Access the full text.
Sign up today, get DeepDyve free for 14 days.
F. Katiraei, R. Iravani, N. Hatziargyriou, A. Dimeas (2008)
Microgrids managementIEEE Power and Energy Magazine, 6
M. Hajiesmaili, Chi-Kin Chau, Minghua Chen, Longbu Huang (2016)
Online microgrid energy generation scheduling revisited: the benefits of randomization and interval predictionProceedings of the Seventh International Conference on Future Energy Systems
Xiangyu Kong, Linquan Bai, Qinran Hu, F. Li, Chengshan Wang (2016)
Day-ahead optimal scheduling method for grid-connected microgrid based on energy storage control strategyJournal of Modern Power Systems and Clean Energy, 4
Qing Li, Zhao Xu, Li Yang (2014)
Recent advancements on the development of microgridsJournal of Modern Power Systems and Clean Energy, 2
Wenbo Shi, Na Li, C. Chu, R. Gadh (2017)
Real-Time Energy Management in MicrogridsIEEE Transactions on Smart Grid, 8
Weicong Kong, Z. Dong, Youwei Jia, D. Hill, Yan Xu, Yuan Zhang (2019)
Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural NetworkIEEE Transactions on Smart Grid, 10
A. Khodaei (2014)
Resiliency-Oriented Microgrid Optimal SchedulingIEEE Transactions on Smart Grid, 5
Peng Li, Pengfei Han, Shuai He, Xiaoxiao Wang (2017)
Double-uncertainty optimal operation of hybrid AC/DC microgrids with high proportion of intermittent energy sourcesJournal of Modern Power Systems and Clean Energy, 5
A. Gholami, Tohid Shekari, F. Aminifar, M. Shahidehpour (2016)
Microgrid Scheduling With Uncertainty: The Quest for ResilienceIEEE Transactions on Smart Grid, 7
Yue Xiang, Junyong Liu, Yilu Liu (2016)
Robust Energy Management of Microgrid With Uncertain Renewable Generation and LoadIEEE Transactions on Smart Grid, 7
F. Farzan, M. Jafari, R. Masiello, Yan Lu (2015)
Toward Optimal Day-Ahead Scheduling and Operation Control of Microgrids Under UncertaintyIEEE Transactions on Smart Grid, 6
C. Papadimitriou, V. Kleftakis, N. Hatziargyriou (2017)
Control strategy for seamless transition from islanded to interconnected operation mode of microgridsJournal of Modern Power Systems and Clean Energy, 5
Lian Lu, Jinlong Tu, Chi-Kin Chau, Minghua Chen, Zhao Xu, Xiaojun Lin (2013)
Towards real-time energy generation scheduling in microgrids with performance guarantee2013 IEEE Power & Energy Society General Meeting
Youwei Jia, Yufei He, Xue Lyu, Songjian Chai, Zhao Xu, Minghua Chen (2018)
Hardware-in-the-loop Implementation of Residential Intelligent Microgrid2018 IEEE Power & Energy Society General Meeting (PESGM)
S. Haider, Guojie Li, Keyou Wang (2018)
A dual control strategy for power sharing improvement in islanded mode of AC microgridProtection and Control of Modern Power Systems, 3
W. Gu, Zhihe Wang, Zhi Wu, Zhao Luo, Yiyuan Tang, J. Wang (2017)
An Online Optimal Dispatch Schedule for CCHP Microgrids Based on Model Predictive ControlIEEE Transactions on Smart Grid, 8
Zhao Luo, W. Gu, Zhi Wu, Zhihe Wang, Yiyuan Tang (2018)
A robust optimization method for energy management of CCHP microgridJournal of Modern Power Systems and Clean Energy, 6
V. Mohan, J. Singh, W. Ongsakul, M.P. Suresh (2016)
Performance enhancement of online energy scheduling in a radial utility distribution microgridInternational Journal of Electrical Power & Energy Systems, 79
Lian Lu, Jinlong Tu, Chi-Kin Chau, Minghua Chen, Xiaojun Lin (2012)
Online energy generation scheduling for microgrids with intermittent energy sources and co-generation
L. Yue (1987)
Solar power generation
Guodong Liu, M. Starke, B. Xiao, K. Tomsovic (2017)
Robust optimisation-based microgrid scheduling with islanding constraintsIet Generation Transmission & Distribution, 11
H. Cai, Qiyu Chen, Zhijian Guan, Jun-hui Huang (2018)
Day-ahead optimal charging/discharging scheduling for electric vehicles in microgridsProtection and Control of Modern Power Systems, 3
F. García-Torres, C. Bordons (2015)
Optimal Economical Schedule of Hydrogen-Based Microgrids With Hybrid Storage Using Model Predictive ControlIEEE Transactions on Industrial Electronics, 62
Yu Zhang, Nikolaos Gatsis, G. Giannakis (2012)
Robust Energy Management for Microgrids With High-Penetration RenewablesIEEE Transactions on Sustainable Energy, 4
B. Kroposki (2017)
Integrating high levels of variable renewable energy into electric power systemsJournal of Modern Power Systems and Clean Energy, 5
J. Leung, Laurie Kelly, James Anderson (2004)
Handbook of Scheduling: Algorithms, Models, and Performance Analysis
