeled with the object-oriented 2.1. Building optimization system modeling framework FINE [65, 71, 72]. The bi- It is assumed that building owners are the decision nary variables are considered to suciently incorporate makers and they consider the energetic supply of the build- the economy of scale of the technologies. The operation is ing from the perspective of a single economic entity. There- modeled in a fully linear and continuous manner for 8760 fore, the goal of the building optimization is to have a hours in a representative year. holistic perspective of single residential buildings and is All in all, the combinatorial consideration of demand- able to consider synergies between di erent solutions, e.g. side and supply-side measures respecting the full opera- demand-side measures are simultaneously considered with tional variety yields a complex mathematical program that supply-side measures, or the operation of the heating sys- is computationally demanding. In order to keep the pro- tem is optimized together with the operation of the elec- gram tractable for many di erent building types and sce- tricity system. narios, the annual time series of weather, occupancy be- havior, and appliance load are aggregated to twelve typi- 2.1.1. Creation of demand and supply time series cal days with a hierarchical aggregation [72, 73]. The days In order to derive the cost optimal supply system for with the smallest temperature and highest electricity load each individual building, rst the temporally varying en- are added as extreme days. Based on these, the optimal ergy demands of electric devices, hot water and thermal choice of the supply technologies and refurbishment mea- comfort, as well as the varying performance of renewable sures is determined with the MILP. The binary decision supply technologies are derived as follows: variables are then xed and a validation and scaling opti- The occupancy behavior and inherited electricity load mization is then performed for the full annual time series of the appliances are created with the help of the CREST [74], similar to that in Bahl et al. [75]. demand model [61{64]. Further, the demand for hot water The detailed model description and their independent is separated from it. The advantage of this model is that validation can be found in Kotzur [65]. it suciently incorporates the high variance of single res- idential load pro les, as well as the stochastic smoothing 2.2. Aggregation of archetype buildings for the case that an agglomeration of households is consid- In order to determine the di erent types of buildings ered. A validation of the model in the context of German the number of their occurance, this section introduces a residential electricity demand is performed in Kotzur [65] new aggregation method to derive spatially-distributed archetype and exhibits sucient accuracy. buildings. Those are used to scale the results of build- The heat load is considered with the 5R1C-model from ing optimizations to a spatially-resolved nationwide per- EN ISO 13799 [66], which was implemented into a MILP spective. Therefore, Section 2.2.1 discusses the relevant by Schutz  et al. [47]. The physical building properties, attributes to describe the energy performance of a build- such as heat transfer coecients for di erent construction ing. The aggregation algorithm itself is sketched in Sec- years are taken from IWU [52]. This is able to account for tion 2.2.2 and its application is illustrated in Section 2.2.3. the thermal building mass for a exible supply system op- 2.2.1. Relevant building attributes eration. Furthermore, potential refurbishment measures are part of the solution space, such as the addition of wall In general, four categories of building attributes are or roof insulation, the replacement of windows, or the in- emphasized in the literature, while the concrete nomen- tegration of smart thermostats. The con guration of the clature varies [49, 50]: The Form describes the physical buildings is introduced in detail in Kotzur [65]. exterior shape of the building, including orientation, wall The time series for PV and solar thermal are created area, window area and roof areas. The Envelope charac- with the PV-Lib [67]. The weather data is derived from terizes the physical properties of the materials used in the the DIN EN 12831 [68] by nding the closest location building. The technologies installed in the buildings to sat- listed. Therefrom, the minimal design temperature is de- isfy thermal comfort and other demands are grouped into rived as well as the test reference region of the Deutsche the category of System. Operation, in turn, summarizes Wetter Dienst (DWD) [69]. Alternatively, the weather all extrinsic conditions determining the system operation, from the COSMO rea-6 reanalysis data set [70] is used for such as the local weather or occupancy behavior. real weather years for validation purposes. Aside from the attributes describing the current en- The whole initialization of the building speci c time ergy performance, future energy supply is also of interest, series are published in the open source python package where the category of Adoption summarizes all attributes tsib - Time Series Initializaion for Buildings (https:// referring to the investment capabilities and investment be- github.com/FZJ-IEK3-VSA/tsib). havior of the building owner, such as a potential interest 4 rate. For instance, the model described here considers the cost-optimal technology adoption of the di erent build- ings. The categories and their aggregation de ne a general framework to segment buildings, but the required attributes depend on the model and data availability, e.g., the enve- lope could be described by materials with exact heat con- ductivities and thicknesses, only heat transfer coecients, or by the construction year of the building from which these values are derived. The attributes considered for the aggregation proce- dure of this work are oriented towards the model intro- Figure 3: Structure of the considered attributes that relevant for the duced in Section 2.1 and the data provided by the cen- building energy supply. sus [76]. Figure 2 shows the aggregated Census data for Germany. The total number of buildings is predominated by over 12.3 million Single-Family Houses (SFH) and 6.3 Thereby, two challenges arise: rst, the buildings are multi-family houses, while buildings with more than 12 described by a mixture of categorical and continuous at- apartments have, at 0.21 million, a small share at the total tributes. Approaches exist dealing with this type of aggre- number. The majority of the SFH are detached, consti- gation class, such as the mixture of k-means and k-modes tuting an overall small proportion of terraced and semi- clustering, referred to as k-prototypes [77]. Nevertheless, detached buildings. 23.2 million of the 40.5 million apart- these would rely on a data set consisting of real building ments are rented, while one- and two-person households instances that should be clustered and can then be rep- dominate with together 27.1 million households. These resented, e.g., by its medoid. This does not apply to the also constitute the peak of apartment sizes, with compact attributes distributions, and so a new aggregation method- living areas of 59 to 79 m per household, while larger ology is required that is introduced in Kotzur [65]. single-family houses are spread over a larger grouping. Thereby, a greedy algorithm is introduced with the goal 1000 Owner- Rented ship Owned of determining a locally optimal set of archetype build- Apart- ments ings. It is inspired by the concept of an expectation- Persons per 3 Terraced app. Sur- 2 maximization algorithm, where Lloyd's k-means cluster- Semi rounding Detached 2020 ing algorithm [78, 79] and the k-prototypes algorithm [77] 179 belong as well. Con- Apart- 159 struction 1987 ment 139 year 2 size [m ] 119 before 1969 below 1959 79 1949 59 1919 39 0 5 10 0 10 20 Buildings [Mil.] Appartments [Mil.] Figure 2: Aggregated attribute distribution of the German residen- tial building stock based on the Census [76]. Figure 4: Flow chart of the developed algorithm to determine a All of these distributions are also available on an abso- spatially distributed archetype building stock. lute scale for the municipalities or a 100m grid in Germany and state the data basis for the considered archetype ag- gregation, as shown in Figure 3. The introduction of ad- The idea is to describe the assignment of the archetype ditional values is discussed in detail in Kotzur [65]. buildings to the di erent nodes or municipalities as the expectation step, with the objective of getting a repre- 2.2.2. Building aggregation algorithm sentation of the attribute distributions by the most likely These attribute distributions must be aggregated to a archetype buildings into every municipality. Nevertheless, limited set of archetype buildings to evaluate them ener- the attributes of the archetype buildings themselves are getically. unknown, and so their estimation is de ned as the maxi- In general, the attributes belonging to the di erent mization step, illustrated in Figure 4. This results in two categories are primarily published as aggregated distri- optimization problems that are iteratively solved. butions for di erent administrative boundaries that de- The result is a set of archetype buildings and a matrix ne the spatial resolution. Nevertheless, the real building that de nes the representation of every municipality by instances and their values are unknown wherefore cross- the number of certain archetype buildings. combinations of attributes are not reproducible, e.g., how many terraced buildings have a certain living area. 5 2.2.3. Application and illustration of the results low 2 %. Thus, it would not be ecient to sacri ce an additional archetype building to represent it. The initial guess of the archetype building attributes, In general, a t below 100 % does not imply that the the start solutions for the algorithm, are derived from the expression is highly under-represented on the aggregated current state of the art archetype buildings for Germany level: While an overestimation of an attribute in one re- [52], while missing parameters are randomly generated, gion and an underestimation in the other regions consti- e.g. the number of persons living in an apartment. tute a reduced t, they could add up and compensate each In the ollowing, the algorithm is applied once to di er- other on an aggregated nationwide level, which is further ent prede ned numbers of archetype buildings in order to elaborated in Kotzur [65]. determine how many of these are required for a sucient The tting of the continuous attributes, the latitude representation of the German building stock. and longitude, is qualitatively illustrated in Figure 6 with The quality of the resulting representation of the di er- their exact geographical placement in Germany. For the ent attributes for di erent numbers of archetype buildings case of 5 to 25 archetype buildings, all buildings are pri- is illustrated in Figure 5. This is de ned as the cumulative marily located in the center of the country. The reason deviation of the representation of an attribute expression is that building archetypes are mainly used to represent m for every region in proportion to the total attribute the diversity of categorical attribute combinations that manifestations for the whole of Germany: are spatially distributed across Germany. E.g., a single- P P d d n;m;p n;m;p b;n b;p;m family house from 1960 with a four-person household and n2N b2B f (p; m) = 8 p; m gas boiler supply manifests as an archetype building that n;m;p n2N represents this building type in the north as well as in the (1) south. For higher numbers of archetype buildings from 100 Number of buildings to 800, the geo-spatial location of the archetype buildings 5 25 100 400 is spreading, as similar categorical building types can be 10 50 200 800 instantiated multiple times. For the case of 800 archetype CHP 6 District 5 4 buildings, it is even observable that urban areas are rep- Electric Supply Persons 3 tech- Pellets nology Heat pump 2 resented by more archetype buildings than rural areas. Gas boiler Oil boiler Photo- 0 voltaic Rebuild Apart- ment 139 Rented Owner- size 119 ship Owned below 90.0 67.5 Orien 45.0 tation 22.5 Con- Apart- ments struction 2 year before 1 1969 Number archetypes Terraced 5 Buildings Sur- Semi 10 Buildings rounding Detached 1919 25 Buildings 0 50 100 0 50 100 50 Buildings Attribute fit [%] Attribute fit [%] 100 Buildings 200 Buildings 400 Buildings 800 Buildings Figure 5: Relative t of all categorical attribute expressions for dif- ferent numbers of archetype buildings in Germany. Figure 6: Geographical location of di erent numbers of archetype buildings in the nal iteration step. The gure shows that for some of the attribute ex- pressions, a small number of archetype buildings is able to represent them suciently, for instance single-family The representation of municipalities is further illus- houses with a single apartment or the energy supplied by trated in Figure 7, which introduces the local assignment gas boilers. These are attribute expressions that often oc- to the di erent municipalities of certain example archetype cur in the original data set. Therefore, they are rst rep- buildings, selected from the set of 800. The locations of resented by the archetype buildings to reduce the overall the archetype buildings are the centers of the buildings error. Nevertheless, attributes such as a CHP, heat pump they represent in the di erent municipalities. As these supply, or apartments with a living area smaller than 39 representations are spatially spread over di erent munici- m rarely occur in the Census data set. Therefore, the palities, the archetypes are not placed at municipalities at algorithm has a secondary priority to represent them and the border. focuses instead on building attributes that exist more of- Furthermore, it is recognizable that the areas and amounts ten, e.g., no archetype building was created with a heat that are represented di er between the archetypes: While pump supply for 5, 25 and 50 archetype buildings because an archetype building supplied with heat pumps must rep- the overall share of heat pump supply in Germany is be- 6 resent buildings over a large area, archetype single-family Energy carrier Electricity houses supplied by gas boilers have a de nite local assign- Renewables ment area. The reason for this is that more archetype District heating Oil buildings with gas boilers are selected, as more buildings Gas 200 Coal with gas boilers also exist in reality. Therefore, the algo- rithm chooses a higher spatial separation for them to min- imize the overall error, while accepting a higher geospatial estimation error for the few buildings with heat pumps. Archetype buildings Archetype building Archetype building Figure 8: Final residential energy demand predicted for di erent with heat pump with gas boiler numbers of archetype buildings. can be seen in Figure 8, the model is able to roughly pre- Location arche- type building dict with ve to ten archetype buildings the demand of the 1000 assignments 100 assignments three dominant energy carriers, but the appearance of mi- 10 assignments nor energy supply carriers is not suciently included. This improves with an increasing number of archetype build- ings while the best t can be achieved with 800 buildings. Nevertheless, the resulting demands of 64.8 TWh/a for re- newables and 44 TWh/a for district heating are still under- estimated. This deviation is constituted by the aggrega- Figure 7: Location of the most northern (blue) and most southern tion, which tries to capture the most frequently appearing (green) single-family house archetype with heat pump supply (left) archetype buildings and neglects rarely occurring building and gas boiler supply (right) of the set of 800 archetype buildings, types. Nevertheless, these missing energy demands for re- and their assignment to the di erent municipalities. newables and district heating are compensated by gas and oil demands, which are slightly overestimated with 286 and 165 TWh/a. This compensation e ect already appears for 25 archetype buildings, where all cases between 25 and 3. Validation of the method 800 archetype buildings predict the total nal energy de- mand in a similar magnitude as the AGEB [2]. Above For the validation of the two-staged methodology, the 200 archetype buildings, the share of the di erent energy di erent numbers of aggregated archetype buildings are carriers also aligns well with the structure of the AGEB independently optimized for the status quo and then mul- [2]. tiplied with their appearance in Germany. The choice The prediction with 50 archetype buildings overesti- of the technologies is prede ned by the archetype def- mates the demand for oil by 27.4% and underestimates the inition [65], but the technology scale and operation are demand for gas by 17.1%, while 100 archetype buildings optimized such that the building-speci c energy demands on the contrary overestimate the gas demand by 25.5% are met. It de nes the Reference, or status quo, of the and underestimate the oil demand by 16.1%. This switch residential energy supply and validates the model to na- shows a drawback of the aggregation: Some archetype tional energy demand statistics. The techno-economic as- buildings appear often and therefore have a high impact sumptions are introduced in Appendix A. The di er- on the overall energy load. If the majority of the buildings ent sets of archetype buildings together with their spa- supplied, e.g., with gas boilers have a construction year tial distribution can also be found in the tsib (https: before 1960 while the more modern buildings are supplied //github.com/FZJ-IEK3-VSA/tsib). with oil, an overestimation of the gas demand and an un- derestimation of the oil demand results, although the abso- 3.1. Impact of the number of archetype buildings lute number of the di erent boiler types is well represented. The impact of choosing di erent number of archetype Nevertheless, this e ect is reduced with an increasing num- buildings to the resulting nal energy demands, aggregated ber of archetype buildings, as single archetypes represent to di erent energy carriers, is illustrated in Figure 8. They attribute distributions on a more granular level. In conse- are validated against the nal energy demand provided by quence, the spatial di erences, e.g., of construction years, AGEB [2]. are better tted and intrinsic correlations of the input data The dominant energy carriers for the residential sector are represented with higher accuracy. are gas, oil and electricity with 268, 162 and 136 TWh/a This work assumes 200 archetype buildings as a su- per year [2]. The demand for renewable energy or district cient trade-o between accuracy and computational load, heating is secondary with 84 and 51 TWh/a per year. As since they already capture the main diversity of the en- Energy demand 2013 [TWh/a] AGEB ergy carriers and the statistical balancing e ects between when Southern Germany's energy demand was 9% above the buildings. the regional average, while northern Germany lies just 5% above the average. This highlights the importance of hav- 3.2. Impact of di erent weather years ing a spatially resolved building stock model, as a single location is not able to represent this variation suciently. The impact of di erent weather years from 2010 to 2015 on the energy demand of the buildings is illustrated in Figure 9 for 200 archetype buildings and validated again to the nal energy demand values provided by AGEB [2]. According to AGEB [2], the total residential energy de- mand varies from 743 TWh in 2010 at a maximum to 608 TWh for 2014 as a minimum. 0 20 1.11 1.17 0.92 0.95 1.00 1.02 2010 2011 2012 2013 2014 2015 TRY Avg. demand Change of Change of Change of 2010 to avg. 2011 to avg. 2012 to avg. [GWh/a/km ] Energy carrier Electricity Renewables 500 District heating Oil Gas Coal 0 1.05 1.09 0.88 0.90 0.92 0.98 1.0 1.2 Change of Change of Change of Change of 2013 to avg. 2014 to avg. 2015 to avg. TRY to avg. Figure 10: Spatial distribution of the nal energy demand, averaged Figure 9: Final energy demand for di erent weather years predicted for the considered years 2010 until 2015, and the relative regional with 200 archetype buildings and compared to the values reported deviation from the average value. by the AGEB [2]. In 2014, no signi cant di erences due to the geo-position For all di erent weather years, the systematic overes- are observed. Nevertheless, it becomes clear that the cities timation of gas demand and underestimation of district are less sensitive to the weather patterns (11.5% below the heating demand is observed, as already discussed in the average in 2014) than the rural areas (13.5% below the previous section. Nevertheless, the relative deviation dif- average in 2014). The reason is that the relative share fers between the years. While the total nal energy de- of energy demand for space heating to the overall energy mand ts well for 2010 with an underestimation of below demand is smaller in the cities than in the rural areas, re- 2%, the deviation increases in the year 2011 up to 7.4%. It ducing the relative impact of weather years on the total reduces again to 2.8% in the year 2013 while 2014 again has energy demand. a value of 5.8%. The di erences are mainly constituted by The analysis illustrates that the novel spatially resolved the di erent demands for the energy carriers that are used approach is able to identify local extreme weather pat- to supply the space heat, while the electricity demands terns. While it was only shown here for the aggregated remain almost constant for all periods. It seems that in annual demand, the model also predicts the temporal de- relatively mild weather years, the deviation is higher than mand of the energy carriers in all municipalities and can in colder weather years. A probable explanation of the be used for the identi cation of local peak demands that varying deviations could be an adaptive occupancy behav- are relevant for the infrastructure's design. ior, e.g., the ventilation rates could be reduced in colder winters, which is not taken in to account in the model. 