Jingjie Yang, Bingqing Guo, B. Qu (2018)
Economic optimization on two time scales for a hybrid energy system based on virtual storageJournal of Modern Power Systems and Clean Energy, 6
Wencong Su, Jianhui Wang, J. Roh (2014)
Stochastic Energy Scheduling in Microgrids With Intermittent Renewable Energy ResourcesIEEE Transactions on Smart Grid, 5
S. Dutta, P. Sadhu, M. Reddy, D. Mohanta (2018)
Shifting of research trends in islanding detection method - a comprehensive surveyProtection and Control of Modern Power Systems, 3
C. Wan, Jian Zhao, Yonghua Song, Zhao Xu, Jin Lin, Zechun Hu (2015)
Photovoltaic and solar power forecasting for smart grid energy managementCSEE Journal of Power and Energy Systems, 1
Yan Du, W. Pei, Naishi Chen, X. Ge, Hao Xiao (2017)
Real-time microgrid economic dispatch based on model predictive control strategyJournal of Modern Power Systems and Clean Energy, 5
Wei Liu, W. Gu, Xiaodong Yuan, Kaifeng Zhang (2018)
Fully distributed control to coordinate charging efficiencies for energy storage systemsJournal of Modern Power Systems and Clean Energy, 6
B.S. Gisin, Q. Gu, J. Mitsche, Simon Tam, Hong Chen (2010)
“Perfect Dispatch” - as the measure of PJM real time grid operational performanceIEEE PES General Meeting
(2015)
he was a postdoctoral fellow at The Hong Kong Polytechnic University. He is currently an Assistant Professor with the Department of Electrical and Electronic Engineering
J. Mod. Power Syst. Clean Energy (2019) 7(6):1608–1618 https://doi.org/10.1007/s40565-019-00574-2 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1 1,2 3,4 2 Youwei JIA , Xue LYU , Chun Sing LAI , Zhao XU , Minghua CHEN Abstract As an emerging paradigm in distributed power were conducted using 3-year-long real-life system data, systems, microgrids provide promising solutions to local and the results of simulations show that the cost difference renewable energy generation and load demand satisfaction. between the proposed retroactive approach and perfect However, the intermittency of renewables and temporal dispatch is less than 11% on average, which suggests better uncertainty in electrical load create great challenges to performance than model predictive control with the cost energy scheduling, especially for small-scale microgrids. difference at 30% compared to the perfect dispatch. Instead of deploying stochastic models to cope with such challenges, this paper presents a retroactive approach to Keywords Microgrid, Renewables, Real-time scheduling, real-time energy scheduling, which is prediction-indepen- Perfect dispatch, Grid integration dent and computationally efficient. Extensive case studies 1 Introduction CrossCheck date: 19 September 2019 Received: 18 November 2018 / Accepted: 19 September 2019 / Recently, modern electricity supply systems have been Published online: 16 November 2019 undergoing a dramatic transition toward decentralized, The Author(s) 2019 decarbonized, and democratized power systems. Driven by & Youwei JIA this trend, microgrids have emerged as a promising solu- jiayw@sustech.edu.cn tion for settling the generation-demand balance locally. In Xue LYU general, a microgrid integrates distributed energy resources xue.lyu@connect.polyu.hk (DERs), including non-dispatchable renewables and dis- Chun Sing LAI patchable local generators, and flexible operation modes, c.s.lai@leeds.ac.uk i.e., grid-connected or stand-alone modes [1–3]. To realize Zhao XU the economic operation of microgrids, it is indispensable to eezhaoxu@polyu.edu.hk optimize energy scheduling, for coordinating the internal Minghua CHEN DERs of microgrids and an external grid, to minimize the minghua@ie.cuhk.edu.hk overall operation cost while satisfying the time-varying Department of Electrical and Electronic Engineering, load demand [4, 5]. Southern University of Science and Technology, Shenzhen, Unlike traditional power grids, microgrids have unique China drawbacks that preclude the optimization of economic Department of Electrical Engineering, The Hong Kong scheduling [6, 7]. Firstly, the abrupt changes of weather Polytechnic University, Hong Kong, China conditions affect non-dispatchable generation, i.e., solar Department of Electrical Engineering, School of Automation, photovoltaic (PV) and wind turbines. Secondly, temporal Guangdong University of Technology, Guangzhou, China uncertainty is always associated with volatile load profiles. School of Civil Engineering, Faculty of Engineering, Hence, it becomes challenging to make accurate University of Leeds, Leeds, UK The Chinese University of Hong Kong, Hong Kong, China 123 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1609 forecasting regarding the future state of supply and competitive microgrid scheduling algorithm, CHASE, is demand, which is essential for successful microgrid firstly proposed in [26, 27]. Using this algorithm, online scheduling [8]. decisions are made by tracking the underlying PD solutions Existing methods of uncertainty management in micro- in a retroactive manner. Distinguished from proactive grids for energy scheduling can be categorized into three approaches, the CHASE algorithm relies on no future classes. The first class of methods formulates the energy information while the scheduling performance has a strong management problem as a day-ahead scheduling problem, guarantee. In a pioneering work described in [26–28], the the solution to which is obtained by solving an offline foundation is laid and a new direction is suggested for deterministic problem using the tools of stochastic pro- managing microgrid operation uncertainties based on the gramming [9–14]. For the methods in this class, a sufficient theoretical performance guarantee. Based on CHASE, number of operation scenarios are typically required as rCHASE and iCHASE are proposed in [29] to further model inputs. Thus, the solution process can be computa- determine the advantages of randomization and interval tionally expensive. Furthermore, the effectiveness of such prediction for microgrid scheduling. Normally, the CHASE methods greatly relies on the representativeness of selec- algorithm is designed for applications to single or multiple tive operation scenarios. For a small-scale microgrid, homogeneous local generators. In this paper, we describe a stochastic scenarios may not be able to capture time- general retrospection-inspired scheduling algorithm, changing conditions, as the typical loading profile is highly hCHASE, which complements and extends the CHASE volatile and the aggregation effect vanishes. Robust opti- algorithm by extending the applicability to scheduling mization characterizes the second class of methods, which multiple heterogeneous generators. consider worst-case conditions in the presence of the highest forecasting error [15–18]. Although robust opti- mization can treat inherent operation uncertainties, the 2 System model and problem formulation obtained scheduling solutions may be too conservative and uneconomic. The last class of methods includes the We consider a small-scale microgrid as a single-bus methods for energy scheduling in shorter periods of time, system comprising of multiple local generators, renew- typically ranging from a few minutes to one hour. This is ables, and an aggregated load. This system connects to an also known as real-time scheduling. Many approaches have external grid, which can complement power imbalance of been considered, for example, using online optimization the microgrid in a due course. The system model is shown algorithms [19–21] or model predictive control (MPC) in Fig. 1. [22–24]. These methods can timely adapt to the fast- In this paper, the microgrid energy scheduling problem changing operation conditions, while they still require is formulated based on the following assumptions: the information on the future. In Section 3, we discuss 1) The microgrid system operates in a self-sustained MPC methodology for microgrid real-time applications. manner. The power mismatch between scheduled Microgrid energy scheduling can be in general repre- power and system net load is met by the external grid. sented as an across-time correlation problem, in which the 2) Renewable generators, including solar PV and wind future conditions are considered as necessary inputs. turbines, are free-running and non-dispatchable. Obviously, the above-mentioned approaches naturally lead 3) The microgrid system operates in a time-slotted to proactive decisions based on the forecasting information, manner. Dispatch orders are periodically sent to the applicability of which is passively restricted by the dispatchable units at the beginning of each time slot forecasting accuracy. In this sense, prediction-independent and take effect without time delay. scheduling approaches are of high interest to small-scale 4) We deploy a