4. Future energy supply scenario Figure 10 shows the spatial distribution of the nal energy demand averaged for all of the weather years con- In order to predict the change of the supply struc- sidered. It clusters in the cities as expected. Addition- ture in the future, the overall model is applied with 200 ally, the relative changes of the nal energy demand for archetype buildings for the techno-economic assumptions the di erent weather years are illustrated for the di erent in 2050, which are de ned in Appendix A.2. The choice, municipalities. The overall magnitudes of the di erences scale, and operation of the considered energy supply tech- align with the di erences shown in Figure 9. Neverthe- nologies are optimized together with the heating system less, it is clearly recognizable that di erent weather years and potential refurbishment measures, implying that the spatially impact the annual energy demand spatially dif- building owners have a technology adoption and operation ferently: While the year 2010 was generally a cold year, in approach that minimizes their energy cost and act as homo northern Germany the nal energy demand was 17% above economicus. We de ne this as Min Cost scenario. the average while in south-west Germany it was only 11% higher than the average. This is the opposite to 2013, Energy demand [TWh/a] AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. The results de ne the overall state that the residential the relatively small cost share of the eciency measures energy supply system is converging on if the assumed en- is that most measures are chosen for buildings that are ergy prices and the techno-economic assumptions for the anyway in the refurbishment cycle due to their lifetime. technologies arise. Besides the incentives included in the Therefore, the costs for the sole eciency measures are scenario, no additional ones are given by regulators. The relatively small because installation costs, such as scaf- demand for the use of electrical devices, hot water demand folding construction, are seperately considered for these and thermal comfort level are asumed to stay the same as buildings. the status quo described in Section 3 with the assumptions The overall results are aggregated from the optimal described in Appendix A.2. Existing technologies are as- system design of the di erent archetype buildings, whose sumed to get replaced until 2050, wherefore each building cost structure is illustrated in Figure 12. The total annual has a green eld optimization regarding the supply side. cost of the buildings is scaled by the number of house- holds in the buildings to show di erent sizes of buildings 4.1. Design, costs and operation on a similar scale. In order to expose patterns between the buildings, they are manually clustered into four groups The resulting total expenditures related to investments based on their resulting supply system. The Single Family in the supply structure or energetic refurbishment mea- Houses (SFHs) are manually di erentiated between those sures are visualized in Figure 11. The aggregated capaci- with and those without heat pumps, while the Multi Fam- ties and energy ows are shown in Appendix B. To real- ily Houses (MFHs) are distinguished between those with ize the technology portfolio, an overall investment of 382.3 and without fuel cells. billion Euro is needed. The largest share is photovoltaics, with a total investment of 104.6 billion Euro and a to- tal capacity of 133.4 GW. 89.7 TWh/a of the generated electricity is used for self-consumption, which is the main incentive to deploy photovoltaic. Solar thermal Photovoltaic The second highest investment is for heat pumps with Battery 88 billion Euro indicating that these are the main supplier Fuel cell Log supply of space heating. Fuel cells are the chosen exible co- Electric heater generation option and amount to 5.4% of the annual costs. Heat pump Heat storage The heat storage systems make up 1.9% of the annual costs Gas supply and have a total investment of 21.8 billion Euro. The Gas boiler Electricity grid investment in the batteries is signi cantly lower with 6.9 Roof billion Euro, amounting to 1% of the annual costs. The log Windows Walls wood supply for the replaces amounts to 2.1%, while the Control electric heaters have a minor share. District heating, oil 0 500 1000 1500 2000 2500 3000 3500 boilers and pellet boilers are not chosen in the solution, as Total annual cost per household [Euro/a/hh] they are not competitive in comparison to the heat pumps Figure 12: Cost composition of the di erent archetype buildings for or gas boilers. the Min Cost scenario in 2050. They are grouped by Single-Family House (SFH) with and without (wo.) heatpumps and Multi-Family 104.6 Houses (MFH) with and without (wo.) fuel cells. 88.0 53.6 In general, the only technology that is chosen for al- 31.1 26.7 21.8 most all of the buildings is rooftop photovoltaic. With the 19.5 12.8 9.7 6.9 6.0 1.6 predicted small cost of the photovoltaic panels and high 0.1 electricity price, these are in the cost-optimal solution for various scales but independent of the roof orientation of the building. Except for one SFH that has a completely self-sucient electricity supply, no other SFH has a fuel cell installed. Figure 11: Total investments into the di erent measures in the resi- Since the demand pro le of a single-family house is highly dential buildings for the Min Cost scenario in 2050. volatile, the achievable full load hours for a self-sucient electricity supply are too low for a fuel cell to become eco- Together, the energy-related refurbishment measures nomically feasible. Furthermore, the required capacities account for 97.2 billion Euro while more than half of them of the fuel cell would be small, increasing the speci c cost are determined by the walls. The occupancy control sys- due to a missing economy of scale. tems have a relatively high share with 12.8 billion Euro, Moreover, it is striking that the occupancy controllers followed by the windows and the roofs. One reason for are primarily installed in SFHs with gas boilers. The build- Investment [Bil. Euro] Control Walls Windows Roof Floor Gas boiler Heat storage Heat pump Electric heater Fuel cell Battery Photovoltaic Solar thermal MFH wo. MFH with SFH wo. SFH with fuel cell fuel cell heat pump heat pump ing cluster with gas boilers is dominated by compact build- load hours an investment into a gas boiler could be advan- ings where only a few rooms need to be equipped with the tageous. From a central infrastructure perspective, both thermostats, constituting small investment costs. More- options have an intrinsic economic issue, as they need the over, the heat capacity of those buildings is small and layout of an infrastructure that will be used in its maximal constitutes limited thermal inertia. This is bene cial for capacity for only a few hours per year. the occupancy controller, as the building can cool down and heat up faster in the case of vacant occupants. For 4.2. Changing electricity grid load large buildings with a high thermal mass, an occupancy The resulting electricity grid exchange, de ned by the controller only o ers limited bene t. electricity imported for the heat pump, the conventional All MFHs with a fuel cell have an additional heat pump electricity demand, and the photovoltaic feed-in is illus- installed. The cheap self-supply with electricity bene ts trated in Figure 14. For comparison purposes, the grid electrical heat generation. Some of the MFHs add a bat- exchange of the Reference scenario is shown as well. The tery system to increase the share of the photovoltaic and aggregated electricity load of the Reference scenario is CHP electricity that can then be self-consumed. dominated by the occupant activities in the morning and The di erent full load hours and capacities of the tech- evening. A small variation between winter and summer nologies in the di erent archetype buildings are shown in then appears. The overall load peaks in the evening hours Figure 13. The scale of the dots indicate how often the during winter with 36.4 GW. This aggregated load signi - archetype buildings are assigned in total in Germany. In cantly changes for the 2050 scenario, when during the sum- general, it can be seen that although photovoltaics are in- mer the load demand is reduced to values below 10 GW, stalled in all buildings, the achievable full load hours vary and also for the evening hours, while high daytime feed-in from 683 to 1025 depending on the roof orientation and rates of the photovoltaic occur with up to 43.1 GW, ex- location of the archetype building. ceeding the peak demand of the Reference scenario. The The highest full load hours are around 5000, and achieved impact of the photovoltaic gets reduced during the win- by the fuel cells. It is observed that a larger fuel cell capac- ter but still reduces the load at noon for most days. The ity correlates with higher achievable full load hours. This evening hours in winter are still the peak demand times mainly relates to the occupancy pro les: due to statistical with a load of up to 32.3 GW for the 2050 scenario, which balancing e ects, larger buildings have atter pro les that is in a similar in magnitude to the Reference scenario. can be covered by higher self-generation rates. Opposing e ects are observed for the peak generators, such as the Reference gas boiler with around 2000 full load hours and the electric heater with less than 1000 full load hours: the larger the 10 Min. cost installed capacities are, the smaller are the achievable full load hours. For the heat pump, no such e ect is observed. It is operated with between 3000 and 4000 full load hours 0 50 100 150 200 250 300 350 for small capacities as well as large ones. Day Figure 14: Aggregated grid exchange of the national residential Fuel cell 4000 Heat pump building stock for the Reference and the Min Cost scenario. Gas boiler As introduced previously, the technology installations Photovoltaic depend on the building type. Therefore, the changes in the Electric heater 500 grid load also vary spatially depending on the local build- ing topology, as illustrated in Figure 15. As expected, 0 1 2 10 10 10 the majority of the regions reduce their annual electricity Capacity [kW] demand with the help of self-generation by photovoltaics and fuel cells. Nevertheless, regional di erences are high: Figure 13: Full load hours and capacity of the installed technologies while urban areas are able to reduce their electricity de- in the di erent archetype buildings for the Min Cost scenario in 2050. The size of the scatter is related to the overall appearance of mand by 60%, some rural areas even increase it. The high the archetypes in Germany. photovoltaic installations in rural areas are not sucient to compensate for the increased electricity demand from the heat pumps. This e ect intensi es for the case of the The distribution of scales and full load hours indicates peak load, as almost no photovoltaic feed-in exists in the that the heat pumps signi cantly rely on a peak boiler, winter days, while the heat pumps are being operated in since their scaling to the maximal heat load would be more full load. Therefore, regions characterized by large SFHs expensive. Nevertheless, it is open as to which peak boiler double their peak load. This is di erent for the urban ar- is chosen in the model. For a few full load hours, the eas that even reduce their peak load because the fuel cells electric heater is more cost-e ective, while for many peak Full load Full load hours [h/a] hours [h/a] Hour Hour Residential grid load [GW] Solar thermal exceed the electrical capacity of the heat pumps and are Photovoltaic Battery synchronously operated. Equivalent regional trends are Fuel cell Bio methane observed for the feed-in: The rural areas feed up to 40% Pellet boiler Pellet supply of the original electricity demand into the grid, while the Log supply Electric heater urban areas have only small feed-in rates of 10%. Further, Heat pump Heat storage a gradient between north and south is recognizable due to District heating CHP the di erent solar irradiance. Gas supply Gas boiler 0 Electricity grid Ventilation Floor Roof Windows Walls Control Figure 16: Annual cost of the Min Cost scenario and the resulting aggregated system cost if the solution space is constrained. 100 0 100 100 0 100 0 15 30 45 Change of Change of Feed-in in ratio peak demand [%] annual demand [%] to Reference demand [%] they are used in small scales in the case that fossil gas is excluded. The amount of occupancy controllers is also Figure 15: Spatial change of the peak electricity demand and the reduced without fossil gas. The reason is that the heat change of the cumulative positive demand from the Reference sce- nario to the Min Cost scenario. Furthermore, the amount of elec- pump is intensively used during the day in order to uti- tricity feed-in to the grid in the Min Cost scenario is shown in ratio lize photovoltaic electricity while heating up the building. to the cumulative electricity demand in the Reference scenario. Nevertheless, the occupancy controller lowers the comfort temperature especially during the day when the occupants are absent. These two temporally opposing e ects reduce In summary, the results indicate that the change of the the value of an occupancy controller for the buildings sup- energy supply in the rural areas is more challenging with plied with heat pumps. The exclusion of fossil gas also respect to the electricity grid operation than the changes constitutes the smallest GHG footprint that gets reduced in the urban areas. Nevertheless, adaptions in the tari from 51.9 Mt/a in the Min Cost scenario to 19.9 Mt/a, as design could dampen this e ect. shown in Figure 17. 4.3. Value of analysis Electricity grid In order to evaluate the robustness of the Min Cost Bio methane Pellet supply scenario for 2050, 200 archetype buildings are again opti- Gas supply mized, but parts of the technologies are excluded or forced District heating into the solution space. Figure 16 illustrates the aggregated resulting annual cost composition of the di erent cases that were consid- ered for the analysis. Gas supply, fuel cell, photovoltaic, Figure 17: GHG footprint of the Min Cost scenario and the resulting heat pump and refurbishment measures are each excluded aggregated GHG footprints if the solution space is constrained. from the solution space, and the other cost minimal so- lutions are compared to the original Min Cost scenario with all technologies available. The increase in the total The structural changes to the Min Cost scenario are systems costs can be interpreted as the Value Of the inte- fairly small for the case that the fuel cell is excluded from gration of a certain technology. As an additional case, the the solution space. The net electricity import increases as full package of refurbishment measures are enforced for all in the previous scenario and compensates for the missing buildings that are in the refurbishment cycle in order to self-generation, but the photovoltaic capacities merely in- reach lower demands for space heating. crease from 133.4 GW to 142.3 GW, while the heat pump In case the fossil gas supply is excluded from the so- capacities remain in a similar magnitude. This is di erent lution space, the electricity purchase doubles, as no fuel to the previous case and indicates that the value of fur- cells for self-consumption are installed. Bio-methane or ther photovoltaic capacities is mainly correlated to higher another renewable fuel is too expensive in the considered heat pump capacities and not to smaller fuel cell capaci- scenario to replace the fossil gas in the fuel cells. Instead, ties. The battery capacities are reduced from 16.9 GWh to higher capacities of photovoltaics are integrated into the 12.8 GWh, although the photovoltaic capacity is increas- solution, aggregating up to 160.3 GW. Moreover, the ag- ing. This implies that their operation partially comple- gregated cost for heat pumps increases by 38% as their ments the fuel cell operation. share of the heat supply increases. While the Min Cost Signi cant shifts and cost increases are recognizable solution did not include district heating or pellet boilers, Min cost No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment GHG-footprint [Mt/a] Annual cost [Bil. Euro/a] Min cost No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment in the case that the photovoltaic is excluded, while the The reverse e ect occurs for the forced refurbishment electricity purchase only increases from 71.9 TWh/a to case: Installed heat pump capacities are reduced while 80.9 TWh/a, the gas import almost doubles from 172.5 TWh/a gas boiler capacities increase. The overall demand for gas to 285.7 TWh/a with the e ect of a GHG footprint of up and electricity is reduced as the space heat demand drops to 81.3 Mt/a. High gas boiler capacities compensate for to 209.1 TWh/a, in comparison to the 309.8 TWh/a in the reduction of the heat pump capacities from 60.4 to the Min Cost scenario and the 449 TWh/a in the Refer- 46.9 GW . This indicates the enforcing e ect between the ence scenario. Nevertheless, the demand reduction is not th heat pump and the photovoltaic supply, which is econom- able to compensate for the high cost of the refurbishment ically advantageous in the case of self-consumption with measures, resulting in an overall annual cost increase of photovoltaics is available. No battery capacities are in- 29.7%. In particular, ventilation systems with heat recov- stalled, supporting the statement that their main economic ery amount to almost half of the eciency measure costs. driver for installation is the photovoltaic, although they Ventilation systems do not bene t from the refurbishment are also partially used to increase the self-consumption cycle, as their integration cost in the building is mostly with fuel cell electricity. independent of any outside renovation measures. Also no- In the case that a heat pump is excluded from the so- ticeable is the fact that the amount of occupancy control lution space, the amount of gas increases by 113.2 TWh/a systems drops: If the heat demand is reduced anyway, while the electricity demand only gets reduced by 13.5 TWh/a. additional measures such as temporally reducing the inner In consequence, the GHG footprints increase by up to 76.0 air temperature have a minor e ect, making the occupancy Mt/a. Furthermore, the investment in refurbishment mea- controller economically unfavorable. The GHG footprint sures increases by 30%, dominated by more occupancy is only reduced to 45.1 Mt/a, which is much smaller than controller and more wall insulation and vice versa, indi- expected but explained by the switch of many buildings to cating that especially cheap heat produced by the heat gas boilers. This indicates a potential rebound e ect that pumps lowers the motivation to invest in eciency mea- might occur in the future, in that the reduced demand for sures. Also, the re wood supply increases from 21.2 TWh/a space heating lowers the economic incentive to invest in to 51.44 TWh/a, since it is a cheaper fuel than fossil gas ecient but expensive heat supply technologies. in the scenario. Remarkable is the reduced investment in It is striking that the aggregated cost gain is moderate fuel cells, cutting their capacity from 12.7 to 5.1 GW . for all considered cases, except for the forced refurbish- el This illustrates that major fuel cell capacities are built to ment case. This indicates that the prediction of the total supply the heat pumps with electricity. cost is robust and not sensitive to the available technolo- The exclusion of refurbishment measures from the so- gies in the future, although, this robustness only accounts lution space constitutes an increased investment in heat for an aggregated German-wide level, as illustrated in Fig- pumps and a reduced investment in gas boilers. This is ure 18. The gure shows the total cost increase and the surprising, as an enforcing e ect between the heat pump distribution of the cost increase of the single buildings. and refurbishment measures could be expected because the While the total cost in Germany only increases by 3.65% refurbishment measures decrease the required supply tem- for the case that no fossil gas supply is available, one of perature in the building, and vice versa, increasing the the archetype buildings has a 22.9% higher energy costs, eciency of heat pumps. Nevertheless, the economic ef- while some other buildings are not a ected at all, as they fects predominate, the heat pumps have higher investment were also not supplied with gas in the Min Cost scenario. costs than the gas boilers, while on the other hand the Similar e ects are observed for the other sensitivity analy- operational energy cost for the gas supply is higher. In ses: The sensitivities for single buildings are high, but the consequence, heat pumps are favored in the case of high cost of the aggregated result is robust. heat demands and their deployment increases for the case The impact on the electricity grid of the di erent cases of no refurbishment. is further illustrated in Figure 19. It shows the sorted grid load for the Reference scenario, the Min Cost scenario, Total in Germany and all related sensitivity analyses. The highest peak load Single archetypes occurs if the gas supply is completely excluded from the solution. No signi cant gas boiler capacities are able to satisfy the peak heat demand and no fuel cells can dimin- ish the additional electricity load of the heat pumps. In consequence, the peak load almost doubles to 55.9 GW in comparison to the Min Cost scenario with 32.3 GW. The second highest demand is reached if no fuel cell is included and the peak load increases by 14.6 GW rela- Figure 18: Annual cost increase for the case that certain technologies tive to the Min Cost scenario. This shows the importance are excluded or added to the solution space in the Min Cost for 2050. of decentralized exible electricity generation in order to The change is once shown for the total aggregated cost and once for the single archetype buildings. compensate for the increased electricity demand by the heat pumps. As is to be expected, the case without heat No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment Annual cost increase [%] pumps has the lowest peak load, with 25.7 GW. service sector including commercial buildings, or also rep- resentative fueling stations, whose detailed models could be upscaled to a nationwide perspective while respecting Reference Min cost the spatially varying conditions to supply them. 20 No Gas supply No Fuel cell No Photovoltaic 5.2. Buildings as economic entities No Heat pump No refurbishment A limitation is the current scope of the building model: Forced 40 refurbishment Mobility demand should be included in future since it has also a high impact on the grid load and provides further 0 2000 4000 6000 8000 Hours [h] exibility. Thereby, also further technology should be in- cluded, such as ground-source heat pumps or hydrogen Figure 19: Sorted grid load of the Min Cost scenario and the grid storage technologies. While those technologies can be eas- load if the solution space is constrained. ily added, the single archetype approach does not allow a good evaluation of district scale technologies since the entities are considered independently and occure in the The amount of photovoltaic feed-in is for all cases that municipalities in di erent combinations, wherefore a sys- include photovoltaics in a similar range. The maximal tematic evaluation, e.g. of heating networks, is hardly feed-in is reached with 55.4 GW for the exclusion of the gas possible. supply, although the single buildings have the constraint Supplementary, the sole nancial agent decision mak- to limit the feed-in to 50% of their maximal capacity. ing is known to perform well in rate of adoption and cu- mulative adoption but underestimates social and attitudi- 5. Limitations and outlook nal components in uencing the technology adoption [82]. In consequence, only a few economically dominant tech- As the previous sections showed the capability of the nologies are chosen in the Min Cost case, although even modeling approach, it has also some limitations. a higher diversity would exist in reality due to individu- ally varying information levels. As a case in point, pellet 5.1. Chosen archetypes boilers were down-selected, although those are considered As concluded by Corgnati et al. [49], the aggregation in di erent scenarios in the literature [15, 19]. This dom- and descriptions of the archetype buildings is highly de- inance is related to the scenario, and a consideration of pendent on the available data. The algorithm in this work di erent biomass prices could change this. Nevertheless, is tailored for the data structure of the German Census in reality an adoption of pellets would also be expected [76] that describes the statistical distributions of building without being economically competitive. Such non-cost parameters at the municipality level. The methodology is optimal adoption behaviors are better included in sim- transferable to other countries where similar data struc- ple adoption models such as Invert tool [83, 84] that in- ture exists, but in case that exact building samples are cludes a statistical randomness in the adoption process, available, it would still be recommended to use conven- but neglects, e.g., the temporal operation. Nevertheless, tional cluster algorithms for the aggregation. The tting the information basis for investment decisions is improving of archetype buildings to meet attribute distributions, as in wherefore the assumption in this work of a cost optimal this work, has the drawback that theoretical building con- investment stays reasonable. gurations are created that meet the distribution values but can signi cantly deviate from real building instances. 5.3. Further scenarios Further, while the set of archetype buildings is able to The good alignment with the reported nal energy de- respect the variety of households with di erent cumulative mand values of today makes the model further suitable to electricity demands due to their di erent appliance equip- develop transformation paths of the building stock until ment and di erent household sizes, the appliance equip- the year 2050. E.g., additional projected supply years in ment and adoption rate are not altered by the socioeco- 2020, 2030 and 2040 could be modeled. Such an approach nomic background of the households. Nevertheless, this could better consider the di erent lifetimes of the tech- is a signi cant descriptor to determine electricity demand nologies and respect the inertia in the adoption process of variation [80, 81] and probably also technology adoption the building owners. rates. The approach would allow this description, but the Thereby, also short term policy design could be better required data is not publicly available. supported. The resulting grid load for the 2050 scenario For more holistic analysis, the algorithm should be in this work is highly related to the at energy tari s con- transferred to other sectors to derive spatially-distributed sidered. Nevertheless, the peak and feed-in values can be sector speci c representatives. Thus, a cross-sectoral spa- controlled by the di erent tari structure, that could give tially resolved bottom-up model can be derived that re- incentives to shift consumption in time in order to atten spects the individual economic entities. Examples are the the load pro les. Therefore, upcoming works will apply Sorted grid load [GW] the model to di erent regulatory regimes and market en- Appendix A. Techno-economic assumptions vironments. A coupling with grid models is promising in The main input assumptions to parametrize the opti- order to determine an economically optimal market design. mization models are introduced in the following section, e.g. residential energy prices and eciencies of the di er- 6. Conclusions ent technologies. Section Appendix A.1 de nes the pa- rameters for the year 2015, while Appendix A.2 extends In this paper, a novel bottom-up model was introduced and adapts them to the year 2050. that is based on an aggregation of archetype buildings and a related optimization model to predict the a spatially- Appendix A.1. Assumptions for 2015 resolved technology adoption and operation. In order to achieve a valid comparison of today's res- The model approach allows the evaluation of the in u- idential energy supply to the changes that will occur in ence of regulatory decisions on energy cost, green house the future, a valid scenario framework is introduced that gas emissions, or grid load under the assumption of cost represents today's cost and operation parameters for the optimal behavior. The novelty is that it is able to analyze residential energy supply systems. the impact regulatory regimes and market environments The economic parameters considered for the supply for single buildings, simultaneously with a nationwide eco- technologies are illustrated in Table A.1, while their de- nomic perspective. tailed derivation is discussed in Kotzur [65]. The struc- As show case, the model was applied and validated ture of the investment costs is oriented around the cost for the residential building stock of Germany and a future model introduced in Lindberg et al. [37]. It di erentiates scenario frame for 2050. The main techno-economic con- between the xed investment costs that occur in the case clusions drawn from the future scenario are the following: of installation and the speci c investment costs that are added and related to the scale of the installations. • The key technologies for reducing the GHG emis- sions in the building stock are photovoltaic and heat- Table A.1: Assumed economic parameters of the energy supply tech- pumps that will signi cantly increase the seasonal nologies for the Reference scenario. variation of the residential electricity load, as their feed-in and demand do not temporally match, result- Technology CAPEX CAPEX OPEX Lifetime Source x speci c %CAPEX/a a ing in a doubling of the peak electricity load in the Gas boiler 2800 Euro 100 Euro/kW 1.5 20 [65] th winter hours in rural areas. Oil boiler 2800 Euro 100 Euro/kW 1.5 20 [65] th Pellet boiler 10000 Euro 300 Euro/kW 3.0 20 [65] th Heat pump 5000 Euro 600 Euro/kW 2.0 20 [65] th • The urban areas can compensate the increasing elec- Heat storage 800 Euro 1200 Euro/m 0.0 25 [65] Photovoltaic 1000 Euro 1400 Euro/kW 1.0 20 [65] el tricity demand by ecient co-generation units, e.g., IC CHP 15000 Euro 1000 Euro/kW 7.0 15 [65] el Solar thermal 4000 Euro 350 Euro/m 1.0 20 [65] in form of fuel cells, which cannot achieve sucient Electric heater 0 Euro 60 Euro/kW 2.0 30 [37] th economies of scale in single family houses in the rural areas. Although the model allows for the modeling of di erent • Signi cant amounts of photovoltaic electricity (for interest rates for di erent building types to take into ac- Germany up to 90 TWh/a) can be self-consumed count of the di erent investment behavior of the building while the majority is used for internal heat supply owners [85], it is here simpli ed to a single interest rate of applications. Batteries are hardly deployed, as the 3%, which lays between the 2% to 5% considered in the heat applications provide enough exibility for self- literature [32, 37, 47, 86, 87]. consumption. The energy and resource prices are illustrated in Ta- • Refurbishment measures are expensive and only cho- ble A.2. The majority of the prices are derived from the sen in cases when the building is in the cosmetic study Energieezienzstrategie Gebude [15, 88]. Their as- refurbishment cycle. Therefore, space heat demand sumptions de ne the basic scenario framework for this only decreases by 30% from 2015 to 2050 in the sce- thesis and rely themselves on the Energiereferenzprognose nario. Instead, a shift towards an ecient integrated [89]. The majority of the resource prices assumed in the energy supply system, e.g. combinations of fuel cells study align with the energy prices observed for 2016 [18, with heat pumps, is favored. 90]. Nevertheless, the assumed gas price overshoots the observed price of 2016 by more than 1 ct/kWh, and was adapted in this work to the values reported for 2016 by Acknowledgments the Bundesnetzagentur [90]. This work was supported by the Helmholtz Associa- The GHG footprint includes the emissions of the previ- tion under the Joint Initiative Energy System 2050 A ous conversion processes, such as in the extraction of fuels, Contribution of the Research Field Energy . Parts of this or the GHG emissions of the German power plant mix. article are derived from the doctoral thesis Future Grid Furthermore, the price structure is modi ed from a sole Load of the Residential Building Sector . energy price (Euro/kWh) structure to a combination of a 14 active at home. The night reduction temperature is set Table A.2: Assumed residential energy prices including taxes, levies, for all buildings to 18C. and network charges based on Energieezienzstrategie Gebude [15, 88, 89] while missing parameters are derived from Lindberg et al. [91], KWKG [92], EEG [93]. The gas price is corrected to the observed gas Appendix A.2. Assumptions for 2050 prices in 2016 [90]. The GHG footprint and primary energy factors The techno-economic assumptions for the future en- (PE) are taken from prognos [88]. FiT refers to Feed-in Tari . ergy supply through 2050 are introduced in the following Technology OPEX-var OPEX- x GHG PE Comment section. While many parameters are estimated to stay at a - Euro/kWh Euro/a kg/kWh kWh/kWh - similar magnitude as in the Reference case in Section Ap- Electricity 0.246 170 0.525 1.8 0.292 Euro/kWh supply for 3700 kWh/a pendix A.1, this section describes only the assumptions Gas supply 0.065 0 0.250 1.1 that are changing for the case of 2050. All prices and costs Oil supply 0.064 0 0.320 1.1 Pellet supply 0.060 0 0.014 0.2 are provided as real prices in 2015. Heat pump 0.190 70 0.525 1.8 tari While no major changes are expected for conventional FiT CHP -0.08 0 0.000 2.8 for less than 50 heat generators, further learning rates and cost reductions kWel FiT PV -0.108 0 0.000 1.8 are considered for photovoltaic and electrochemical tech- District 0.074 327 0.270 0.7 0.096 Euro/kWh heating for 15.000 kWh/a nologies, as shown in Table A.4. Their detailed derivation Log supply 0.050 0 0.000 0.2 and discussion is also performed in Kotzur [65]. Table A.4: Change and addition of economic parameters of the en- at price (Euro/a) and an energy price (Euro/kWh). This ergy supply technologies for the year 2050. is important because the savings due to self-consumption, Technology CAPEX CAPEX OPEX Lifetime Source e.g. of photovoltaic electricity, would be overestimated x speci c %CAPEX/a a with the sole energy price. Additionally, this structure Photovoltaic 1000 Euro 650 Euro/kW 1.0 20 [65] el Battery 1000 Euro 300 Euro/kWh 2.0 15 [65] respects that speci c wholesale prices decrease with larger Fuel cell 4000 Euro 1500 Euro/kW 3.0 15 [65] el energy consumptions rates [90]. The technical performance of the technologies is sum- marized in Table A.3. The eciencies are given for the The technical assumptions for 2050 are shown in Ta- Lower Heating Value (LHV) of gas, oil or pellets. The ble A.5. The eciency of the heat pumps is expected to electrical and thermal CHP eciencies are de ned for a increase further in the future [100], for which this work xed operation ratio and cannot be varied in between. assumes an increase of the quality grade to 0.45, which The values are chosen such that the di erent age struc- is the upper bound of today's systems [65]. The photo- tures of the technologies are respected, e.g. an eciency voltaic eciency is assumed to increase to a value of 30% is assumed for the gas boiler that refers to the eciency of [22]. Primarily, this impacts the space coverage on the condensing boilers, while for the oil boiler a lower eciency rooftop and increases the photovoltaic potential that can is considered that is related to older boiler technologies. be installed. The technical parameters of the batteries are derived from a prediction until 2050 [101], but some of today's residential storage systems already achieve similar Table A.3: Summary of the main technical parameters of the energy supply technologies eciencies [102]. Technology Eciency Comment and Reference Table A.5: Summary of the main technical parameters of the energy Gas boiler 0.96 Condensing boiler supply technologies for 2050. [94] Oil boiler 0.84 [95] Pellet boiler 0.9 [37] Technology Eciency Comment and Reference Heat pump dynamic [65] quality grade of 0.4 Heat pump dynamic [65] Heat storage 0.99 charge [37] quality grade of 0.45 0.99 discharge 0.6%/h self-discharge [96] Photovoltaic 0.3 average 2050 [22] Photovoltaic 0.15 based on Hanwha HSL 60 S [97] 2 with 3.5 m /kWp with 7 m /kWp Battery 0.95 charge [101] IC CHP 0.6 thermal [98] 0.95 discharge [101] 0.25 electric [98] Electric heater 0.98 [95] 0.01%/h self-discharge [101] Solar thermal dynamic [37] 0.5 kW/kWh capacity factor Fireplace 0.83 [95, 99] Fuel cell 0.33 thermal [65] 0.52 electric [65] The comfort temperature inside the buildings is as- sumed to have a value of 21C for when occupants are 15 The electrical eciency of the fuel cell is assumed to Table A.7: Techno-economic assumptions for the insulation measures be 52% and positions itself between the eciency that can of a single building. The two measure levels are derived from Schutz  be achieved from Solid Oxide Fuel Cell (SOFC) systems et al. [47] while the exact cost and lambda are taken from BMVBS [105].(* thickness equivalent. ** capital expenditures related only to and the eciency of the Proton Exchange Membrane Fuel energetic measures.) Cells (PEMFC), as discussed in detail in Kotzur [65]. A fully exible operation is assumed for the year 2050. The Component Measure Thickness* Lambda CAPEX CAPEX energy ** 2 2 eciencies are considered to be the same for natural gas, - m W/m/K Euro/m Euro/m biogas or hydrogen as potential alternative fuels [103]. Wall Base 0.15 0.035 124.0 51.5 Future 0.22 0.035 140.9 68.5 The energy prices for 2050 are shown in Table A.6 Roof Base 0.24 0.035 237.6 53.0 and also rely on the Energieezienzstrategie Gebude [15, Future 0.36 0.035 270.0 79.6 88] and Energiereferenzprognose [89]. The Energiereferen- Floor Base 0.08 0.035 51.7 - zprognose considers a carbon price of 76 Euro/ton for the year 2050, which, e.g., increases the gas price by 1.9 ct/kWh. been refurbished anyway, as discussed in Kotzur [65]. The costs relate to the exterior surface area of the building Table A.6: Assumed energy prices, GHG footprints and primary component. Two levels of potential insulation measures energy factors (PE) based on the Energieezienzstrategie Gebude are considered and di er by the thickness of the insulation [15, 88, 89] for 2050. layer, and are referred to as Base and Future. Technology OPEX-var OPEX- x GHG PE Comment The cost for envelope refurbishment measures di ers - Euro/kWh Euro/a kg/kWh kWh/kWh - between buildings that are in the refurbishment cycle and Electricity 0.220 170 0.122 0.4 0.266 Euro/kWh buildings that are not, as discussed in detail in Kotzur [65]. supply for 3700 kWh/a Gas supply 0.096 0 0.250 1.1 The costs of replacing the windows and changing the Bio-methane 0.138 0 0.014 0.2 Oil supply 0.124 0 0.320 1.1 solar and thermal transmittance of the di erent window Pellet supply 0.080 0 0.014 0.2 types are shown in Table A.8 and rely on the BMVBS HP Tarif 0.190 70 0.122 0.4 FiTCHP -0.010 0 0.000 0.4 [105] as well. The costs are speci c to the window area of FiTPV -0.010 0 0.000 0.4 District 0.085 327 0.144 0.5 0.107 Euro/kWh the building. Again, a di erentiation is made between the heating for 15000 kWh/a Base and Future levels. Log supply 0.065 0 0.000 0.2 Table A.8: Techno-economic assumptions for the windows. The transmittance are based on [47] and the cost based on [105] Furthermore, a bio-methane purchase is integrated with a price of 13.8 ct/kWh, which can be either a synthetic Measure Solar transmittance Thermal transmittance CAPEX gas or biogas. As no sucient predictions for bio-methane 2 2 - W/m /K Euro/m prices in 2050 are te be found, its price is derived from the Base 0.575 1.1 313 production cost of bio-methane for the feed-in into the gas Future 0.5 0.7 361.5 grid of 7.5 ct/kWh in 2013 [104], plus the surcharge for grid fees, tax etc. This surcharge is considered to be 6.3 ct/kWh, which is the di erence between the gas market All envelope measures have a lifetime of 40 years with price of 3.3 ct/kWh and the residential gas price of 9.6 zero operational costs. ct/kWh in 2050 [89]. All in all, it results in a price of 13.8 In addition to the conventional refurbishment measures ct/kWh for the bio-methane, which is signi cantly above at the envelope of the building, a heat recovery for the the fossil gas price. ventilation is assumed with a speci c investment of 65 No values for future feed-in tari s were identi ed. There- Euro/m per living area, a lifetime of 25 years and opera- fore, the feed-in is only marginally subsidized, as it is tional cost of 4% per year according to the BMVBS [105] highly dependent on the future market environment. A as a ratio to the original investment. If integrated, 80% of marginal value of 0.01 eur/kWh is chosen in order to guar- the heat losses due to ventilation would be recovered. antee that photovoltaic generation is not curtailed and is Lastly, an occupancy controller can be installed that instead fed-in to the grid. reduces the comfort temperature in case of vacant occu- The cost and energetic impact of the refurbishment pants [65]. Based on the cost of Controme [106], they measures for the opaque building envelope are shown in are assumed with a xed investment of 1000 Euro for the Table A.7. central system controller and 3 Euro/m per living area All measures are additional layers to the existing en- for the di erent thermostats in the rooms, including their velope of the building. The costs are average values taken installation costs. A lifetime of 15 years is assumed. from a survey about subsided refurbishment measures in Germany [105]. They di er between the entire CAPEX of a refurbishment measure and the sole additional CAPEX of energy eciency measures if the building would have 16 17.6 53.9 40.0 103.7 4.7 5.2 69.2 Heat node to Space heat Appendix B. Supplementary results The composition of the total annual residential energy Heat pump to Heat node cost for the Min Cost scenario in 2050 scenario are shown Gas boiler to Heat node in Figure B.20. Floor Electric heater to Heat node Electricity grid Roof Windows 28.4% Fire place to Heat node Walls Control 3.6% Fuel cell to Heat node Total annual Gas boiler 3.4% expenditures: Photovoltaic 0 12.4% 65.0 bil. Euro 0 50 100 150 200 250 300 350 Day Annual expenditures per household: 1603.0 Euro Battery 5.4% Figure B.22: Heat ows of the relevant heat generators to the heat 25.5% Fuel cell 2.1% node and the connected demand for space heating for Germany in Gas supply the Min Cost scenario in 2050. Log supply 11.8% Electric heater Heat pump Heat storage Table B.9: Aggregated energy ows [TWh/a] between the di erent Figure B.20: Composition of the total annual costs over the whole technologies for Value of analysis. of Germany for the Min Cost scenario in 2050. Min No No No No No Forced Cost Gas supply Fuel cell Photovoltaic Heat pump refurbishment refurbishment AC Node to Battery 5.2 2.0 3.7 0.0 5.1 6.1 5.2 AC Node to Building 112.8 112.8 112.8 112.8 112.8 112.8 112.8 AC Node to Electric heater 40.8 56.1 43.3 4.6 48.1 45.0 40.0 The resulting annual energy ows between the di erent AC Node to Heat pump 34.3 26.9 16.2 24.6 0.0 50.1 21.0 AC Node to Hot water 6.8 6.8 6.8 6.8 6.8 6.8 6.8 technologies for the Min Cost scenario in 2050 are illus- Battery to AC Node 4.7 1.8 3.3 0.0 4.6 5.5 4.7 CHP to AC Node 0.0 0.0 0.2 0.0 0.0 0.0 0.0 Cool supply to Building 19.8 21.4 19.6 19.0 18.5 29.0 31.6 trated in Figure B.21 for the aggregated level of the whole Electricity supply to AC Node 51.5 100.5 84.7 63.5 57.7 47.3 52.7 Fuel cell to AC Node 53.9 0.0 0.0 85.3 20.1 69.1 45.7 of Germany. Gas supply to CHP 0.0 0.0 0.7 0.0 0.0 0.0 0.0 Gas supply to Fuel cell 103.7 0.0 0.0 164.1 38.6 132.9 87.9 Gas supply to Gas boiler 68.7 0.0 91.2 121.6 226.6 26.8 65.7 HP Tarif to Heat pump 20.4 46.2 40.2 17.4 0.0 42.1 10.5 Log supply to Fire place 21.2 19.4 23.0 23.6 51.4 21.3 28.8 21.2 Pellet supply to Pellet boiler 0.0 10.3 0.0 0.0 3.1 0.0 0.0 Log supply Fire place District heating Renewable gas to Gas boiler 0.0 8.7 0.0 0.0 0.0 0.0 0.0 Heat pump to Building 225.1 296.5 231.0 169.5 0.0 363.9 126.7 CHP to HNode 0.0 0.0 0.4 0.0 0.0 0.0 0.0 District heating to HNode 0.0 7.9 0.8 0.0 15.0 0.0 0.0 Electric heater to HNode 40.0 55.0 42.5 4.5 47.1 44.1 39.2 Pellet supply Pellet boiler Solar thermal Fire place to HNode 17.6 16.1 19.1 19.6 42.7 17.7 23.9 Fuel cell to HNode 34.2 0.0 0.0 54.2 12.7 43.9 29.0 Gas boiler to HNode 66.0 8.3 87.6 116.7 217.5 25.7 63.1 HNode to Building 84.7 23.4 78.1 123.3 265.8 57.4 82.4 Oil supply Oil boiler Heat storage HNode to Hot water 69.2 69.2 69.2 69.2 69.2 69.2 69.2 Pellet boiler to HNode 0.0 9.3 0.0 0.0 2.8 0.0 0.0 68.7 66.0 309.8 HNode to Heat storage 54.5 41.8 38.3 77.2 52.2 59.6 52.9 Heat storage to HNode 50.5 37.8 35.2 74.5 49.2 54.8 49.3 Gas supply Gas boiler Space heat Photovoltaic to AC Node 89.7 102.4 94.7 0.0 90.4 98.9 82.7 Photovoltaic to FiTPV 24.1 33.9 26.4 0.0 19.2 27.3 20.7 Solar thermal to HNode 0.1 0.0 0.0 0.1 0.1 0.0 0.0 Renewable gas Fuel cell CHP Heat pump Electric heater Hot water FiTCHP HP Tarif 51.5 112.8 References Electricity supply Electricity 24.1 [1] IEA, CO2 Emissions from Fuel Combustion - 2015 Edition, FiTPV Photovoltaic Battery Technical Report, International Energy Agency, 2015. [2] AGEB, Auswertungstabellen zur energiebilanz deutschland Figure B.21: Annual energy ows in TWh between the di erent 1990 bis 2016, 2017. technologies aggregated for the whole of Germany for the Min Cost [3] RWI, Erstellung der Anwendungsbilanzen 2015 und 2016 fur  scenario den Sektor der Privaten Haushalte und den Verkehrssektor in Deutschland, Technical Report, RWI Leibniz-Institut fur  Wirtschaftsforschung e.V., 2017. [4] BMWi, Zahlen und Fakten Energiedaten, Technical Report, The aggregated temporal operation of heat technolo- Bundesministerium fur  Wirtschaft und Energie, 2016. gies in the Min Cost scenario in 2050 are illustrated in [5] UN, Paris agreement - cop21, 2015. Figure B.22. [6] S. Solomon, G.-K. Plattner, R. Knutti, P. Friedlingstein, Irre- versible climate change due to carbon dioxide emissions, Pro- The aggregated annual energy ows between the di er- ceedings of the national academy of sciences 106 (2009) 1704{ ent considered system components are listed in Table B.9 for the di erent scenarios in 2050. [7] EU, 2010. URL: https://eur-lex.europa.eu/ Table B.10 shows the aggregated installed capacities legal-content/en/TXT/?uri=CELEX:32010L0031. [8] EU, 2012. URL: http://eur-lex.europa.eu/legal-content/ for all residential buildings for the di erent sensitivity cases. EN/TXT/PDF/?uri=CELEX:32012L0027&from=DE. [9] REHVA, How to de ne nearly net zero energy build- ings nZEB, Technical Report, 2011. URL: https: 20.4 6.8 50.5 54.5 34.2 40.8 0.1 89.7 34.3 225.1 Hour Hour Hour Hour Hour Hour GW GW GW GW GW GW [23] B. Nykvist, M. Nilsson, Rapidly falling costs of battery packs Table B.10: Aggregated installed capacities of the di erent technolo- for electricvehicles, Nature Climate Change 5 (2015) 329{332. gies for the Value of analysis. doi:10.1038/nclimate2564. [24] C. Breyer, A. Gerlach, Global overview on grid-parity, Min No No No No No Forced Cost Gas supply Fuel cell Photovoltaic Heat pump refurbishment refurbishment Progress in Photovoltaics: Research and Applications 21 Gas boiler [GW ] 36.6 5.1 47.6 83.9 111.9 14.7 36.3 th (2013) 121{136. doi:10.1002/pip.1254. Oil boiler [GWth] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 CHP [GW ] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 el [25] K. Mainzer, K. Fath, R. McKenna, J. Stengel, W. Ficht- District heating [GW ] 0.0 7.5 0.9 0.0 13.1 0.0 0.0 th Heat storage [GWh ] 215.6 291.4 216.6 63.4 173.9 265.4 188.6 ner, F. Schultmann, A high-resolution determination of the th Heat pump [GWth] 60.4 83.1 60.4 47.0 0.0 95.5 34.6 technical potential for residential-roof-mounted photovoltaic Electric heater [GW ] 99.7 97.1 90.3 83.6 75.4 121.0 84.4 th Solar thermal [GW ] 0.2 0.0 0.0 0.2 0.2 0.0 0.0 th systems in Germany, Solar Energy 105 (2014) 715{731. Pellet boiler [GW ] 0.0 2.4 0.0 0.0 0.8 0.0 0.0 th Battery [GWhel] 16.9 7.1 12.8 0.0 17.1 20.0 17.4 doi:10.1016/j.solener.2014.04.015. Fuel cell [GW ] 12.7 0.0 0.0 14.5 5.1 16.9 10.7 el Photovoltaic [GW ] 133.4 160.3 142.3 0.0 127.8 147.6 121.0 el [26] ISI, Netzentwicklungsplan Strom - Entwicklung der re- gionalen Stromnachfrage und Lastpro le, Technical Report, Fraunhofer-Institut fur  System- und Innovationsforschung, [27] M. G. Beer, Regionalisiertes Energiemodell zur Analyse der //www.rehva.eu/fileadmin/hvac-dictio/03-2011/How_ exiblen Betriebsweise von Kraft-W arme-Kopplungsanlagen, to_define_nearly_net_zero_energy_buildings_nZEB.pdf. Thesis, 2012. [10] A. J. Marszal, P. Heiselberg, J. S. Bourrelle, E. Musall, [28] P. Seljom, K. B. Lindberg, A. Tomasgard, G. Doorman, K. Voss, I. Sartori, A. Napolitano, Zero Energy Building {A re- I. Sartori, The impact of Zero Energy Buildings on the view of de nitions and calculation methodologies, Energy and Scandinavian energy system, Energy 118 (2017) 284{296. Buildings 43 (2011) 971{979. doi:10.1016/j.enbuild.2010. doi:10.1016/j.energy.2016.12.008. 12.022. [29] P. Balcombe, D. Rigby, A. Azapagic, Investigating the im- [11] J. OCallaghan, S. Mezger, M. Coughlin, K. Isles, A. Welsh, portance of motivations and barriers related to microgenera- C. K. Lyttle, Utility of the future - A customer-led shift in the tion uptake in the UK, Applied Energy 130 (2014) 403{418. electricity sector, Technical Report, 2014. doi:10.1016/j.apenergy.2014.05.047. [12] S. Agnew, P. Dargusch, E ect of residential solar and stor- [30] M. Achtnicht, R. Madlener, Factors in uencing German house age on centralized electricity supply systems, Nature Climate owners'preferences on energy retro ts, Energy Policy 68 (2014) Change 5 (2015) 315{318. doi:10.1038/nclimate2523. 254{263. doi:10.1016/j.enpol.2014.01.006. [13] W. Rickerson, T. Couture, G. Barbose, D. Jacobs, G. Parkin- [31] C. Milan, C. Bojesen, M. P. Nielsen, A cost optimization model son, E. Chessin, A. Belden, H. Wilson, H. Barrett, Residen- for 100% renewable residential energy supply systems, Energy tial prosumers - Drivers and policy options, Technical Report, 48 (2012) 118{127. doi:10.1016/j.energy.2012.05.034. [32] D. Lauinger, P. Caliandro, J. Van herle, D. Kuhn, A lin- [14] R. McKenna, E. Merkel, D. Fehrenbach, S. Mehne, W. Ficht- ear programming approach to the optimization of residential ner, Energy eciency in the German residential sector: A energy systems, Journal of Energy Storage 7 (2016) 24{37. bottom-up building-stock-model-based analysis in the context doi:10.1016/j.est.2016.04.009. of energy-political targets, Building and Environment 62 [33] A. Ashouri, S. S. Fux, M. J. Benz, L. Guzzella, Optimal design (2013) 77{88. doi:10.1016/j.buildenv.2013.01.002. and operation of building services using mixed-integer linear [15] BMWi, Energieezienzstrategie Geb aude, Technical Report, programming techniques, Energy 59 (2013) 365{376. doi:10. 1016/j.energy.2013.06.053. [16] N. Diefenbach, T. Loga, B. Stein, Reaching the climate pro- [34] E. D. Mehleri, H. Sarimveis, N. C. Markatos, L. G. Papageor- tection targets for the heat supply of the German residential giou, Optimal design and operation of distributed energy sys- building stock: How and how fast?, Energy and Buildings 132 tems: Application to Greek residential sector, Renewable En- (2016) 53{73. doi:10.1016/j.enbuild.2016.06.095. ergy 51 (2013) 331{342. doi:10.1016/j.renene.2012.09.009. [17] Beuth, Ableitung eines Korridors fur  den Ausbau der erneuer- [35] S. Fazlollahi, P. Mandel, G. Becker, F. Mar echal, Methods baren W arme im Geb audebereich, Technical Report, Beuth for multi-objective investment and operating optimization of Hoschule fur  Technik, ifeu - Institut fur  Energie- und Umwelt- complex energy systems, Energy 45 (2012) 12{22. doi:10.1016/ forsschung Heidelberg GmbH, 2017. j.energy.2012.02.046. [18] BMWi, Energiedaten - Tabelle 30: Interna- [36] H. Harb, J.-N. Paprott, P. Matthes, T. Schutz,  R. Streblow, tionaler Preisvergleich, Technical Report, 2018. D. Muller,  Decentralized scheduling strategy of heating sys- URL: https://www.bmwi.de/Redaktion/DE/Binaer/ tems for balancing the residual load, Building and Environ- Energiedaten/Energiepreise-und-Energiekosten/ ment 86 (2015) 132{140. doi:10.1016/j.buildenv.2014.12. energiedaten-energiepreise-5-xls.xls?__blob= publicationFile&v=17. [37] K. B. Lindberg, G. Doorman, D. Fischer, M. Korp as, [19] IWES, Interaktion EE-Strom, W arme und Verkehr, Tech- A. Anestad, I. Sartori, Methodology for optimal energy sys- nical Report, Fraunhofer-Institut fur  Windenergie und tem design of Zero Energy Buildings using mixed-integer lin- Energiesystemtechnik (Fraunhofer IWES), 2015. URL: ear programming, Energy and Buildings 127 (2016) 194{205. https://www.iee.fraunhofer.de/content/dam/iwes-neu/ doi:10.1016/j.enbuild.2016.05.039. energiesystemtechnik/de/Dokumente/Veroeffentlichungen/ [38] G. Cardoso, M. Stadler, S. Mashayekh, E. Hartvigsson, The 2015/Interaktion_EEStrom_Waerme_Verkehr_Endbericht. impact of ancillary services in optimal DER investment deci- pdf. sions, Energy 130 (2017) 99{112. doi:10.1016/j.energy.2017. [20] UBA, Datenbasis zur Bewertung von Energieezienzmanah- 04.124. men in der Zeitreihe 2005 {2014, Technical Report, Umwelt- [39] T. Schutz,  M. H. Schraven, S. Remy, J. Granacher, bundesamt, 2017. D. Kemetmuller,  M. Fuchs, D. Muller,  Optimal design of [21] H. Lund, A. N. Andersen, P. A. stergaard, B. V. Mathiesen, energy conversion units for residential buildings considering D. Connolly, From electricity smart grids to smart energy sys- German market conditions, Energy 139 (2017) 895{915. tems {A market operation based approach and understanding, doi:10.1016/j.energy.2017.08.024. Energy 42 (2012) 96{102. doi:10.1016/j.energy.2012.04.003. [40] M. Hamdy, A. Hasan, K. Siren, A multi-stage optimization [22] ISE, Current and Future Cost of Photovoltaics. Long-term Sce- method for cost-optimal and nearly-zero-energy building solu- narios for Market Development, System Prices and LCOE of tions in line with the EPBD-recast 2010, Energy and Buildings Utility-Scale PV Systems, Technical Report, Study on behalf 56 (2013) 189{203. doi:10.1016/j.enbuild.2012.08.023. of Agora Energiewende, 2015. 18 [41] R. Evins, Multi-level optimization of building design, energy 648{662. doi:10.1016/j.compchemeng.2013.10.016. system sizing and operation, Energy 90 (2015) 1775{1789. [60] J. A. Fonseca, A. Schlueter, Integrated model for characteriza- doi:10.1016/j.energy.2015.07.007. tion of spatiotemporal building energy consumption patterns [42] M. Stadler, G. Cardoso, S. Mashayekh, T. Forget, N. DeForest, in neighborhoods and city districts, Applied Energy 142 (2015) A. Agarwal, A. Sch onbein, Value streams in microgrids: A 247{265. doi:10.1016/j.apenergy.2014.12.068. literature review, Applied Energy 162 (2016) 980{989. doi:10. [61] E. McKenna, M. Thomson, High-resolution stochastic inte- 1016/j.apenergy.2015.10.081. grated thermal{electrical domestic demand model, Applied [43] X. Han, H. Zhang, X. Yu, L. Wang, Economic evaluation of Energy 165 (2016) 445{461. doi:10.1016/j.apenergy.2015. grid-connected micro-grid system with photovoltaic and en- 12.089. ergy storage under di erent investment and nancing models, [62] I. Richardson, M. Thomson, D. In eld, C. Cli ord, Domes- Applied Energy 184 (2016) 103{118. doi:10.1016/j.apenergy. tic electricity use: A high-resolution energy demand model, 2016.10.008. Energy and Buildings 42 (2010) 1878{1887. doi:10.1016/j. [44] R. Streblow, K. Ansorge, Genetischer Algorithmuszur kombi- enbuild.2010.05.023. natorischen Optimierung von Geb audehulle  und Anlagentech- [63] I. Richardson, M. Thomson, D. In eld, A. Delahunty, Domes- nik, Technical Report, 2017. tic lighting: A high-resolution energy demand model, Energy [45] R. Wu, G. Mavromatidis, K. Orehounig, J. Carmeliet, Multi- and Buildings 41 (2009) 781{789. doi:10.1016/j.enbuild. objective optimisation of energy systems and building envelope 2009.02.010. retro t in a residential community, Applied Energy 190 (2017) [64] I. Richardson, M. Thomson, D. In eld, A high-resolution do- 634{649. doi:10.1016/j.apenergy.2016.12.161. mestic building occupancy model for energy demand simula- [46] M. Stadler, M. Groissb ock, G. Cardoso, C. Marnay, Optimiz- tions, Energy and Buildings 40 (2008) 1560{1566. doi:10.1016/ ing Distributed Energy Resources and building retro ts with j.enbuild.2008.02.006. the strategic DER-CAModel, Applied Energy 132 (2014) 557{ [65] L. Kotzur, Future Grid Load of the Residential Building 567. doi:10.1016/j.apenergy.2014.07.041. Sector, Thesis, 2018. URL: http://juser.fz-juelich.de/ [47] T. Schutz,  L. Schi er, H. Harb, M. Fuchs, D. Muller,  Optimal record/858675. design of energy conversion units and envelopes for residential [66] D. EN ISO, Thermal performance and energy use in the build building retro ts using a comprehensive MILP model, Ap- environment (ISO 13790:2008) (2008). plied Energy 185 (2017) 1{15. doi:10.1016/j.apenergy.2016. [67] R. W. Andrews, J. S. Stein, C. Hansen, D. Riley, Introduction 10.049. to the Open Source PV-LIB for Python Photovoltaic System [48] M. Kavgic, A. Mavrogianni, D. Mumovic, A. Summer eld, Modeling Package (2014). Z. Stevanovic, M. Djurovic-Petrovic, A review of bottom-up [68] DIN, Din en 12831 - heating systems and water based cooling building stock models for energy consumption in the residen- systems in buildings method for calculation of the design heat tial sector, Building and Environment 45 (2010) 1683{1697. load, 2014. doi:10.1016/j.buildenv.2010.01.021. [69] DWD, Testreferenzjahre (TRY) (2012). [49] S. P. Corgnati, E. Fabrizio, M. Filippi, V. Monetti, Reference [70] C. Bollmeyer, J. D. Keller, C. Ohlwein, S. Wahl, S. Crewell, buildings for cost optimal analysis: Method of de nition and P. Friederichs, A. Hense, J. Keune, S. Kneifel, I. Pscheidt, application, Applied Energy 102 (2013) 983{993. doi:10.1016/ S. Redl, S. Steinke, Towards a high-resolution regional re- j.apenergy.2012.06.001. analysis for the European CORDEX domain, Quarterly Jour- [50] E. Mata, A. Sasic Kalagasidis, F. Johnsson, Building-stock nal of the Royal Meteorological Society 141 (2015) 1{15. aggregation through archetype buildings: France, Germany, doi:10.1002/qj.2486. Spain and the UK, Building and Environment 81 (2014) 270{ [71] L. Welder, D. S. Ryberg, L. Kotzur, T. Grube, M. Robinius, 282. doi:10.1016/j.buildenv.2014.06.013. D. Stolten, Spatio-temporal optimization of a future energy [51] IWU, Deutsche Gebaeudetypologie. Systematik und Daten- system for power-to-hydrogen applications in Germany, En- saetze, Technical Report, Institut Wohnen und Umwelt ergy 158 (2018) 1130{1149. doi:10.1016/j.energy.2018.05. (IWU), 2005. URL: http://www.iwu.de/fileadmin/user_ 059. upload/dateien/energie/klima_altbau/Gebaeudetypologie_ [72] L. Kotzur, P. Markewitz, M. Robinius, D. Stolten, Impact of Deutschland.pdf. di erent time series aggregation methods on optimal energy [52] IWU, Datenbasis Geb audebestand - Datenerhebung zur en- system design, Renewable Energy 117 (2018) 474{487. doi:10. ergetischen Qualit at und zu den Modernisierungstrends im 1016/j.renene.2017.10.017. deutschen Wohngeb audebestand, Technical Report, Institut [73] L. Kotzur, P. Markewitz, M. Robinius, D. Stolten, Time se- Wohnen und Umwelt Bremer Energie Institut, 2010. ries aggregation for energy system design: Modeling seasonal [53] EPISCOPE, 2016. URL: http://episcope.eu/iee-project/ storage, Applied Energy 213 (2018) 123{135. doi:10.1016/j. episcope/. apenergy.2018.01.023. [54] R. Hendron, C. Engebrecht, Building America Research [74] T. Kannengieer, M. Ho mann, L. Kotzur, P. Stenzel, Benchmark De nition, Technical Report, National Renewable P. Markewitz, F. Schutz  , K. Peters, S. Nykamp, D. Stolten, Energy Laboratory, 2010. URL: https://www.nrel.gov/docs/ M. Robinius, Reducing computational load for mixed inte- fy10osti/47246.pdf. ger linear programming: An example for a district and an [55] P. Torcellini, M. Deru, B. Grith, K. Benne, Doe commercial island energy system (2019). URL: https://www.preprints. building benchmark models, 2008. org/manuscript/201905.0116/download/final_file. [56] G. DallO, A. Galante, G. Pasetti, A methodology for eval- [75] B. Bahl, A. Kump  el, H. Seele, M. Lampe, A. Bardow, Time- uating the potential energy savings of retro tting residential series aggregation for synthesis problems by bounding error in building stocks, Sustainable Cities and Society 4 (2012) 12{21. the objective function, Energy 135 (2017) 900{912. doi:10. doi:10.1016/j.scs.2012.01.004. 1016/j.energy.2017.06.082. [57] C. Cerezo, J. Sokol, C. Reinhart, A. Al-Mumin, 2015. URL: [76] S. Bundesamt, 2011. URL: https://www.zensus2011.de/DE/ http://www.ibpsa.org/proceedings/BS2015/p2435.pdf. Home/home_node.html. [58] J. Sokol, C. Cerezo Davila, C. F. Reinhart, Validation of [77] Z. Huang, Extensions to the k-means algorithm for clus- a Bayesian-based method for de ning residential archetypes tering large data sets with categorical values, 1998. in urban building energy models, Energy and Buildings 134 URL: https://link.springer.com/content/pdf/10.1023/A: (2017) 11{24. doi:10.1016/j.enbuild.2016.10.050. 1009769707641.pdf. [59] S. Fazlollahi, G. Becker, F. Mar echal, Multi-objectives, multi- [78] S. Lloyd, Least squares quantization in PCM, IEEE Transac- period optimization of district energy systems: II-Daily ther- tions on Information Theory 28 (1982) 129{137. doi:10.1109/ mal storage, Computers & Chemical Engineering 71 (2014) TIT.1982.1056489. 19 [79] A. K. Jain, Data clustering: 50 years beyond K-means, Pat- Procedure to Determine Coecients for the Sandia Array Per- tern Recognition Letters 31 (2010) 651{666. doi:10.1016/j. formance Model, Technical Report, Sandia National Labora- patrec.2009.09.011. tories, 2016. [80] A. Druckman, T. Jackson, Household energy consumption in [98] ASUE, BHKW-Kenndaten 2014/2015 - Module, Anbieter, the UK: A highly geographically and socio-economically dis- Kosten, Technical Report, Arbeitsgemeinschaft fur  sparsamen aggregated model, Energy Policy 36 (2008) 3177{3192. und umweltfreundlichen Energieverbrauch e.V., 2015. [81] A. Elnakat, J. D. Gomez, N. Booth, A zip code study of so- [99] Olsberg, Produktdatenblatt Olsberg Palena Cmpact, Tech- cioeconomic, demographic, and household gendered in uence nical Report, 2018. URL: https://www.olsberg.com/ on the residential energy sector, Energy Reports 2 (2016) 21{ 578912-wee-de-wAssets/docs/kaminoefen/palena-compact/ 27. doi:10.1016/j.egyr.2016.01.003. anleitungen-zertifikate-ersatzteillisten/ [82] S. A. Robinson, V. Rai, Determinants of spatio-temporal pat- Produktdatenblatt_Palena-Compact_23-575_ohne-OEC.pdf. terns of energy technology adoption: An agent-based modeling [100] H. Willem, Y. Lin, A. Lekov, Review of energy eciency and approach, Applied Energy 151 (2015) 273{284. doi:10.1016/ system performance of residential heat pump water heaters, j.apenergy.2015.04.071. Energy and Buildings 143 (2017) 191{201. doi:10.1016/j. [83] L. Kranzl, M. Hummel, A. Muller,  J. Steinbach, Renewable enbuild.2017.02.023. heating: Perspectives and the impact of policy instruments, [101] P. Elsner, D. U. Sauer, Energiespeicher - Technologiesteckbrief Energy Policy 59 (2013) 44{58. doi:10.1016/j.enpol.2013. zur Analyse "Flexibilit atskonzepte fur  die Stromversorgung 03.050. 2050", Technical Report, 2015. [84] A. Mul  ler, Energy Demand Assessment for Space Condition- [102] K. Kairies, D. Haberschusz, J. v. Ouwerkerk, J. Strebel, ing and Domestic Hot Water: A Case Study for the Austrian O. Wessels, D. Magnor, J. Badeda, D. U. Sauer, Wis- Building Stock, Thesis, 2015. URL: http://www.invert.at/ senschaftliches Mess- und Evaluierungsprogramm Solarstrom- Dateien/Dissertation_AndreasM.pdf. speicher - Jahresbericht 2016, Technical Report, Institut [85] J. Schleich, X. Gassmann, C. Faure, T. Meissner, Making fur  Stromrichtertechnik und Elektrische Antriebe der RWTH the implicit explicit: A look inside the implicit discount rate, Aachen, 2016. Energy Policy 97 (2016) 321{331. doi:10.1016/j.enpol.2016. [103] R. Peters, R. Deja, M. Engelbracht, M. Frank, V. N. Nguyen, 07.044. L. Blum, D. Stolten, Eciency analysis of a hydrogen-fueled [86] A.-L. Klingler, Self-consumption with PV + Battery systems: solid oxide fuel cell system with anode o -gas recirculation, A market di usion model considering individual consumer be- Journal of Power Sources 328 (2016) 105{113. doi:10.1016/j. haviour and preferences, Applied Energy 205 (2017) 1560{ jpowsour.2016.08.002. 1570. doi:10.1016/j.apenergy.2017.08.159. [104] Bundesnetzagentur, Biogas- Monitoringbericht 2014, Tech- [87] A. Lahnaoui, P. Stenzel, J. Linssen, Techno-economic analysis nical Report, 2014. URL: https://www.bundesnetzagentur. of photovoltaic battery system con guration and location, Ap- de/SharedDocs/Downloads/DE/Sachgebiete/Energie/ plied Energy (2017). doi:10.1016/j.apenergy.2017.09.093. Unternehmen_Institutionen/ErneuerbareEnergien/Biogas/ [88] prognos, Hintergrundpapier Energieezienzstrategie Geb aude, Biogas_Monitoring/Biogas_Monitoringbericht_2014.pdf?__ Technical Report, 2015. blob=publicationFile&v=1. [89] EWI, Entwicklung der Energiemaerkte - Energiereferenzprog- [105] BMVBS, Kosten energierelevanter Bau- und Anlagenteile bei nose, Technical Report, On behalf of the Federal Ministry for der energetischen Modernisierung von Wohngeb auden, Tech- Economic A airs and Energy (BMWi), 2014. nical Report, 2012. URL: http://www.bbsr.bund.de/BBSR/ [90] Bundesnetzagentur, Monitoringbericht 2017, Technical Re- DE/Veroeffentlichungen/BMVBS/Online/2012/DL_ON072012. port, 2017. URL: https://www.bundesnetzagentur.de/ pdf?__blob=publicationFile&v=2. SharedDocs/Downloads/DE/Allgemeines/Bundesnetzagentur/ [106] Controme, Controme - Fubodenheizung Set-Funk, Techni- Publikationen/Berichte/2017/Monitoringbericht_2017. cal Report, 2018. URL: https://www.controme.com/produkt/ pdf?__blob=publicationFile&v=4. fussbodenheizung-set-funk-2/. [91] K. B. Lindberg, D. Fischer, G. Doorman, M. Korp as, I. Sar- tori, Cost-optimal energy system design in Zero Energy Build- ings with resulting grid impact: A case study of a German multi-family house, Energy and Buildings 127 (2016) 830{845. doi:10.1016/j.enbuild.2016.05.063. [92] KWKG, Gesetz ur  die erhaltung, die modernisierung und den ausbau der kraft-w arme-kopplung, 2016. URL: https://www. gesetze-im-internet.de/. [93] EEG, Gesetz fur  den ausbau erneuerbarer energien, 2017. URL: https://www.gesetze-im-internet.de/. [94] H.-M. Henning, A. Palzer, A comprehensive model for the German electricity and heat sector in a future energy system with a dominant contribution from renewable energy technolo- gies|Part I: Methodology, Renewable and Sustainable Energy Reviews 30 (2014) 1003{1018. doi:10.1016/j.rser.2013.09. [95] UBA, Klimaneutraler Geb audebestand 2050 - En- ergieezienzpotentiale und die Auswirkungen des Klimawandels auf den Geb audebestand, Techni- cal Report, Umweltbundesamt, 2017. URL: https: //www.umweltbundesamt.de/sites/default/files/medien/ 1410/publikationen/2017-11-06_climate-change_26-2017_ klimaneutraler-gebaeudebestand-ii.pdf. [96] T. Schutz,  R. Streblow, D. Mulle  r, A comparison of thermal energy storage models for building energy system optimiza- tion, Energy and Buildings 93 (2015) 23{31. doi:10.1016/j. enbuild.2015.02.031. [97] B. H. King, C. W. Hansen, D. Riley, C. D. Robinson, L. Pratt, http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Physics arXiv (Cornell University)

Bottom-up energy supply optimization of a national building stock

Loading next page...
 
/lp/arxiv-cornell-university/bottom-up-energy-supply-optimization-of-a-national-building-stock-WX8ioXoEoP
ISSN
0378-7788
eISSN
ARCH-3341
DOI
10.1016/j.enbuild.2019.109667
Publisher site
See Article on Publisher Site

Abstract

The installation and operation distributed energy resources (DER) and the electri cation of the heat supply signi cantly changes the interaction of the residential building stock with the grid infrastructure. Evaluating the mass deployment of DER at the national level would require analyzing millions of individual buildings, entailing signi cant computational burden. To overcome this, this work proposes a novel bottom-up model that consists of an aggregation algorithm to create a spatially distributed set of typical residential buildings from census data. Each typical building is then optimized with a Mixed-Integer Linear Program to derive its cost optimal technology adoption and operation, determining its changing grid load in future scenarios. The model is validated for Germany, with 200 typical buildings considered to suciently represent the diversity of the residential building stock. In a future scenario for 2050, photovoltaic and heat pumps are predicted to be the most economically and ecologically robust supply solutions for the di erent building types. Nevertheless, their electricity generation and demand temporally do not match, resulting in a doubling of the peak electricity grid load in the rural areas during the winter. The urban areas can compensate this with ecient co-generation units, which are not cost-ecient in the rural areas. Keywords: energy systems, typical buildings, Mixed Integer Linear Programming (MILP), Zero Energy Buildings (ZEB), building stock, aggregation 1. Introduction well as governments [13]. Therefore, analyses are needed that predict technological development and their system 1.1. Background integration: Many works for the building sector [14{18] The residential building sector is responsible for 17% solely focus on GHG reduction strategies for heat demand. of worldwide CO emissions [1]. In Germany, it was the They conclude that signi cant energy saving potentials source of 10% of the total Greenhouse Gases (GHG) through can be accessed by increased refurbishment rates and that fossil fuel combustion in the year 2015. Moreover, it was residual heat can be supplied with renewable energy. responsible for 12% of total emissions due to the GHG Nevertheless, in relation the heat demand of the build- footprint of its energy imports [2{4]. These emissions must ing sector can no longer be regarded as being more isolated be cut in order to reach the overall goal of net zero GHG from the other energy system: Heat pumps are seen as a emissions in the second half of this century [5] with the key option to eciently provide space heat [19, 20], while goal of minimizing the impact of anthropogenic climate Combined Heat and Power (CHP) generation allows an change [6]. Therefore, the European Union introduced the ecient usage of chemical energy carriers while providing concept of "Zero Energy Buildings" (ZEB) in the context exibility to the grid [21]. Furthermore, a trend towards of its energy performance of buildings directive [7, 8] with an increased self-supply of residential buildings is appar- the goal of deploying GHG-neutral buildings that com- ent with the rapidly falling prices of photovoltaics [22] and pensate for their emissions by exporting on-site generated batteries [23] constituting grid parity [24], meaning that renewable energy [9, 10]. the levelized cost of self-generated electricity is below the While the objectives are clear, the pathway to GHG- retail electricity grid price. neutral building stock is uncertain and the integration of Both trends, i.e., the changing heat supply and the in- new technological solutions unsettles utilities [11, 12] as creasing self-suciency of the buildings, will signi cantly change the future grid demand and challenge the feasibil- ity of current electric grid design. Therefore, new analyses Corresponding author. Email: l.kotzur@fz-juelich.de Preprint submitted to Energy & Buildings January 7, 2020 arXiv:2001.01554v1 [physics.soc-ph] 2 Jan 2020 Nomenclature CAPEX CAPital EXpenditure OPEX OPerational EXpenditure CHP Combined Heat and Power QIP Quadratic Integer Program CRF Capital Recovery Factor RMSE Root Mean Squared Error DER Distributed Energy Resources SFH Single-Family House GHG Greenhouse Gas UB Upper Bound LB Lower Bound WACC Weighted Average Cost of Capital MILP Mixed-Integer Linear Program MFH Multi-Family House ZEB Zero Energy Building are required that consider the adoption and operation of is guaranteed. Some of the models even include the in- new supply technologies and eciency measures, predict- vestment decision into eciency measures by changing the ing the spatially- and temporally-varying impacts on the buildings envelope [41, 45{47]. grid infrastructure. All models would enable the analysis of the impact of technology adoption and operation on the local infrastruc- 1.2. Related works tures: for instance, Lindberg et al. [37] apply a MILP to design the supply system of a Multi-Family-House (MFH) Various works have already perforemed top-down anal- and analyze the resulting electricity grid load for cost- yses of the load change due to single Distributed Energy optimal system operation under current German regula- Resources (DER), like photovoltaic [25], heat pumps and tions. Schutz  et al. [39] use a model to evaluate the opti- battery electric vehicles [26], or Combined Heat and Power mal technology adoption with currently considered incen- (CHP) [27]. Seljom et al. [28] analyze the impact of pho- tives and market conditions for the case of three reference tovoltaics deployed in ZEB in for the future Scandinavien buildings. energy system. Nevertheless, as demand, exibility op- tions and generation are closely connected to the building 1.2.2. Archetype buildings supply systems, these technologies can not be evaluated Although these analyses and models already provide independently and must be get holistically regarded and many insights for the application cases, a further gener- modeled as system entities. This can primarily be done by alization would be required to upscale these results to an means of detailed bottom-up models that simultaneously aggregated nationwide perspective. Furthermore, the spa- consider investment decisions and the operation of DER as tial variation due to regionally di ering building topologies well as eciency measures and demand side management. would be required to integrate the results to grid mod- els. Therefore, a set of representative buildings is required 1.2.1. Building optimization that characterizes the spatial diversity of building stock Thereby, the models must account for cost optimality, and that can be used for the previously described building as the main motivations of building owners to adopt di er- models. ent supply technologies are savings or earnings emanating In general, such typical buildings are often referred to from their installation [29]. This also applies for the ef- as archetype buildings and are commonly used for model- ciency measures or energy retro ts, where the need of ing GHG reduction strategies in the building sector [48], replacement or nancial pro tability are the main activa- as described in the Energy Performance of Buildings Di- tors for the adoption [30]. rective [7, 8] of the European Union. Therefore, many di erent optimization models have In this context, Corgnati et al. [49] introduced di erent been proposed for determining the cost optimal invest- pathways to determine representative reference buildings ment decisions and operation of building supply systems: for the analysis of cost optimal refurbishment measures, either as Linear Programs (LP) [31, 32] with the advan- but conclude that in reality most often a mixture is used tage of good computational tractability but the disadvan- due to the di erent available data for di erent buildings tage of not being able to account for economies of scale; stocks. or as a Mixed Integer Linear Program (MILP) model [33{ Mata et al. [50] proposed an analytical methodology 39]. Furthermore, two-level approaches that determine at to aggregate archetype building stocks based on publicly least a part of investment decisions with a meta-heuristic available data. The steps include a segmentation based solver and operation with a simulation or optimization are on categories such as construction year, a technical char- popular [40{45]. The last approach can account for very acterization such as the thermal transmittance as input detailed physical models but no global optimal solution values for energy performance models, as well as a quan- 2 ti cation to scale the buildings up to a nationwide level. 1. aggregate a spatially resolved building stock with The methodology is applied to France, Germany, Spain, a limited number of archetype buildings which and the UK, with the resulting nal energy demand show- are described by a set of attribtues related to their ing a deviation of less than -4% to + 2% to the aggregated energy supply and demand statistical values. 2. and to optimize those di erent buildings in par- Various nation-speci c works exist to quantify the en- allel, considering a superstructure of supply tech- ergy consumption of the building sector with the help of nologies and potential eciency measures. archetype buildings: the German residential building stock Thereby, di erent regulatory regimes can be considered, is described by a framework developed by the Institute fr resulting in di erent future technology installations and Wohnen und Umwelt (IWU) [51, 52]. This schema cat- operations, and grid demands. The spatial assignment of egorizes the stock into classes that di er by construction the archetype buildings allows then for the local evalua- year and building size that are represented with detailed tion of the changing energy demand. The general idea is technical parameters. This stock description has been ex- visualized Figure 1. tended to other European countries in the framework of the EPISCOPE project [53]. The US Department of En- ergy (DOE) introduced archetype buildings for the resi- dential sector [54] and the service sector [55] in the USA, referred to as benchmark or prototype buildings. An ad- vantage over the European database [53] is the fact that additional time series data are provided for the di erent building demands, including electricity, hot water, cool- ing or heating demand for typical days in di erent climate zones. The aggregation to archetype buildings is also widespread in the context of urban energy models: DallO et al. [56] present a work ow to derive archetype buildings with Figure 1: Structure of the bottom-up model to optimize the a spa- a combination of statistical data and a survey applied tially resolved building stock and determine the changing infrastruc- to sample buildings. Meanwhile. Cerezo et al. [57] and ture usage. Sokol et al. [58] introduce methods to estimate unknown attributes of the proposed archetype buildings, such as With the help of this modeling approach, the poten- comfort temperature levels, based on a probability distri- tial for self-sucient energy supply systems in residential bution. This approach can make use of measured energy buildings can be eciently evaluated and its large-scale data in di erent buildings, such as annual or monthly gas techno-economic impact on the grid demand can be de- demand, and ts the uncertain attributes to it. More- rived. To do so, Section 2 introduces the used aggregation over, Fazlollahi et al. [59] and Fonseca and Schlueter [60] tool, the used data sets and the parallel building optimiza- use k-means clustering methods to group similar buildings tion for this model. The approach is applied for a reference in urban areas by using the location of the buildings and scenario in Section 3 and validated to the available aggre- spatially resolved statistics. The advantage of clustering is gated energy demand data in Germany. Thereby, a trade- that the simulation models or optimization models can be o is made regarding the number of archetype buildings applied to the zones instead of the single buildings, which to describe the diversity of building stock on one hand, reduces the number of variables and the computational but limiting the computational by the number of Mixed- load of the related models. Integer Linear Programs to solve. In order to derive the changing load, a future supply scenario is introduced in 1.3. Own approach and structure Section 4 for the year 2050. Section 6 critically recaps the In summary, many works exist that consider detailed work and draws the main conclusions. integrated building optimization models to determine the cost optimal technology adoption and operation. There also exist many di erent approaches for the aggregation 2. Methods and data and segmentation of archetype buildings. Nevertheless, to the knowledge of the authors the combination of both The model consists of two main components: for the purpose of analyzing the spatially and temporally 1. The building optimization model, described in Sec- changing demand of the infrastructure, i.e. gas grid and tion 2.1, represents the decision making of the build- electricity grid, does not exist. ing owner regarding the design and operation of the energy supply and energy demand. 1.3.1. Therefore, this work proposes a two stage framework to 3 2. The aggregation and distribution of archetype build- 2.1.2. Optimizing structure, scale and operation ings, introduced in Section 2.2, makes the build- The building optimization is based on a typical Mixed- ing optimization results generalizable to the perspec- Integer Linear Program (MILP) with the objective of min- tive of regulators and grid operators on a nationwide imizing the annual energy cost of a single building as pro- scale. posed in the vast literature. The operation and design of the supply system is modeled with the object-oriented 2.1. Building optimization system modeling framework FINE [65, 71, 72]. The bi- It is assumed that building owners are the decision nary variables are considered to suciently incorporate makers and they consider the energetic supply of the build- the economy of scale of the technologies. The operation is ing from the perspective of a single economic entity. There- modeled in a fully linear and continuous manner for 8760 fore, the goal of the building optimization is to have a hours in a representative year. holistic perspective of single residential buildings and is All in all, the combinatorial consideration of demand- able to consider synergies between di erent solutions, e.g. side and supply-side measures respecting the full opera- demand-side measures are simultaneously considered with tional variety yields a complex mathematical program that supply-side measures, or the operation of the heating sys- is computationally demanding. In order to keep the pro- tem is optimized together with the operation of the elec- gram tractable for many di erent building types and sce- tricity system. narios, the annual time series of weather, occupancy be- havior, and appliance load are aggregated to twelve typi- 2.1.1. Creation of demand and supply time series cal days with a hierarchical aggregation [72, 73]. The days In order to derive the cost optimal supply system for with the smallest temperature and highest electricity load each individual building, rst the temporally varying en- are added as extreme days. Based on these, the optimal ergy demands of electric devices, hot water and thermal choice of the supply technologies and refurbishment mea- comfort, as well as the varying performance of renewable sures is determined with the MILP. The binary decision supply technologies are derived as follows: variables are then xed and a validation and scaling opti- The occupancy behavior and inherited electricity load mization is then performed for the full annual time series of the appliances are created with the help of the CREST [74], similar to that in Bahl et al. [75]. demand model [61{64]. Further, the demand for hot water The detailed model description and their independent is separated from it. The advantage of this model is that validation can be found in Kotzur [65]. it suciently incorporates the high variance of single res- idential load pro les, as well as the stochastic smoothing 2.2. Aggregation of archetype buildings for the case that an agglomeration of households is consid- In order to determine the di erent types of buildings ered. A validation of the model in the context of German the number of their occurance, this section introduces a residential electricity demand is performed in Kotzur [65] new aggregation method to derive spatially-distributed archetype and exhibits sucient accuracy. buildings. Those are used to scale the results of build- The heat load is considered with the 5R1C-model from ing optimizations to a spatially-resolved nationwide per- EN ISO 13799 [66], which was implemented into a MILP spective. Therefore, Section 2.2.1 discusses the relevant by Schutz  et al. [47]. The physical building properties, attributes to describe the energy performance of a build- such as heat transfer coecients for di erent construction ing. The aggregation algorithm itself is sketched in Sec- years are taken from IWU [52]. This is able to account for tion 2.2.2 and its application is illustrated in Section 2.2.3. the thermal building mass for a exible supply system op- 2.2.1. Relevant building attributes eration. Furthermore, potential refurbishment measures are part of the solution space, such as the addition of wall In general, four categories of building attributes are or roof insulation, the replacement of windows, or the in- emphasized in the literature, while the concrete nomen- tegration of smart thermostats. The con guration of the clature varies [49, 50]: The Form describes the physical buildings is introduced in detail in Kotzur [65]. exterior shape of the building, including orientation, wall The time series for PV and solar thermal are created area, window area and roof areas. The Envelope charac- with the PV-Lib [67]. The weather data is derived from terizes the physical properties of the materials used in the the DIN EN 12831 [68] by nding the closest location building. The technologies installed in the buildings to sat- listed. Therefrom, the minimal design temperature is de- isfy thermal comfort and other demands are grouped into rived as well as the test reference region of the Deutsche the category of System. Operation, in turn, summarizes Wetter Dienst (DWD) [69]. Alternatively, the weather all extrinsic conditions determining the system operation, from the COSMO rea-6 reanalysis data set [70] is used for such as the local weather or occupancy behavior. real weather years for validation purposes. Aside from the attributes describing the current en- The whole initialization of the building speci c time ergy performance, future energy supply is also of interest, series are published in the open source python package where the category of Adoption summarizes all attributes tsib - Time Series Initializaion for Buildings (https:// referring to the investment capabilities and investment be- github.com/FZJ-IEK3-VSA/tsib). havior of the building owner, such as a potential interest 4 rate. For instance, the model described here considers the cost-optimal technology adoption of the di erent build- ings. The categories and their aggregation de ne a general framework to segment buildings, but the required attributes depend on the model and data availability, e.g., the enve- lope could be described by materials with exact heat con- ductivities and thicknesses, only heat transfer coecients, or by the construction year of the building from which these values are derived. The attributes considered for the aggregation proce- dure of this work are oriented towards the model intro- Figure 3: Structure of the considered attributes that relevant for the duced in Section 2.1 and the data provided by the cen- building energy supply. sus [76]. Figure 2 shows the aggregated Census data for Germany. The total number of buildings is predominated by over 12.3 million Single-Family Houses (SFH) and 6.3 Thereby, two challenges arise: rst, the buildings are multi-family houses, while buildings with more than 12 described by a mixture of categorical and continuous at- apartments have, at 0.21 million, a small share at the total tributes. Approaches exist dealing with this type of aggre- number. The majority of the SFH are detached, consti- gation class, such as the mixture of k-means and k-modes tuting an overall small proportion of terraced and semi- clustering, referred to as k-prototypes [77]. Nevertheless, detached buildings. 23.2 million of the 40.5 million apart- these would rely on a data set consisting of real building ments are rented, while one- and two-person households instances that should be clustered and can then be rep- dominate with together 27.1 million households. These resented, e.g., by its medoid. This does not apply to the also constitute the peak of apartment sizes, with compact attributes distributions, and so a new aggregation method- living areas of 59 to 79 m per household, while larger ology is required that is introduced in Kotzur [65]. single-family houses are spread over a larger grouping. Thereby, a greedy algorithm is introduced with the goal 1000 Owner- Rented ship Owned of determining a locally optimal set of archetype build- Apart- ments ings. It is inspired by the concept of an expectation- Persons per 3 Terraced app. Sur- 2 maximization algorithm, where Lloyd's k-means cluster- Semi rounding Detached 2020 ing algorithm [78, 79] and the k-prototypes algorithm [77] 179 belong as well. Con- Apart- 159 struction 1987 ment 139 year 2 size [m ] 119 before 1969 below 1959 79 1949 59 1919 39 0 5 10 0 10 20 Buildings [Mil.] Appartments [Mil.] Figure 2: Aggregated attribute distribution of the German residen- tial building stock based on the Census [76]. Figure 4: Flow chart of the developed algorithm to determine a All of these distributions are also available on an abso- spatially distributed archetype building stock. lute scale for the municipalities or a 100m grid in Germany and state the data basis for the considered archetype ag- gregation, as shown in Figure 3. The introduction of ad- The idea is to describe the assignment of the archetype ditional values is discussed in detail in Kotzur [65]. buildings to the di erent nodes or municipalities as the expectation step, with the objective of getting a repre- 2.2.2. Building aggregation algorithm sentation of the attribute distributions by the most likely These attribute distributions must be aggregated to a archetype buildings into every municipality. Nevertheless, limited set of archetype buildings to evaluate them ener- the attributes of the archetype buildings themselves are getically. unknown, and so their estimation is de ned as the maxi- In general, the attributes belonging to the di erent mization step, illustrated in Figure 4. This results in two categories are primarily published as aggregated distri- optimization problems that are iteratively solved. butions for di erent administrative boundaries that de- The result is a set of archetype buildings and a matrix ne the spatial resolution. Nevertheless, the real building that de nes the representation of every municipality by instances and their values are unknown wherefore cross- the number of certain archetype buildings. combinations of attributes are not reproducible, e.g., how many terraced buildings have a certain living area. 5 2.2.3. Application and illustration of the results low 2 %. Thus, it would not be ecient to sacri ce an additional archetype building to represent it. The initial guess of the archetype building attributes, In general, a t below 100 % does not imply that the the start solutions for the algorithm, are derived from the expression is highly under-represented on the aggregated current state of the art archetype buildings for Germany level: While an overestimation of an attribute in one re- [52], while missing parameters are randomly generated, gion and an underestimation in the other regions consti- e.g. the number of persons living in an apartment. tute a reduced t, they could add up and compensate each In the ollowing, the algorithm is applied once to di er- other on an aggregated nationwide level, which is further ent prede ned numbers of archetype buildings in order to elaborated in Kotzur [65]. determine how many of these are required for a sucient The tting of the continuous attributes, the latitude representation of the German building stock. and longitude, is qualitatively illustrated in Figure 6 with The quality of the resulting representation of the di er- their exact geographical placement in Germany. For the ent attributes for di erent numbers of archetype buildings case of 5 to 25 archetype buildings, all buildings are pri- is illustrated in Figure 5. This is de ned as the cumulative marily located in the center of the country. The reason deviation of the representation of an attribute expression is that building archetypes are mainly used to represent m for every region in proportion to the total attribute the diversity of categorical attribute combinations that manifestations for the whole of Germany: are spatially distributed across Germany. E.g., a single- P P d d n;m;p n;m;p b;n b;p;m family house from 1960 with a four-person household and n2N b2B f (p; m) = 8 p; m gas boiler supply manifests as an archetype building that n;m;p n2N represents this building type in the north as well as in the (1) south. For higher numbers of archetype buildings from 100 Number of buildings to 800, the geo-spatial location of the archetype buildings 5 25 100 400 is spreading, as similar categorical building types can be 10 50 200 800 instantiated multiple times. For the case of 800 archetype CHP 6 District 5 4 buildings, it is even observable that urban areas are rep- Electric Supply Persons 3 tech- Pellets nology Heat pump 2 resented by more archetype buildings than rural areas. Gas boiler Oil boiler Photo- 0 voltaic Rebuild Apart- ment 139 Rented Owner- size 119 ship Owned below 90.0 67.5 Orien 45.0 tation 22.5 Con- Apart- ments struction 2 year before 1 1969 Number archetypes Terraced 5 Buildings Sur- Semi 10 Buildings rounding Detached 1919 25 Buildings 0 50 100 0 50 100 50 Buildings Attribute fit [%] Attribute fit [%] 100 Buildings 200 Buildings 400 Buildings 800 Buildings Figure 5: Relative t of all categorical attribute expressions for dif- ferent numbers of archetype buildings in Germany. Figure 6: Geographical location of di erent numbers of archetype buildings in the nal iteration step. The gure shows that for some of the attribute ex- pressions, a small number of archetype buildings is able to represent them suciently, for instance single-family The representation of municipalities is further illus- houses with a single apartment or the energy supplied by trated in Figure 7, which introduces the local assignment gas boilers. These are attribute expressions that often oc- to the di erent municipalities of certain example archetype cur in the original data set. Therefore, they are rst rep- buildings, selected from the set of 800. The locations of resented by the archetype buildings to reduce the overall the archetype buildings are the centers of the buildings error. Nevertheless, attributes such as a CHP, heat pump they represent in the di erent municipalities. As these supply, or apartments with a living area smaller than 39 representations are spatially spread over di erent munici- m rarely occur in the Census data set. Therefore, the palities, the archetypes are not placed at municipalities at algorithm has a secondary priority to represent them and the border. focuses instead on building attributes that exist more of- Furthermore, it is recognizable that the areas and amounts ten, e.g., no archetype building was created with a heat that are represented di er between the archetypes: While pump supply for 5, 25 and 50 archetype buildings because an archetype building supplied with heat pumps must rep- the overall share of heat pump supply in Germany is be- 6 resent buildings over a large area, archetype single-family Energy carrier Electricity houses supplied by gas boilers have a de nite local assign- Renewables ment area. The reason for this is that more archetype District heating Oil buildings with gas boilers are selected, as more buildings Gas 200 Coal with gas boilers also exist in reality. Therefore, the algo- rithm chooses a higher spatial separation for them to min- imize the overall error, while accepting a higher geospatial estimation error for the few buildings with heat pumps. Archetype buildings Archetype building Archetype building Figure 8: Final residential energy demand predicted for di erent with heat pump with gas boiler numbers of archetype buildings. can be seen in Figure 8, the model is able to roughly pre- Location arche- type building dict with ve to ten archetype buildings the demand of the 1000 assignments 100 assignments three dominant energy carriers, but the appearance of mi- 10 assignments nor energy supply carriers is not suciently included. This improves with an increasing number of archetype build- ings while the best t can be achieved with 800 buildings. Nevertheless, the resulting demands of 64.8 TWh/a for re- newables and 44 TWh/a for district heating are still under- estimated. This deviation is constituted by the aggrega- Figure 7: Location of the most northern (blue) and most southern tion, which tries to capture the most frequently appearing (green) single-family house archetype with heat pump supply (left) archetype buildings and neglects rarely occurring building and gas boiler supply (right) of the set of 800 archetype buildings, types. Nevertheless, these missing energy demands for re- and their assignment to the di erent municipalities. newables and district heating are compensated by gas and oil demands, which are slightly overestimated with 286 and 165 TWh/a. This compensation e ect already appears for 25 archetype buildings, where all cases between 25 and 3. Validation of the method 800 archetype buildings predict the total nal energy de- mand in a similar magnitude as the AGEB [2]. Above For the validation of the two-staged methodology, the 200 archetype buildings, the share of the di erent energy di erent numbers of aggregated archetype buildings are carriers also aligns well with the structure of the AGEB independently optimized for the status quo and then mul- [2]. tiplied with their appearance in Germany. The choice The prediction with 50 archetype buildings overesti- of the technologies is prede ned by the archetype def- mates the demand for oil by 27.4% and underestimates the inition [65], but the technology scale and operation are demand for gas by 17.1%, while 100 archetype buildings optimized such that the building-speci c energy demands on the contrary overestimate the gas demand by 25.5% are met. It de nes the Reference, or status quo, of the and underestimate the oil demand by 16.1%. This switch residential energy supply and validates the model to na- shows a drawback of the aggregation: Some archetype tional energy demand statistics. The techno-economic as- buildings appear often and therefore have a high impact sumptions are introduced in Appendix A. The di er- on the overall energy load. If the majority of the buildings ent sets of archetype buildings together with their spa- supplied, e.g., with gas boilers have a construction year tial distribution can also be found in the tsib (https: before 1960 while the more modern buildings are supplied //github.com/FZJ-IEK3-VSA/tsib). with oil, an overestimation of the gas demand and an un- derestimation of the oil demand results, although the abso- 3.1. Impact of the number of archetype buildings lute number of the di erent boiler types is well represented. The impact of choosing di erent number of archetype Nevertheless, this e ect is reduced with an increasing num- buildings to the resulting nal energy demands, aggregated ber of archetype buildings, as single archetypes represent to di erent energy carriers, is illustrated in Figure 8. They attribute distributions on a more granular level. In conse- are validated against the nal energy demand provided by quence, the spatial di erences, e.g., of construction years, AGEB [2]. are better tted and intrinsic correlations of the input data The dominant energy carriers for the residential sector are represented with higher accuracy. are gas, oil and electricity with 268, 162 and 136 TWh/a This work assumes 200 archetype buildings as a su- per year [2]. The demand for renewable energy or district cient trade-o between accuracy and computational load, heating is secondary with 84 and 51 TWh/a per year. As since they already capture the main diversity of the en- Energy demand 2013 [TWh/a] AGEB ergy carriers and the statistical balancing e ects between when Southern Germany's energy demand was 9% above the buildings. the regional average, while northern Germany lies just 5% above the average. This highlights the importance of hav- 3.2. Impact of di erent weather years ing a spatially resolved building stock model, as a single location is not able to represent this variation suciently. The impact of di erent weather years from 2010 to 2015 on the energy demand of the buildings is illustrated in Figure 9 for 200 archetype buildings and validated again to the nal energy demand values provided by AGEB [2]. According to AGEB [2], the total residential energy de- mand varies from 743 TWh in 2010 at a maximum to 608 TWh for 2014 as a minimum. 0 20 1.11 1.17 0.92 0.95 1.00 1.02 2010 2011 2012 2013 2014 2015 TRY Avg. demand Change of Change of Change of 2010 to avg. 2011 to avg. 2012 to avg. [GWh/a/km ] Energy carrier Electricity Renewables 500 District heating Oil Gas Coal 0 1.05 1.09 0.88 0.90 0.92 0.98 1.0 1.2 Change of Change of Change of Change of 2013 to avg. 2014 to avg. 2015 to avg. TRY to avg. Figure 10: Spatial distribution of the nal energy demand, averaged Figure 9: Final energy demand for di erent weather years predicted for the considered years 2010 until 2015, and the relative regional with 200 archetype buildings and compared to the values reported deviation from the average value. by the AGEB [2]. In 2014, no signi cant di erences due to the geo-position For all di erent weather years, the systematic overes- are observed. Nevertheless, it becomes clear that the cities timation of gas demand and underestimation of district are less sensitive to the weather patterns (11.5% below the heating demand is observed, as already discussed in the average in 2014) than the rural areas (13.5% below the previous section. Nevertheless, the relative deviation dif- average in 2014). The reason is that the relative share fers between the years. While the total nal energy de- of energy demand for space heating to the overall energy mand ts well for 2010 with an underestimation of below demand is smaller in the cities than in the rural areas, re- 2%, the deviation increases in the year 2011 up to 7.4%. It ducing the relative impact of weather years on the total reduces again to 2.8% in the year 2013 while 2014 again has energy demand. a value of 5.8%. The di erences are mainly constituted by The analysis illustrates that the novel spatially resolved the di erent demands for the energy carriers that are used approach is able to identify local extreme weather pat- to supply the space heat, while the electricity demands terns. While it was only shown here for the aggregated remain almost constant for all periods. It seems that in annual demand, the model also predicts the temporal de- relatively mild weather years, the deviation is higher than mand of the energy carriers in all municipalities and can in colder weather years. A probable explanation of the be used for the identi cation of local peak demands that varying deviations could be an adaptive occupancy behav- are relevant for the infrastructure's design. ior, e.g., the ventilation rates could be reduced in colder winters, which is not taken in to account in the model. 4. Future energy supply scenario Figure 10 shows the spatial distribution of the nal energy demand averaged for all of the weather years con- In order to predict the change of the supply struc- sidered. It clusters in the cities as expected. Addition- ture in the future, the overall model is applied with 200 ally, the relative changes of the nal energy demand for archetype buildings for the techno-economic assumptions the di erent weather years are illustrated for the di erent in 2050, which are de ned in Appendix A.2. The choice, municipalities. The overall magnitudes of the di erences scale, and operation of the considered energy supply tech- align with the di erences shown in Figure 9. Neverthe- nologies are optimized together with the heating system less, it is clearly recognizable that di erent weather years and potential refurbishment measures, implying that the spatially impact the annual energy demand spatially dif- building owners have a technology adoption and operation ferently: While the year 2010 was generally a cold year, in approach that minimizes their energy cost and act as homo northern Germany the nal energy demand was 17% above economicus. We de ne this as Min Cost scenario. the average while in south-west Germany it was only 11% higher than the average. This is the opposite to 2013, Energy demand [TWh/a] AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. AGEB 200 bdgs. The results de ne the overall state that the residential the relatively small cost share of the eciency measures energy supply system is converging on if the assumed en- is that most measures are chosen for buildings that are ergy prices and the techno-economic assumptions for the anyway in the refurbishment cycle due to their lifetime. technologies arise. Besides the incentives included in the Therefore, the costs for the sole eciency measures are scenario, no additional ones are given by regulators. The relatively small because installation costs, such as scaf- demand for the use of electrical devices, hot water demand folding construction, are seperately considered for these and thermal comfort level are asumed to stay the same as buildings. the status quo described in Section 3 with the assumptions The overall results are aggregated from the optimal described in Appendix A.2. Existing technologies are as- system design of the di erent archetype buildings, whose sumed to get replaced until 2050, wherefore each building cost structure is illustrated in Figure 12. The total annual has a green eld optimization regarding the supply side. cost of the buildings is scaled by the number of house- holds in the buildings to show di erent sizes of buildings 4.1. Design, costs and operation on a similar scale. In order to expose patterns between the buildings, they are manually clustered into four groups The resulting total expenditures related to investments based on their resulting supply system. The Single Family in the supply structure or energetic refurbishment mea- Houses (SFHs) are manually di erentiated between those sures are visualized in Figure 11. The aggregated capaci- with and those without heat pumps, while the Multi Fam- ties and energy ows are shown in Appendix B. To real- ily Houses (MFHs) are distinguished between those with ize the technology portfolio, an overall investment of 382.3 and without fuel cells. billion Euro is needed. The largest share is photovoltaics, with a total investment of 104.6 billion Euro and a to- tal capacity of 133.4 GW. 89.7 TWh/a of the generated electricity is used for self-consumption, which is the main incentive to deploy photovoltaic. Solar thermal Photovoltaic The second highest investment is for heat pumps with Battery 88 billion Euro indicating that these are the main supplier Fuel cell Log supply of space heating. Fuel cells are the chosen exible co- Electric heater generation option and amount to 5.4% of the annual costs. Heat pump Heat storage The heat storage systems make up 1.9% of the annual costs Gas supply and have a total investment of 21.8 billion Euro. The Gas boiler Electricity grid investment in the batteries is signi cantly lower with 6.9 Roof billion Euro, amounting to 1% of the annual costs. The log Windows Walls wood supply for the replaces amounts to 2.1%, while the Control electric heaters have a minor share. District heating, oil 0 500 1000 1500 2000 2500 3000 3500 boilers and pellet boilers are not chosen in the solution, as Total annual cost per household [Euro/a/hh] they are not competitive in comparison to the heat pumps Figure 12: Cost composition of the di erent archetype buildings for or gas boilers. the Min Cost scenario in 2050. They are grouped by Single-Family House (SFH) with and without (wo.) heatpumps and Multi-Family 104.6 Houses (MFH) with and without (wo.) fuel cells. 88.0 53.6 In general, the only technology that is chosen for al- 31.1 26.7 21.8 most all of the buildings is rooftop photovoltaic. With the 19.5 12.8 9.7 6.9 6.0 1.6 predicted small cost of the photovoltaic panels and high 0.1 electricity price, these are in the cost-optimal solution for various scales but independent of the roof orientation of the building. Except for one SFH that has a completely self-sucient electricity supply, no other SFH has a fuel cell installed. Figure 11: Total investments into the di erent measures in the resi- Since the demand pro le of a single-family house is highly dential buildings for the Min Cost scenario in 2050. volatile, the achievable full load hours for a self-sucient electricity supply are too low for a fuel cell to become eco- Together, the energy-related refurbishment measures nomically feasible. Furthermore, the required capacities account for 97.2 billion Euro while more than half of them of the fuel cell would be small, increasing the speci c cost are determined by the walls. The occupancy control sys- due to a missing economy of scale. tems have a relatively high share with 12.8 billion Euro, Moreover, it is striking that the occupancy controllers followed by the windows and the roofs. One reason for are primarily installed in SFHs with gas boilers. The build- Investment [Bil. Euro] Control Walls Windows Roof Floor Gas boiler Heat storage Heat pump Electric heater Fuel cell Battery Photovoltaic Solar thermal MFH wo. MFH with SFH wo. SFH with fuel cell fuel cell heat pump heat pump ing cluster with gas boilers is dominated by compact build- load hours an investment into a gas boiler could be advan- ings where only a few rooms need to be equipped with the tageous. From a central infrastructure perspective, both thermostats, constituting small investment costs. More- options have an intrinsic economic issue, as they need the over, the heat capacity of those buildings is small and layout of an infrastructure that will be used in its maximal constitutes limited thermal inertia. This is bene cial for capacity for only a few hours per year. the occupancy controller, as the building can cool down and heat up faster in the case of vacant occupants. For 4.2. Changing electricity grid load large buildings with a high thermal mass, an occupancy The resulting electricity grid exchange, de ned by the controller only o ers limited bene t. electricity imported for the heat pump, the conventional All MFHs with a fuel cell have an additional heat pump electricity demand, and the photovoltaic feed-in is illus- installed. The cheap self-supply with electricity bene ts trated in Figure 14. For comparison purposes, the grid electrical heat generation. Some of the MFHs add a bat- exchange of the Reference scenario is shown as well. The tery system to increase the share of the photovoltaic and aggregated electricity load of the Reference scenario is CHP electricity that can then be self-consumed. dominated by the occupant activities in the morning and The di erent full load hours and capacities of the tech- evening. A small variation between winter and summer nologies in the di erent archetype buildings are shown in then appears. The overall load peaks in the evening hours Figure 13. The scale of the dots indicate how often the during winter with 36.4 GW. This aggregated load signi - archetype buildings are assigned in total in Germany. In cantly changes for the 2050 scenario, when during the sum- general, it can be seen that although photovoltaics are in- mer the load demand is reduced to values below 10 GW, stalled in all buildings, the achievable full load hours vary and also for the evening hours, while high daytime feed-in from 683 to 1025 depending on the roof orientation and rates of the photovoltaic occur with up to 43.1 GW, ex- location of the archetype building. ceeding the peak demand of the Reference scenario. The The highest full load hours are around 5000, and achieved impact of the photovoltaic gets reduced during the win- by the fuel cells. It is observed that a larger fuel cell capac- ter but still reduces the load at noon for most days. The ity correlates with higher achievable full load hours. This evening hours in winter are still the peak demand times mainly relates to the occupancy pro les: due to statistical with a load of up to 32.3 GW for the 2050 scenario, which balancing e ects, larger buildings have atter pro les that is in a similar in magnitude to the Reference scenario. can be covered by higher self-generation rates. Opposing e ects are observed for the peak generators, such as the Reference gas boiler with around 2000 full load hours and the electric heater with less than 1000 full load hours: the larger the 10 Min. cost installed capacities are, the smaller are the achievable full load hours. For the heat pump, no such e ect is observed. It is operated with between 3000 and 4000 full load hours 0 50 100 150 200 250 300 350 for small capacities as well as large ones. Day Figure 14: Aggregated grid exchange of the national residential Fuel cell 4000 Heat pump building stock for the Reference and the Min Cost scenario. Gas boiler As introduced previously, the technology installations Photovoltaic depend on the building type. Therefore, the changes in the Electric heater 500 grid load also vary spatially depending on the local build- ing topology, as illustrated in Figure 15. As expected, 0 1 2 10 10 10 the majority of the regions reduce their annual electricity Capacity [kW] demand with the help of self-generation by photovoltaics and fuel cells. Nevertheless, regional di erences are high: Figure 13: Full load hours and capacity of the installed technologies while urban areas are able to reduce their electricity de- in the di erent archetype buildings for the Min Cost scenario in 2050. The size of the scatter is related to the overall appearance of mand by 60%, some rural areas even increase it. The high the archetypes in Germany. photovoltaic installations in rural areas are not sucient to compensate for the increased electricity demand from the heat pumps. This e ect intensi es for the case of the The distribution of scales and full load hours indicates peak load, as almost no photovoltaic feed-in exists in the that the heat pumps signi cantly rely on a peak boiler, winter days, while the heat pumps are being operated in since their scaling to the maximal heat load would be more full load. Therefore, regions characterized by large SFHs expensive. Nevertheless, it is open as to which peak boiler double their peak load. This is di erent for the urban ar- is chosen in the model. For a few full load hours, the eas that even reduce their peak load because the fuel cells electric heater is more cost-e ective, while for many peak Full load Full load hours [h/a] hours [h/a] Hour Hour Residential grid load [GW] Solar thermal exceed the electrical capacity of the heat pumps and are Photovoltaic Battery synchronously operated. Equivalent regional trends are Fuel cell Bio methane observed for the feed-in: The rural areas feed up to 40% Pellet boiler Pellet supply of the original electricity demand into the grid, while the Log supply Electric heater urban areas have only small feed-in rates of 10%. Further, Heat pump Heat storage a gradient between north and south is recognizable due to District heating CHP the di erent solar irradiance. Gas supply Gas boiler 0 Electricity grid Ventilation Floor Roof Windows Walls Control Figure 16: Annual cost of the Min Cost scenario and the resulting aggregated system cost if the solution space is constrained. 100 0 100 100 0 100 0 15 30 45 Change of Change of Feed-in in ratio peak demand [%] annual demand [%] to Reference demand [%] they are used in small scales in the case that fossil gas is excluded. The amount of occupancy controllers is also Figure 15: Spatial change of the peak electricity demand and the reduced without fossil gas. The reason is that the heat change of the cumulative positive demand from the Reference sce- nario to the Min Cost scenario. Furthermore, the amount of elec- pump is intensively used during the day in order to uti- tricity feed-in to the grid in the Min Cost scenario is shown in ratio lize photovoltaic electricity while heating up the building. to the cumulative electricity demand in the Reference scenario. Nevertheless, the occupancy controller lowers the comfort temperature especially during the day when the occupants are absent. These two temporally opposing e ects reduce In summary, the results indicate that the change of the the value of an occupancy controller for the buildings sup- energy supply in the rural areas is more challenging with plied with heat pumps. The exclusion of fossil gas also respect to the electricity grid operation than the changes constitutes the smallest GHG footprint that gets reduced in the urban areas. Nevertheless, adaptions in the tari from 51.9 Mt/a in the Min Cost scenario to 19.9 Mt/a, as design could dampen this e ect. shown in Figure 17. 4.3. Value of analysis Electricity grid In order to evaluate the robustness of the Min Cost Bio methane Pellet supply scenario for 2050, 200 archetype buildings are again opti- Gas supply mized, but parts of the technologies are excluded or forced District heating into the solution space. Figure 16 illustrates the aggregated resulting annual cost composition of the di erent cases that were consid- ered for the analysis. Gas supply, fuel cell, photovoltaic, Figure 17: GHG footprint of the Min Cost scenario and the resulting heat pump and refurbishment measures are each excluded aggregated GHG footprints if the solution space is constrained. from the solution space, and the other cost minimal so- lutions are compared to the original Min Cost scenario with all technologies available. The increase in the total The structural changes to the Min Cost scenario are systems costs can be interpreted as the Value Of the inte- fairly small for the case that the fuel cell is excluded from gration of a certain technology. As an additional case, the the solution space. The net electricity import increases as full package of refurbishment measures are enforced for all in the previous scenario and compensates for the missing buildings that are in the refurbishment cycle in order to self-generation, but the photovoltaic capacities merely in- reach lower demands for space heating. crease from 133.4 GW to 142.3 GW, while the heat pump In case the fossil gas supply is excluded from the so- capacities remain in a similar magnitude. This is di erent lution space, the electricity purchase doubles, as no fuel to the previous case and indicates that the value of fur- cells for self-consumption are installed. Bio-methane or ther photovoltaic capacities is mainly correlated to higher another renewable fuel is too expensive in the considered heat pump capacities and not to smaller fuel cell capaci- scenario to replace the fossil gas in the fuel cells. Instead, ties. The battery capacities are reduced from 16.9 GWh to higher capacities of photovoltaics are integrated into the 12.8 GWh, although the photovoltaic capacity is increas- solution, aggregating up to 160.3 GW. Moreover, the ag- ing. This implies that their operation partially comple- gregated cost for heat pumps increases by 38% as their ments the fuel cell operation. share of the heat supply increases. While the Min Cost Signi cant shifts and cost increases are recognizable solution did not include district heating or pellet boilers, Min cost No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment GHG-footprint [Mt/a] Annual cost [Bil. Euro/a] Min cost No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment in the case that the photovoltaic is excluded, while the The reverse e ect occurs for the forced refurbishment electricity purchase only increases from 71.9 TWh/a to case: Installed heat pump capacities are reduced while 80.9 TWh/a, the gas import almost doubles from 172.5 TWh/a gas boiler capacities increase. The overall demand for gas to 285.7 TWh/a with the e ect of a GHG footprint of up and electricity is reduced as the space heat demand drops to 81.3 Mt/a. High gas boiler capacities compensate for to 209.1 TWh/a, in comparison to the 309.8 TWh/a in the reduction of the heat pump capacities from 60.4 to the Min Cost scenario and the 449 TWh/a in the Refer- 46.9 GW . This indicates the enforcing e ect between the ence scenario. Nevertheless, the demand reduction is not th heat pump and the photovoltaic supply, which is econom- able to compensate for the high cost of the refurbishment ically advantageous in the case of self-consumption with measures, resulting in an overall annual cost increase of photovoltaics is available. No battery capacities are in- 29.7%. In particular, ventilation systems with heat recov- stalled, supporting the statement that their main economic ery amount to almost half of the eciency measure costs. driver for installation is the photovoltaic, although they Ventilation systems do not bene t from the refurbishment are also partially used to increase the self-consumption cycle, as their integration cost in the building is mostly with fuel cell electricity. independent of any outside renovation measures. Also no- In the case that a heat pump is excluded from the so- ticeable is the fact that the amount of occupancy control lution space, the amount of gas increases by 113.2 TWh/a systems drops: If the heat demand is reduced anyway, while the electricity demand only gets reduced by 13.5 TWh/a. additional measures such as temporally reducing the inner In consequence, the GHG footprints increase by up to 76.0 air temperature have a minor e ect, making the occupancy Mt/a. Furthermore, the investment in refurbishment mea- controller economically unfavorable. The GHG footprint sures increases by 30%, dominated by more occupancy is only reduced to 45.1 Mt/a, which is much smaller than controller and more wall insulation and vice versa, indi- expected but explained by the switch of many buildings to cating that especially cheap heat produced by the heat gas boilers. This indicates a potential rebound e ect that pumps lowers the motivation to invest in eciency mea- might occur in the future, in that the reduced demand for sures. Also, the re wood supply increases from 21.2 TWh/a space heating lowers the economic incentive to invest in to 51.44 TWh/a, since it is a cheaper fuel than fossil gas ecient but expensive heat supply technologies. in the scenario. Remarkable is the reduced investment in It is striking that the aggregated cost gain is moderate fuel cells, cutting their capacity from 12.7 to 5.1 GW . for all considered cases, except for the forced refurbish- el This illustrates that major fuel cell capacities are built to ment case. This indicates that the prediction of the total supply the heat pumps with electricity. cost is robust and not sensitive to the available technolo- The exclusion of refurbishment measures from the so- gies in the future, although, this robustness only accounts lution space constitutes an increased investment in heat for an aggregated German-wide level, as illustrated in Fig- pumps and a reduced investment in gas boilers. This is ure 18. The gure shows the total cost increase and the surprising, as an enforcing e ect between the heat pump distribution of the cost increase of the single buildings. and refurbishment measures could be expected because the While the total cost in Germany only increases by 3.65% refurbishment measures decrease the required supply tem- for the case that no fossil gas supply is available, one of perature in the building, and vice versa, increasing the the archetype buildings has a 22.9% higher energy costs, eciency of heat pumps. Nevertheless, the economic ef- while some other buildings are not a ected at all, as they fects predominate, the heat pumps have higher investment were also not supplied with gas in the Min Cost scenario. costs than the gas boilers, while on the other hand the Similar e ects are observed for the other sensitivity analy- operational energy cost for the gas supply is higher. In ses: The sensitivities for single buildings are high, but the consequence, heat pumps are favored in the case of high cost of the aggregated result is robust. heat demands and their deployment increases for the case The impact on the electricity grid of the di erent cases of no refurbishment. is further illustrated in Figure 19. It shows the sorted grid load for the Reference scenario, the Min Cost scenario, Total in Germany and all related sensitivity analyses. The highest peak load Single archetypes occurs if the gas supply is completely excluded from the solution. No signi cant gas boiler capacities are able to satisfy the peak heat demand and no fuel cells can dimin- ish the additional electricity load of the heat pumps. In consequence, the peak load almost doubles to 55.9 GW in comparison to the Min Cost scenario with 32.3 GW. The second highest demand is reached if no fuel cell is included and the peak load increases by 14.6 GW rela- Figure 18: Annual cost increase for the case that certain technologies tive to the Min Cost scenario. This shows the importance are excluded or added to the solution space in the Min Cost for 2050. of decentralized exible electricity generation in order to The change is once shown for the total aggregated cost and once for the single archetype buildings. compensate for the increased electricity demand by the heat pumps. As is to be expected, the case without heat No Gas supply No Fuel cell No Photovoltaic No Heat pump No refurbishment Forced refurbishment Annual cost increase [%] pumps has the lowest peak load, with 25.7 GW. service sector including commercial buildings, or also rep- resentative fueling stations, whose detailed models could be upscaled to a nationwide perspective while respecting Reference Min cost the spatially varying conditions to supply them. 20 No Gas supply No Fuel cell No Photovoltaic 5.2. Buildings as economic entities No Heat pump No refurbishment A limitation is the current scope of the building model: Forced 40 refurbishment Mobility demand should be included in future since it has also a high impact on the grid load and provides further 0 2000 4000 6000 8000 Hours [h] exibility. Thereby, also further technology should be in- cluded, such as ground-source heat pumps or hydrogen Figure 19: Sorted grid load of the Min Cost scenario and the grid storage technologies. While those technologies can be eas- load if the solution space is constrained. ily added, the single archetype approach does not allow a good evaluation of district scale technologies since the entities are considered independently and occure in the The amount of photovoltaic feed-in is for all cases that municipalities in di erent combinations, wherefore a sys- include photovoltaics in a similar range. The maximal tematic evaluation, e.g. of heating networks, is hardly feed-in is reached with 55.4 GW for the exclusion of the gas possible. supply, although the single buildings have the constraint Supplementary, the sole nancial agent decision mak- to limit the feed-in to 50% of their maximal capacity. ing is known to perform well in rate of adoption and cu- mulative adoption but underestimates social and attitudi- 5. Limitations and outlook nal components in uencing the technology adoption [82]. In consequence, only a few economically dominant tech- As the previous sections showed the capability of the nologies are chosen in the Min Cost case, although even modeling approach, it has also some limitations. a higher diversity would exist in reality due to individu- ally varying information levels. As a case in point, pellet 5.1. Chosen archetypes boilers were down-selected, although those are considered As concluded by Corgnati et al. [49], the aggregation in di erent scenarios in the literature [15, 19]. This dom- and descriptions of the archetype buildings is highly de- inance is related to the scenario, and a consideration of pendent on the available data. The algorithm in this work di erent biomass prices could change this. Nevertheless, is tailored for the data structure of the German Census in reality an adoption of pellets would also be expected [76] that describes the statistical distributions of building without being economically competitive. Such non-cost parameters at the municipality level. The methodology is optimal adoption behaviors are better included in sim- transferable to other countries where similar data struc- ple adoption models such as Invert tool [83, 84] that in- ture exists, but in case that exact building samples are cludes a statistical randomness in the adoption process, available, it would still be recommended to use conven- but neglects, e.g., the temporal operation. Nevertheless, tional cluster algorithms for the aggregation. The tting the information basis for investment decisions is improving of archetype buildings to meet attribute distributions, as in wherefore the assumption in this work of a cost optimal this work, has the drawback that theoretical building con- investment stays reasonable. gurations are created that meet the distribution values but can signi cantly deviate from real building instances. 5.3. Further scenarios Further, while the set of archetype buildings is able to The good alignment with the reported nal energy de- respect the variety of households with di erent cumulative mand values of today makes the model further suitable to electricity demands due to their di erent appliance equip- develop transformation paths of the building stock until ment and di erent household sizes, the appliance equip- the year 2050. E.g., additional projected supply years in ment and adoption rate are not altered by the socioeco- 2020, 2030 and 2040 could be modeled. Such an approach nomic background of the households. Nevertheless, this could better consider the di erent lifetimes of the tech- is a signi cant descriptor to determine electricity demand nologies and respect the inertia in the adoption process of variation [80, 81] and probably also technology adoption the building owners. rates. The approach would allow this description, but the Thereby, also short term policy design could be better required data is not publicly available. supported. The resulting grid load for the 2050 scenario For more holistic analysis, the algorithm should be in this work is highly related to the at energy tari s con- transferred to other sectors to derive spatially-distributed sidered. Nevertheless, the peak and feed-in values can be sector speci c representatives. Thus, a cross-sectoral spa- controlled by the di erent tari structure, that could give tially resolved bottom-up model can be derived that re- incentives to shift consumption in time in order to atten spects the individual economic entities. Examples are the the load pro les. Therefore, upcoming works will apply Sorted grid load [GW] the model to di erent regulatory regimes and market en- Appendix A. Techno-economic assumptions vironments. A coupling with grid models is promising in The main input assumptions to parametrize the opti- order to determine an economically optimal market design. mization models are introduced in the following section, e.g. residential energy prices and eciencies of the di er- 6. Conclusions ent technologies. Section Appendix A.1 de nes the pa- rameters for the year 2015, while Appendix A.2 extends In this paper, a novel bottom-up model was introduced and adapts them to the year 2050. that is based on an aggregation of archetype buildings and a related optimization model to predict the a spatially- Appendix A.1. Assumptions for 2015 resolved technology adoption and operation. In order to achieve a valid comparison of today's res- The model approach allows the evaluation of the in u- idential energy supply to the changes that will occur in ence of regulatory decisions on energy cost, green house the future, a valid scenario framework is introduced that gas emissions, or grid load under the assumption of cost represents today's cost and operation parameters for the optimal behavior. The novelty is that it is able to analyze residential energy supply systems. the impact regulatory regimes and market environments The economic parameters considered for the supply for single buildings, simultaneously with a nationwide eco- technologies are illustrated in Table A.1, while their de- nomic perspective. tailed derivation is discussed in Kotzur [65]. The struc- As show case, the model was applied and validated ture of the investment costs is oriented around the cost for the residential building stock of Germany and a future model introduced in Lindberg et al. [37]. It di erentiates scenario frame for 2050. The main techno-economic con- between the xed investment costs that occur in the case clusions drawn from the future scenario are the following: of installation and the speci c investment costs that are added and related to the scale of the installations. • The key technologies for reducing the GHG emis- sions in the building stock are photovoltaic and heat- Table A.1: Assumed economic parameters of the energy supply tech- pumps that will signi cantly increase the seasonal nologies for the Reference scenario. variation of the residential electricity load, as their feed-in and demand do not temporally match, result- Technology CAPEX CAPEX OPEX Lifetime Source x speci c %CAPEX/a a ing in a doubling of the peak electricity load in the Gas boiler 2800 Euro 100 Euro/kW 1.5 20 [65] th winter hours in rural areas. Oil boiler 2800 Euro 100 Euro/kW 1.5 20 [65] th Pellet boiler 10000 Euro 300 Euro/kW 3.0 20 [65] th Heat pump 5000 Euro 600 Euro/kW 2.0 20 [65] th • The urban areas can compensate the increasing elec- Heat storage 800 Euro 1200 Euro/m 0.0 25 [65] Photovoltaic 1000 Euro 1400 Euro/kW 1.0 20 [65] el tricity demand by ecient co-generation units, e.g., IC CHP 15000 Euro 1000 Euro/kW 7.0 15 [65] el Solar thermal 4000 Euro 350 Euro/m 1.0 20 [65] in form of fuel cells, which cannot achieve sucient Electric heater 0 Euro 60 Euro/kW 2.0 30 [37] th economies of scale in single family houses in the rural areas. Although the model allows for the modeling of di erent • Signi cant amounts of photovoltaic electricity (for interest rates for di erent building types to take into ac- Germany up to 90 TWh/a) can be self-consumed count of the di erent investment behavior of the building while the majority is used for internal heat supply owners [85], it is here simpli ed to a single interest rate of applications. Batteries are hardly deployed, as the 3%, which lays between the 2% to 5% considered in the heat applications provide enough exibility for self- literature [32, 37, 47, 86, 87]. consumption. The energy and resource prices are illustrated in Ta- • Refurbishment measures are expensive and only cho- ble A.2. The majority of the prices are derived from the sen in cases when the building is in the cosmetic study Energieezienzstrategie Gebude [15, 88]. Their as- refurbishment cycle. Therefore, space heat demand sumptions de ne the basic scenario framework for this only decreases by 30% from 2015 to 2050 in the sce- thesis and rely themselves on the Energiereferenzprognose nario. Instead, a shift towards an ecient integrated [89]. The majority of the resource prices assumed in the energy supply system, e.g. combinations of fuel cells study align with the energy prices observed for 2016 [18, with heat pumps, is favored. 90]. Nevertheless, the assumed gas price overshoots the observed price of 2016 by more than 1 ct/kWh, and was adapted in this work to the values reported for 2016 by Acknowledgments the Bundesnetzagentur [90]. This work was supported by the Helmholtz Associa- The GHG footprint includes the emissions of the previ- tion under the Joint Initiative Energy System 2050 A ous conversion processes, such as in the extraction of fuels, Contribution of the Research Field Energy . Parts of this or the GHG emissions of the German power plant mix. article are derived from the doctoral thesis Future Grid Furthermore, the price structure is modi ed from a sole Load of the Residential Building Sector . energy price (Euro/kWh) structure to a combination of a 14 active at home. The night reduction temperature is set Table A.2: Assumed residential energy prices including taxes, levies, for all buildings to 18C. and network charges based on Energieezienzstrategie Gebude [15, 88, 89] while missing parameters are derived from Lindberg et al. [91], KWKG [92], EEG [93]. The gas price is corrected to the observed gas Appendix A.2. Assumptions for 2050 prices in 2016 [90]. The GHG footprint and primary energy factors The techno-economic assumptions for the future en- (PE) are taken from prognos [88]. FiT refers to Feed-in Tari . ergy supply through 2050 are introduced in the following Technology OPEX-var OPEX- x GHG PE Comment section. While many parameters are estimated to stay at a - Euro/kWh Euro/a kg/kWh kWh/kWh - similar magnitude as in the Reference case in Section Ap- Electricity 0.246 170 0.525 1.8 0.292 Euro/kWh supply for 3700 kWh/a pendix A.1, this section describes only the assumptions Gas supply 0.065 0 0.250 1.1 that are changing for the case of 2050. All prices and costs Oil supply 0.064 0 0.320 1.1 Pellet supply 0.060 0 0.014 0.2 are provided as real prices in 2015. Heat pump 0.190 70 0.525 1.8 tari While no major changes are expected for conventional FiT CHP -0.08 0 0.000 2.8 for less than 50 heat generators, further learning rates and cost reductions kWel FiT PV -0.108 0 0.000 1.8 are considered for photovoltaic and electrochemical tech- District 0.074 327 0.270 0.7 0.096 Euro/kWh heating for 15.000 kWh/a nologies, as shown in Table A.4. Their detailed derivation Log supply 0.050 0 0.000 0.2 and discussion is also performed in Kotzur [65]. Table A.4: Change and addition of economic parameters of the en- at price (Euro/a) and an energy price (Euro/kWh). This ergy supply technologies for the year 2050. is important because the savings due to self-consumption, Technology CAPEX CAPEX OPEX Lifetime Source e.g. of photovoltaic electricity, would be overestimated x speci c %CAPEX/a a with the sole energy price. Additionally, this structure Photovoltaic 1000 Euro 650 Euro/kW 1.0 20 [65] el Battery 1000 Euro 300 Euro/kWh 2.0 15 [65] respects that speci c wholesale prices decrease with larger Fuel cell 4000 Euro 1500 Euro/kW 3.0 15 [65] el energy consumptions rates [90]. The technical performance of the technologies is sum- marized in Table A.3. The eciencies are given for the The technical assumptions for 2050 are shown in Ta- Lower Heating Value (LHV) of gas, oil or pellets. The ble A.5. The eciency of the heat pumps is expected to electrical and thermal CHP eciencies are de ned for a increase further in the future [100], for which this work xed operation ratio and cannot be varied in between. assumes an increase of the quality grade to 0.45, which The values are chosen such that the di erent age struc- is the upper bound of today's systems [65]. The photo- tures of the technologies are respected, e.g. an eciency voltaic eciency is assumed to increase to a value of 30% is assumed for the gas boiler that refers to the eciency of [22]. Primarily, this impacts the space coverage on the condensing boilers, while for the oil boiler a lower eciency rooftop and increases the photovoltaic potential that can is considered that is related to older boiler technologies. be installed. The technical parameters of the batteries are derived from a prediction until 2050 [101], but some of today's residential storage systems already achieve similar Table A.3: Summary of the main technical parameters of the energy supply technologies eciencies [102]. Technology Eciency Comment and Reference Table A.5: Summary of the main technical parameters of the energy Gas boiler 0.96 Condensing boiler supply technologies for 2050. [94] Oil boiler 0.84 [95] Pellet boiler 0.9 [37] Technology Eciency Comment and Reference Heat pump dynamic [65] quality grade of 0.4 Heat pump dynamic [65] Heat storage 0.99 charge [37] quality grade of 0.45 0.99 discharge 0.6%/h self-discharge [96] Photovoltaic 0.3 average 2050 [22] Photovoltaic 0.15 based on Hanwha HSL 60 S [97] 2 with 3.5 m /kWp with 7 m /kWp Battery 0.95 charge [101] IC CHP 0.6 thermal [98] 0.95 discharge [101] 0.25 electric [98] Electric heater 0.98 [95] 0.01%/h self-discharge [101] Solar thermal dynamic [37] 0.5 kW/kWh capacity factor Fireplace 0.83 [95, 99] Fuel cell 0.33 thermal [65] 0.52 electric [65] The comfort temperature inside the buildings is as- sumed to have a value of 21C for when occupants are 15 The electrical eciency of the fuel cell is assumed to Table A.7: Techno-economic assumptions for the insulation measures be 52% and positions itself between the eciency that can of a single building. The two measure levels are derived from Schutz  be achieved from Solid Oxide Fuel Cell (SOFC) systems et al. [47] while the exact cost and lambda are taken from BMVBS [105].(* thickness equivalent. ** capital expenditures related only to and the eciency of the Proton Exchange Membrane Fuel energetic measures.) Cells (PEMFC), as discussed in detail in Kotzur [65]. A fully exible operation is assumed for the year 2050. The Component Measure Thickness* Lambda CAPEX CAPEX energy ** 2 2 eciencies are considered to be the same for natural gas, - m W/m/K Euro/m Euro/m biogas or hydrogen as potential alternative fuels [103]. Wall Base 0.15 0.035 124.0 51.5 Future 0.22 0.035 140.9 68.5 The energy prices for 2050 are shown in Table A.6 Roof Base 0.24 0.035 237.6 53.0 and also rely on the Energieezienzstrategie Gebude [15, Future 0.36 0.035 270.0 79.6 88] and Energiereferenzprognose [89]. The Energiereferen- Floor Base 0.08 0.035 51.7 - zprognose considers a carbon price of 76 Euro/ton for the year 2050, which, e.g., increases the gas price by 1.9 ct/kWh. been refurbished anyway, as discussed in Kotzur [65]. The costs relate to the exterior surface area of the building Table A.6: Assumed energy prices, GHG footprints and primary component. Two levels of potential insulation measures energy factors (PE) based on the Energieezienzstrategie Gebude are considered and di er by the thickness of the insulation [15, 88, 89] for 2050. layer, and are referred to as Base and Future. Technology OPEX-var OPEX- x GHG PE Comment The cost for envelope refurbishment measures di ers - Euro/kWh Euro/a kg/kWh kWh/kWh - between buildings that are in the refurbishment cycle and Electricity 0.220 170 0.122 0.4 0.266 Euro/kWh buildings that are not, as discussed in detail in Kotzur [65]. supply for 3700 kWh/a Gas supply 0.096 0 0.250 1.1 The costs of replacing the windows and changing the Bio-methane 0.138 0 0.014 0.2 Oil supply 0.124 0 0.320 1.1 solar and thermal transmittance of the di erent window Pellet supply 0.080 0 0.014 0.2 types are shown in Table A.8 and rely on the BMVBS HP Tarif 0.190 70 0.122 0.4 FiTCHP -0.010 0 0.000 0.4 [105] as well. The costs are speci c to the window area of FiTPV -0.010 0 0.000 0.4 District 0.085 327 0.144 0.5 0.107 Euro/kWh the building. Again, a di erentiation is made between the heating for 15000 kWh/a Base and Future levels. Log supply 0.065 0 0.000 0.2 Table A.8: Techno-economic assumptions for the windows. The transmittance are based on [47] and the cost based on [105] Furthermore, a bio-methane purchase is integrated with a price of 13.8 ct/kWh, which can be either a synthetic Measure Solar transmittance Thermal transmittance CAPEX gas or biogas. As no sucient predictions for bio-methane 2 2 - W/m /K Euro/m prices in 2050 are te be found, its price is derived from the Base 0.575 1.1 313 production cost of bio-methane for the feed-in into the gas Future 0.5 0.7 361.5 grid of 7.5 ct/kWh in 2013 [104], plus the surcharge for grid fees, tax etc. This surcharge is considered to be 6.3 ct/kWh, which is the di erence between the gas market All envelope measures have a lifetime of 40 years with price of 3.3 ct/kWh and the residential gas price of 9.6 zero operational costs. ct/kWh in 2050 [89]. All in all, it results in a price of 13.8 In addition to the conventional refurbishment measures ct/kWh for the bio-methane, which is signi cantly above at the envelope of the building, a heat recovery for the the fossil gas price. ventilation is assumed with a speci c investment of 65 No values for future feed-in tari s were identi ed. There- Euro/m per living area, a lifetime of 25 years and opera- fore, the feed-in is only marginally subsidized, as it is tional cost of 4% per year according to the BMVBS [105] highly dependent on the future market environment. A as a ratio to the original investment. If integrated, 80% of marginal value of 0.01 eur/kWh is chosen in order to guar- the heat losses due to ventilation would be recovered. antee that photovoltaic generation is not curtailed and is Lastly, an occupancy controller can be installed that instead fed-in to the grid. reduces the comfort temperature in case of vacant occu- The cost and energetic impact of the refurbishment pants [65]. Based on the cost of Controme [106], they measures for the opaque building envelope are shown in are assumed with a xed investment of 1000 Euro for the Table A.7. central system controller and 3 Euro/m per living area All measures are additional layers to the existing en- for the di erent thermostats in the rooms, including their velope of the building. The costs are average values taken installation costs. A lifetime of 15 years is assumed. from a survey about subsided refurbishment measures in Germany [105]. They di er between the entire CAPEX of a refurbishment measure and the sole additional CAPEX of energy eciency measures if the building would have 16 17.6 53.9 40.0 103.7 4.7 5.2 69.2 Heat node to Space heat Appendix B. Supplementary results The composition of the total annual residential energy Heat pump to Heat node cost for the Min Cost scenario in 2050 scenario are shown Gas boiler to Heat node in Figure B.20. Floor Electric heater to Heat node Electricity grid Roof Windows 28.4% Fire place to Heat node Walls Control 3.6% Fuel cell to Heat node Total annual Gas boiler 3.4% expenditures: Photovoltaic 0 12.4% 65.0 bil. Euro 0 50 100 150 200 250 300 350 Day Annual expenditures per household: 1603.0 Euro Battery 5.4% Figure B.22: Heat ows of the relevant heat generators to the heat 25.5% Fuel cell 2.1% node and the connected demand for space heating for Germany in Gas supply the Min Cost scenario in 2050. Log supply 11.8% Electric heater Heat pump Heat storage Table B.9: Aggregated energy ows [TWh/a] between the di erent Figure B.20: Composition of the total annual costs over the whole technologies for Value of analysis. of Germany for the Min Cost scenario in 2050. Min No No No No No Forced Cost Gas supply Fuel cell Photovoltaic Heat pump refurbishment refurbishment AC Node to Battery 5.2 2.0 3.7 0.0 5.1 6.1 5.2 AC Node to Building 112.8 112.8 112.8 112.8 112.8 112.8 112.8 AC Node to Electric heater 40.8 56.1 43.3 4.6 48.1 45.0 40.0 The resulting annual energy ows between the di erent AC Node to Heat pump 34.3 26.9 16.2 24.6 0.0 50.1 21.0 AC Node to Hot water 6.8 6.8 6.8 6.8 6.8 6.8 6.8 technologies for the Min Cost scenario in 2050 are illus- Battery to AC Node 4.7 1.8 3.3 0.0 4.6 5.5 4.7 CHP to AC Node 0.0 0.0 0.2 0.0 0.0 0.0 0.0 Cool supply to Building 19.8 21.4 19.6 19.0 18.5 29.0 31.6 trated in Figure B.21 for the aggregated level of the whole Electricity supply to AC Node 51.5 100.5 84.7 63.5 57.7 47.3 52.7 Fuel cell to AC Node 53.9 0.0 0.0 85.3 20.1 69.1 45.7 of Germany. Gas supply to CHP 0.0 0.0 0.7 0.0 0.0 0.0 0.0 Gas supply to Fuel cell 103.7 0.0 0.0 164.1 38.6 132.9 87.9 Gas supply to Gas boiler 68.7 0.0 91.2 121.6 226.6 26.8 65.7 HP Tarif to Heat pump 20.4 46.2 40.2 17.4 0.0 42.1 10.5 Log supply to Fire place 21.2 19.4 23.0 23.6 51.4 21.3 28.8 21.2 Pellet supply to Pellet boiler 0.0 10.3 0.0 0.0 3.1 0.0 0.0 Log supply Fire place District heating Renewable gas to Gas boiler 0.0 8.7 0.0 0.0 0.0 0.0 0.0 Heat pump to Building 225.1 296.5 231.0 169.5 0.0 363.9 126.7 CHP to HNode 0.0 0.0 0.4 0.0 0.0 0.0 0.0 District heating to HNode 0.0 7.9 0.8 0.0 15.0 0.0 0.0 Electric heater to HNode 40.0 55.0 42.5 4.5 47.1 44.1 39.2 Pellet supply Pellet boiler Solar thermal Fire place to HNode 17.6 16.1 19.1 19.6 42.7 17.7 23.9 Fuel cell to HNode 34.2 0.0 0.0 54.2 12.7 43.9 29.0 Gas boiler to HNode 66.0 8.3 87.6 116.7 217.5 25.7 63.1 HNode to Building 84.7 23.4 78.1 123.3 265.8 57.4 82.4 Oil supply Oil boiler Heat storage HNode to Hot water 69.2 69.2 69.2 69.2 69.2 69.2 69.2 Pellet boiler to HNode 0.0 9.3 0.0 0.0 2.8 0.0 0.0 68.7 66.0 309.8 HNode to Heat storage 54.5 41.8 38.3 77.2 52.2 59.6 52.9 Heat storage to HNode 50.5 37.8 35.2 74.5 49.2 54.8 49.3 Gas supply Gas boiler Space heat Photovoltaic to AC Node 89.7 102.4 94.7 0.0 90.4 98.9 82.7 Photovoltaic to FiTPV 24.1 33.9 26.4 0.0 19.2 27.3 20.7 Solar thermal to HNode 0.1 0.0 0.0 0.1 0.1 0.0 0.0 Renewable gas Fuel cell CHP Heat pump Electric heater Hot water FiTCHP HP Tarif 51.5 112.8 References Electricity supply Electricity 24.1 [1] IEA, CO2 Emissions from Fuel Combustion - 2015 Edition, FiTPV Photovoltaic Battery Technical Report, International Energy Agency, 2015. [2] AGEB, Auswertungstabellen zur energiebilanz deutschland Figure B.21: Annual energy ows in TWh between the di erent 1990 bis 2016, 2017. technologies aggregated for the whole of Germany for the Min Cost [3] RWI, Erstellung der Anwendungsbilanzen 2015 und 2016 fur  scenario den Sektor der Privaten Haushalte und den Verkehrssektor in Deutschland, Technical Report, RWI Leibniz-Institut fur  Wirtschaftsforschung e.V., 2017. [4] BMWi, Zahlen und Fakten Energiedaten, Technical Report, The aggregated temporal operation of heat technolo- Bundesministerium fur  Wirtschaft und Energie, 2016. gies in the Min Cost scenario in 2050 are illustrated in [5] UN, Paris agreement - cop21, 2015. Figure B.22. [6] S. Solomon, G.-K. Plattner, R. Knutti, P. Friedlingstein, Irre- versible climate change due to carbon dioxide emissions, Pro- The aggregated annual energy ows between the di er- ceedings of the national academy of sciences 106 (2009) 1704{ ent considered system components are listed in Table B.9 for the di erent scenarios in 2050. [7] EU, 2010. URL: https://eur-lex.europa.eu/ Table B.10 shows the aggregated installed capacities legal-content/en/TXT/?uri=CELEX:32010L0031. [8] EU, 2012. URL: http://eur-lex.europa.eu/legal-content/ for all residential buildings for the di erent sensitivity cases. EN/TXT/PDF/?uri=CELEX:32012L0027&from=DE. [9] REHVA, How to de ne nearly net zero energy build- ings nZEB, Technical Report, 2011. URL: https: 20.4 6.8 50.5 54.5 34.2 40.8 0.1 89.7 34.3 225.1 Hour Hour Hour Hour Hour Hour GW GW GW GW GW GW [23] B. Nykvist, M. Nilsson, Rapidly falling costs of battery packs Table B.10: Aggregated installed capacities of the di erent technolo- for electricvehicles, Nature Climate Change 5 (2015) 329{332. gies for the Value of analysis. doi:10.1038/nclimate2564. [24] C. Breyer, A. Gerlach, Global overview on grid-parity, Min No No No No No Forced Cost Gas supply Fuel cell Photovoltaic Heat pump refurbishment refurbishment Progress in Photovoltaics: Research and Applications 21 Gas boiler [GW ] 36.6 5.1 47.6 83.9 111.9 14.7 36.3 th (2013) 121{136. doi:10.1002/pip.1254. Oil boiler [GWth] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 CHP [GW ] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 el [25] K. Mainzer, K. Fath, R. McKenna, J. Stengel, W. Ficht- District heating [GW ] 0.0 7.5 0.9 0.0 13.1 0.0 0.0 th Heat storage [GWh ] 215.6 291.4 216.6 63.4 173.9 265.4 188.6 ner, F. Schultmann, A high-resolution determination of the th Heat pump [GWth] 60.4 83.1 60.4 47.0 0.0 95.5 34.6 technical potential for residential-roof-mounted photovoltaic Electric heater [GW ] 99.7 97.1 90.3 83.6 75.4 121.0 84.4 th Solar thermal [GW ] 0.2 0.0 0.0 0.2 0.2 0.0 0.0 th systems in Germany, Solar Energy 105 (2014) 715{731. Pellet boiler [GW ] 0.0 2.4 0.0 0.0 0.8 0.0 0.0 th Battery [GWhel] 16.9 7.1 12.8 0.0 17.1 20.0 17.4 doi:10.1016/j.solener.2014.04.015. Fuel cell [GW ] 12.7 0.0 0.0 14.5 5.1 16.9 10.7 el Photovoltaic [GW ] 133.4 160.3 142.3 0.0 127.8 147.6 121.0 el [26] ISI, Netzentwicklungsplan Strom - Entwicklung der re- gionalen Stromnachfrage und Lastpro le, Technical Report, Fraunhofer-Institut fur  System- und Innovationsforschung, [27] M. G. Beer, Regionalisiertes Energiemodell zur Analyse der //www.rehva.eu/fileadmin/hvac-dictio/03-2011/How_ exiblen Betriebsweise von Kraft-W arme-Kopplungsanlagen, to_define_nearly_net_zero_energy_buildings_nZEB.pdf. Thesis, 2012. [10] A. J. Marszal, P. Heiselberg, J. S. Bourrelle, E. Musall, [28] P. Seljom, K. B. Lindberg, A. Tomasgard, G. Doorman, K. Voss, I. Sartori, A. Napolitano, Zero Energy Building {A re- I. Sartori, The impact of Zero Energy Buildings on the view of de nitions and calculation methodologies, Energy and Scandinavian energy system, Energy 118 (2017) 284{296. Buildings 43 (2011) 971{979. doi:10.1016/j.enbuild.2010. doi:10.1016/j.energy.2016.12.008. 12.022. [29] P. Balcombe, D. Rigby, A. Azapagic, Investigating the im- [11] J. OCallaghan, S. Mezger, M. Coughlin, K. Isles, A. Welsh, portance of motivations and barriers related to microgenera- C. K. Lyttle, Utility of the future - A customer-led shift in the tion uptake in the UK, Applied Energy 130 (2014) 403{418. electricity sector, Technical Report, 2014. doi:10.1016/j.apenergy.2014.05.047. [12] S. Agnew, P. Dargusch, E ect of residential solar and stor- [30] M. Achtnicht, R. Madlener, Factors in uencing German house age on centralized electricity supply systems, Nature Climate owners'preferences on energy retro ts, Energy Policy 68 (2014) Change 5 (2015) 315{318. doi:10.1038/nclimate2523. 254{263. doi:10.1016/j.enpol.2014.01.006. [13] W. Rickerson, T. Couture, G. Barbose, D. Jacobs, G. Parkin- [31] C. Milan, C. Bojesen, M. P. Nielsen, A cost optimization model son, E. Chessin, A. Belden, H. Wilson, H. Barrett, Residen- for 100% renewable residential energy supply systems, Energy tial prosumers - Drivers and policy options, Technical Report, 48 (2012) 118{127. doi:10.1016/j.energy.2012.05.034. [32] D. Lauinger, P. Caliandro, J. Van herle, D. Kuhn, A lin- [14] R. McKenna, E. Merkel, D. Fehrenbach, S. Mehne, W. Ficht- ear programming approach to the optimization of residential ner, Energy eciency in the German residential sector: A energy systems, Journal of Energy Storage 7 (2016) 24{37. bottom-up building-stock-model-based analysis in the context doi:10.1016/j.est.2016.04.009. of energy-political targets, Building and Environment 62 [33] A. Ashouri, S. S. Fux, M. J. Benz, L. Guzzella, Optimal design (2013) 77{88. doi:10.1016/j.buildenv.2013.01.002. and operation of building services using mixed-integer linear [15] BMWi, Energieezienzstrategie Geb aude, Technical Report, programming techniques, Energy 59 (2013) 365{376. doi:10. 1016/j.energy.2013.06.053. [16] N. Diefenbach, T. Loga, B. Stein, Reaching the climate pro- [34] E. D. Mehleri, H. Sarimveis, N. C. Markatos, L. G. Papageor- tection targets for the heat supply of the German residential giou, Optimal design and operation of distributed energy sys- building stock: How and how fast?, Energy and Buildings 132 tems: Application to Greek residential sector, Renewable En- (2016) 53{73. doi:10.1016/j.enbuild.2016.06.095. ergy 51 (2013) 331{342. doi:10.1016/j.renene.2012.09.009. [17] Beuth, Ableitung eines Korridors fur  den Ausbau der erneuer- [35] S. Fazlollahi, P. Mandel, G. Becker, F. Mar echal, Methods baren W arme im Geb audebereich, Technical Report, Beuth for multi-objective investment and operating optimization of Hoschule fur  Technik, ifeu - Institut fur  Energie- und Umwelt- complex energy systems, Energy 45 (2012) 12{22. doi:10.1016/ forsschung Heidelberg GmbH, 2017. j.energy.2012.02.046. [18] BMWi, Energiedaten - Tabelle 30: Interna- [36] H. Harb, J.-N. Paprott, P. Matthes, T. Schutz,  R. Streblow, tionaler Preisvergleich, Technical Report, 2018. D. Muller,  Decentralized scheduling strategy of heating sys- URL: https://www.bmwi.de/Redaktion/DE/Binaer/ tems for balancing the residual load, Building and Environ- Energiedaten/Energiepreise-und-Energiekosten/ ment 86 (2015) 132{140. doi:10.1016/j.buildenv.2014.12. energiedaten-energiepreise-5-xls.xls?__blob= publicationFile&v=17. [37] K. B. Lindberg, G. Doorman, D. Fischer, M. Korp as, [19] IWES, Interaktion EE-Strom, W arme und Verkehr, Tech- A. Anestad, I. Sartori, Methodology for optimal energy sys- nical Report, Fraunhofer-Institut fur  Windenergie und tem design of Zero Energy Buildings using mixed-integer lin- Energiesystemtechnik (Fraunhofer IWES), 2015. URL: ear programming, Energy and Buildings 127 (2016) 194{205. https://www.iee.fraunhofer.de/content/dam/iwes-neu/ doi:10.1016/j.enbuild.2016.05.039. energiesystemtechnik/de/Dokumente/Veroeffentlichungen/ [38] G. Cardoso, M. Stadler, S. Mashayekh, E. Hartvigsson, The 2015/Interaktion_EEStrom_Waerme_Verkehr_Endbericht. impact of ancillary services in optimal DER investment deci- pdf. sions, Energy 130 (2017) 99{112. doi:10.1016/j.energy.2017. [20] UBA, Datenbasis zur Bewertung von Energieezienzmanah- 04.124. men in der Zeitreihe 2005 {2014, Technical Report, Umwelt- [39] T. Schutz,  M. H. Schraven, S. Remy, J. Granacher, bundesamt, 2017. D. Kemetmuller,  M. Fuchs, D. Muller,  Optimal design of [21] H. Lund, A. N. Andersen, P. A. stergaard, B. V. Mathiesen, energy conversion units for residential buildings considering D. Connolly, From electricity smart grids to smart energy sys- German market conditions, Energy 139 (2017) 895{915. tems {A market operation based approach and understanding, doi:10.1016/j.energy.2017.08.024. Energy 42 (2012) 96{102. doi:10.1016/j.energy.2012.04.003. [40] M. Hamdy, A. Hasan, K. Siren, A multi-stage optimization [22] ISE, Current and Future Cost of Photovoltaics. Long-term Sce- method for cost-optimal and nearly-zero-energy building solu- narios for Market Development, System Prices and LCOE of tions in line with the EPBD-recast 2010, Energy and Buildings Utility-Scale PV Systems, Technical Report, Study on behalf 56 (2013) 189{203. doi:10.1016/j.enbuild.2012.08.023. of Agora Energiewende, 2015. 18 [41] R. Evins, Multi-level optimization of building design, energy 648{662. doi:10.1016/j.compchemeng.2013.10.016. system sizing and operation, Energy 90 (2015) 1775{1789. [60] J. A. Fonseca, A. Schlueter, Integrated model for characteriza- doi:10.1016/j.energy.2015.07.007. tion of spatiotemporal building energy consumption patterns [42] M. Stadler, G. Cardoso, S. Mashayekh, T. Forget, N. DeForest, in neighborhoods and city districts, Applied Energy 142 (2015) A. Agarwal, A. Sch onbein, Value streams in microgrids: A 247{265. doi:10.1016/j.apenergy.2014.12.068. literature review, Applied Energy 162 (2016) 980{989. doi:10. [61] E. McKenna, M. Thomson, High-resolution stochastic inte- 1016/j.apenergy.2015.10.081. grated thermal{electrical domestic demand model, Applied [43] X. Han, H. Zhang, X. Yu, L. Wang, Economic evaluation of Energy 165 (2016) 445{461. doi:10.1016/j.apenergy.2015. grid-connected micro-grid system with photovoltaic and en- 12.089. ergy storage under di erent investment and nancing models, [62] I. Richardson, M. Thomson, D. In eld, C. Cli ord, Domes- Applied Energy 184 (2016) 103{118. doi:10.1016/j.apenergy. tic electricity use: A high-resolution energy demand model, 2016.10.008. Energy and Buildings 42 (2010) 1878{1887. doi:10.1016/j. [44] R. Streblow, K. Ansorge, Genetischer Algorithmuszur kombi- enbuild.2010.05.023. natorischen Optimierung von Geb audehulle  und Anlagentech- [63] I. Richardson, M. Thomson, D. In eld, A. Delahunty, Domes- nik, Technical Report, 2017. tic lighting: A high-resolution energy demand model, Energy [45] R. Wu, G. Mavromatidis, K. Orehounig, J. Carmeliet, Multi- and Buildings 41 (2009) 781{789. doi:10.1016/j.enbuild. objective optimisation of energy systems and building envelope 2009.02.010. retro t in a residential community, Applied Energy 190 (2017) [64] I. Richardson, M. Thomson, D. In eld, A high-resolution do- 634{649. doi:10.1016/j.apenergy.2016.12.161. mestic building occupancy model for energy demand simula- [46] M. Stadler, M. Groissb ock, G. Cardoso, C. Marnay, Optimiz- tions, Energy and Buildings 40 (2008) 1560{1566. doi:10.1016/ ing Distributed Energy Resources and building retro ts with j.enbuild.2008.02.006. the strategic DER-CAModel, Applied Energy 132 (2014) 557{ [65] L. Kotzur, Future Grid Load of the Residential Building 567. doi:10.1016/j.apenergy.2014.07.041. Sector, Thesis, 2018. URL: http://juser.fz-juelich.de/ [47] T. Schutz,  L. Schi er, H. Harb, M. Fuchs, D. Muller,  Optimal record/858675. design of energy conversion units and envelopes for residential [66] D. EN ISO, Thermal performance and energy use in the build building retro ts using a comprehensive MILP model, Ap- environment (ISO 13790:2008) (2008). plied Energy 185 (2017) 1{15. doi:10.1016/j.apenergy.2016. [67] R. W. Andrews, J. S. Stein, C. Hansen, D. Riley, Introduction 10.049. to the Open Source PV-LIB for Python Photovoltaic System [48] M. Kavgic, A. Mavrogianni, D. Mumovic, A. Summer eld, Modeling Package (2014). Z. Stevanovic, M. Djurovic-Petrovic, A review of bottom-up [68] DIN, Din en 12831 - heating systems and water based cooling building stock models for energy consumption in the residen- systems in buildings method for calculation of the design heat tial sector, Building and Environment 45 (2010) 1683{1697. load, 2014. doi:10.1016/j.buildenv.2010.01.021. [69] DWD, Testreferenzjahre (TRY) (2012). [49] S. P. Corgnati, E. Fabrizio, M. Filippi, V. Monetti, Reference [70] C. Bollmeyer, J. D. Keller, C. Ohlwein, S. Wahl, S. Crewell, buildings for cost optimal analysis: Method of de nition and P. Friederichs, A. Hense, J. Keune, S. Kneifel, I. Pscheidt, application, Applied Energy 102 (2013) 983{993. doi:10.1016/ S. Redl, S. Steinke, Towards a high-resolution regional re- j.apenergy.2012.06.001. analysis for the European CORDEX domain, Quarterly Jour- [50] E. Mata, A. Sasic Kalagasidis, F. Johnsson, Building-stock nal of the Royal Meteorological Society 141 (2015) 1{15. aggregation through archetype buildings: France, Germany, doi:10.1002/qj.2486. Spain and the UK, Building and Environment 81 (2014) 270{ [71] L. Welder, D. S. Ryberg, L. Kotzur, T. Grube, M. Robinius, 282. doi:10.1016/j.buildenv.2014.06.013. D. Stolten, Spatio-temporal optimization of a future energy [51] IWU, Deutsche Gebaeudetypologie. Systematik und Daten- system for power-to-hydrogen applications in Germany, En- saetze, Technical Report, Institut Wohnen und Umwelt ergy 158 (2018) 1130{1149. doi:10.1016/j.energy.2018.05. (IWU), 2005. URL: http://www.iwu.de/fileadmin/user_ 059. upload/dateien/energie/klima_altbau/Gebaeudetypologie_ [72] L. Kotzur, P. Markewitz, M. Robinius, D. Stolten, Impact of Deutschland.pdf. di erent time series aggregation methods on optimal energy [52] IWU, Datenbasis Geb audebestand - Datenerhebung zur en- system design, Renewable Energy 117 (2018) 474{487. doi:10. ergetischen Qualit at und zu den Modernisierungstrends im 1016/j.renene.2017.10.017. deutschen Wohngeb audebestand, Technical Report, Institut [73] L. Kotzur, P. Markewitz, M. Robinius, D. Stolten, Time se- Wohnen und Umwelt Bremer Energie Institut, 2010. ries aggregation for energy system design: Modeling seasonal [53] EPISCOPE, 2016. URL: http://episcope.eu/iee-project/ storage, Applied Energy 213 (2018) 123{135. doi:10.1016/j. episcope/. apenergy.2018.01.023. [54] R. Hendron, C. Engebrecht, Building America Research [74] T. Kannengieer, M. Ho mann, L. Kotzur, P. Stenzel, Benchmark De nition, Technical Report, National Renewable P. Markewitz, F. Schutz  , K. Peters, S. Nykamp, D. Stolten, Energy Laboratory, 2010. URL: https://www.nrel.gov/docs/ M. Robinius, Reducing computational load for mixed inte- fy10osti/47246.pdf. ger linear programming: An example for a district and an [55] P. Torcellini, M. Deru, B. Grith, K. Benne, Doe commercial island energy system (2019). URL: https://www.preprints. building benchmark models, 2008. org/manuscript/201905.0116/download/final_file. [56] G. DallO, A. Galante, G. Pasetti, A methodology for eval- [75] B. Bahl, A. Kump  el, H. Seele, M. Lampe, A. Bardow, Time- uating the potential energy savings of retro tting residential series aggregation for synthesis problems by bounding error in building stocks, Sustainable Cities and Society 4 (2012) 12{21. the objective function, Energy 135 (2017) 900{912. doi:10. doi:10.1016/j.scs.2012.01.004. 1016/j.energy.2017.06.082. [57] C. Cerezo, J. Sokol, C. Reinhart, A. Al-Mumin, 2015. URL: [76] S. Bundesamt, 2011. URL: https://www.zensus2011.de/DE/ http://www.ibpsa.org/proceedings/BS2015/p2435.pdf. Home/home_node.html. [58] J. Sokol, C. Cerezo Davila, C. F. Reinhart, Validation of [77] Z. Huang, Extensions to the k-means algorithm for clus- a Bayesian-based method for de ning residential archetypes tering large data sets with categorical values, 1998. in urban building energy models, Energy and Buildings 134 URL: https://link.springer.com/content/pdf/10.1023/A: (2017) 11{24. doi:10.1016/j.enbuild.2016.10.050. 1009769707641.pdf. [59] S. Fazlollahi, G. Becker, F. Mar echal, Multi-objectives, multi- [78] S. Lloyd, Least squares quantization in PCM, IEEE Transac- period optimization of district energy systems: II-Daily ther- tions on Information Theory 28 (1982) 129{137. doi:10.1109/ mal storage, Computers & Chemical Engineering 71 (2014) TIT.1982.1056489. 19 [79] A. K. Jain, Data clustering: 50 years beyond K-means, Pat- Procedure to Determine Coecients for the Sandia Array Per- tern Recognition Letters 31 (2010) 651{666. doi:10.1016/j. formance Model, Technical Report, Sandia National Labora- patrec.2009.09.011. tories, 2016. [80] A. Druckman, T. Jackson, Household energy consumption in [98] ASUE, BHKW-Kenndaten 2014/2015 - Module, Anbieter, the UK: A highly geographically and socio-economically dis- Kosten, Technical Report, Arbeitsgemeinschaft fur  sparsamen aggregated model, Energy Policy 36 (2008) 3177{3192. und umweltfreundlichen Energieverbrauch e.V., 2015. [81] A. Elnakat, J. D. Gomez, N. Booth, A zip code study of so- [99] Olsberg, Produktdatenblatt Olsberg Palena Cmpact, Tech- cioeconomic, demographic, and household gendered in uence nical Report, 2018. URL: https://www.olsberg.com/ on the residential energy sector, Energy Reports 2 (2016) 21{ 578912-wee-de-wAssets/docs/kaminoefen/palena-compact/ 27. doi:10.1016/j.egyr.2016.01.003. anleitungen-zertifikate-ersatzteillisten/ [82] S. A. Robinson, V. Rai, Determinants of spatio-temporal pat- Produktdatenblatt_Palena-Compact_23-575_ohne-OEC.pdf. terns of energy technology adoption: An agent-based modeling [100] H. Willem, Y. Lin, A. Lekov, Review of energy eciency and approach, Applied Energy 151 (2015) 273{284. doi:10.1016/ system performance of residential heat pump water heaters, j.apenergy.2015.04.071. Energy and Buildings 143 (2017) 191{201. doi:10.1016/j. [83] L. Kranzl, M. Hummel, A. Muller,  J. Steinbach, Renewable enbuild.2017.02.023. heating: Perspectives and the impact of policy instruments, [101] P. Elsner, D. U. Sauer, Energiespeicher - Technologiesteckbrief Energy Policy 59 (2013) 44{58. doi:10.1016/j.enpol.2013. zur Analyse "Flexibilit atskonzepte fur  die Stromversorgung 03.050. 2050", Technical Report, 2015. [84] A. Mul  ler, Energy Demand Assessment for Space Condition- [102] K. Kairies, D. Haberschusz, J. v. Ouwerkerk, J. Strebel, ing and Domestic Hot Water: A Case Study for the Austrian O. Wessels, D. Magnor, J. Badeda, D. U. Sauer, Wis- Building Stock, Thesis, 2015. URL: http://www.invert.at/ senschaftliches Mess- und Evaluierungsprogramm Solarstrom- Dateien/Dissertation_AndreasM.pdf. speicher - Jahresbericht 2016, Technical Report, Institut [85] J. Schleich, X. Gassmann, C. Faure, T. Meissner, Making fur  Stromrichtertechnik und Elektrische Antriebe der RWTH the implicit explicit: A look inside the implicit discount rate, Aachen, 2016. Energy Policy 97 (2016) 321{331. doi:10.1016/j.enpol.2016. [103] R. Peters, R. Deja, M. Engelbracht, M. Frank, V. N. Nguyen, 07.044. L. Blum, D. Stolten, Eciency analysis of a hydrogen-fueled [86] A.-L. Klingler, Self-consumption with PV + Battery systems: solid oxide fuel cell system with anode o -gas recirculation, A market di usion model considering individual consumer be- Journal of Power Sources 328 (2016) 105{113. doi:10.1016/j. haviour and preferences, Applied Energy 205 (2017) 1560{ jpowsour.2016.08.002. 1570. doi:10.1016/j.apenergy.2017.08.159. [104] Bundesnetzagentur, Biogas- Monitoringbericht 2014, Tech- [87] A. Lahnaoui, P. Stenzel, J. Linssen, Techno-economic analysis nical Report, 2014. URL: https://www.bundesnetzagentur. of photovoltaic battery system con guration and location, Ap- de/SharedDocs/Downloads/DE/Sachgebiete/Energie/ plied Energy (2017). doi:10.1016/j.apenergy.2017.09.093. Unternehmen_Institutionen/ErneuerbareEnergien/Biogas/ [88] prognos, Hintergrundpapier Energieezienzstrategie Geb aude, Biogas_Monitoring/Biogas_Monitoringbericht_2014.pdf?__ Technical Report, 2015. blob=publicationFile&v=1. [89] EWI, Entwicklung der Energiemaerkte - Energiereferenzprog- [105] BMVBS, Kosten energierelevanter Bau- und Anlagenteile bei nose, Technical Report, On behalf of the Federal Ministry for der energetischen Modernisierung von Wohngeb auden, Tech- Economic A airs and Energy (BMWi), 2014. nical Report, 2012. URL: http://www.bbsr.bund.de/BBSR/ [90] Bundesnetzagentur, Monitoringbericht 2017, Technical Re- DE/Veroeffentlichungen/BMVBS/Online/2012/DL_ON072012. port, 2017. URL: https://www.bundesnetzagentur.de/ pdf?__blob=publicationFile&v=2. SharedDocs/Downloads/DE/Allgemeines/Bundesnetzagentur/ [106] Controme, Controme - Fubodenheizung Set-Funk, Techni- Publikationen/Berichte/2017/Monitoringbericht_2017. cal Report, 2018. URL: https://www.controme.com/produkt/ pdf?__blob=publicationFile&v=4. fussbodenheizung-set-funk-2/. [91] K. B. Lindberg, D. Fischer, G. Doorman, M. Korp as, I. Sar- tori, Cost-optimal energy system design in Zero Energy Build- ings with resulting grid impact: A case study of a German multi-family house, Energy and Buildings 127 (2016) 830{845. doi:10.1016/j.enbuild.2016.05.063. [92] KWKG, Gesetz ur  die erhaltung, die modernisierung und den ausbau der kraft-w arme-kopplung, 2016. URL: https://www. gesetze-im-internet.de/. [93] EEG, Gesetz fur  den ausbau erneuerbarer energien, 2017. URL: https://www.gesetze-im-internet.de/. [94] H.-M. Henning, A. Palzer, A comprehensive model for the German electricity and heat sector in a future energy system with a dominant contribution from renewable energy technolo- gies|Part I: Methodology, Renewable and Sustainable Energy Reviews 30 (2014) 1003{1018. doi:10.1016/j.rser.2013.09. [95] UBA, Klimaneutraler Geb audebestand 2050 - En- ergieezienzpotentiale und die Auswirkungen des Klimawandels auf den Geb audebestand, Techni- cal Report, Umweltbundesamt, 2017. URL: https: //www.umweltbundesamt.de/sites/default/files/medien/ 1410/publikationen/2017-11-06_climate-change_26-2017_ klimaneutraler-gebaeudebestand-ii.pdf. [96] T. Schutz,  R. Streblow, D. Mulle  r, A comparison of thermal energy storage models for building energy system optimiza- tion, Energy and Buildings 93 (2015) 23{31. doi:10.1016/j. enbuild.2015.02.031. [97] B. H. King, C. W. Hansen, D. Riley, C. D. Robinson, L. Pratt,

Journal

PhysicsarXiv (Cornell University)

Published: Jan 2, 2020

References