steady-state energy model assuming the microgrid applications. generation output of local generators, renewables, and The concept of ‘‘perfect dispatch’’ (PD) has been firstly load demand at each time slot to be constant. proposed in the initiative of PJM interconnection to 5) Renewable generation and electrical load demand are improve the real-time performance of power grids, which merely predictable at a certain accuracy rate d. Sec- refers to the least production cost commitment and dispatch tion 2.1 provides the mathematical model of renew- solution by assuming full knowledge of future conditions able forecasting. Day-ahead forecasting of the system and decisions [25]. PD performs retrospective direct opti- net load is difficult at a satisfactory accuracy rate, mization by spanning the entire dispatching horizon of while a general trend can be obtained as a model input. interest. Unfortunately, PD cannot be fully implemented in This assumption is reasonable based on the state-of- real applications because weather conditions cannot be the-art forecasting techniques of renewables and perfectly forecasted. However, PD solutions can provide residential load [30, 31]. some useful information for scheduling. Inspired by this, a 123 1610 Youwei JIA et al. considered as the only dispatchable units for maintaining the power balance in the microgrid. This study deploys a linear cost model for a specific local generator by inte- grating the start-up, sunk, and incremental operation costs, Up-stream power grid which is formulated as: incr sunk start Cðy ðtÞ; P ðtÞÞ ¼ c P ðtÞy ðtÞþ c y ðtÞþ c ½y ðtÞ i i i i i i i i i y ðt 1Þ Electricity ð4Þ send to/extract from where y ðtÞ and P ðtÞ are the operation on/off status and the i i output power of generation unit i at time slot t, respec- incr sunk start tively; c , c , c are the incremental, sunk, and start- i i i up costs of unit i, respectively; ‘‘?’’ means taking the positive number. Thermal local generators in a microgrid normally operate subject to the following operational constraints. 1) Power limit constraint Aggregated load Renewable generators Local generators min max P P ðtÞ P ð5Þ Operate in aĀ free Being dispatched i i i runningā mode periodically min max where P and P are the minimal power output and the i i Microgrid s ystem power capacity of generation unit i, respectively. 2) Power balance constraint Fig. 1 Microgrid system Given that there are N local generators in the microgrid, we obtain: 2.1 Renewable energy generation and system net load P ðtÞþ P ðtÞ P ðtÞð6Þ i grid load i¼1 In each time slot t, the electrical load and renewable where P ðtÞ is the power extracted from the main grid at grid energy generation (e.g., solar PV and wind energy) are time slot t. denoted as P ðtÞ and P ðtÞ, respectively. The system e renewable 3) Ramp-up and ramp-down constraints net load is defined as: The incremental output of a thermal generator in P ðtÞ¼ P ðtÞ P ðtÞð1Þ load e renewable two consecutive time slots is limited by the ramp-up up down and ramp-down rates P and P , i.e., In practice, a future net load is forecasted with specific i i techniques for energy scheduling purposes. For the up convenience of analysis, we use the following model to P ðt þ 1Þ P ðtÞ P ð7Þ i i approximate the forecasting process: down P ðtÞ P ðt 1Þ P ð8Þ i i P ðt þ rÞ load gðt þ rÞ 2 gðt þ rÞð2Þ 4) Minimum on-time and off-time constraints P ðt þ rÞ load If a thermal generator is committed at time slot t,it where P ðtÞ is the forecasting net load at time slot t; r is load remains committed for at least minimum on-time, and the number of look-ahead steps and g is defined in (3). vice versa. gðt þ rÞ¼ rðd 1Þþ 1 ð3Þ on y ðsÞ 1 t þ 1 s t þ T 1 ð9Þ fy ðtÞ [ y ðt1Þg i i i where d is the forecasting accuracy and d 2½0; 1. off y ðsÞ 1 1 t þ 1 s t þ T 1 ð10Þ i fy ðtÞ\y ðt1Þg i i i 2.2 Conventional generation where 1 denotes the indicator function; s is the time fg off on on In this paper, conventional thermal generators (e.g., gas instance between t?1 and t þ T 1. T and T are the i i i turbines, steam turbines, and diesel generators) are 123 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1611 minimum on-time and off-time of local generator, 3 MPC respectively. The classical MPC can be used for the real-time scheduling In general, thermal generators scheduled in real time in a problem of the power system. As illustrated in Fig. 2,the small-scale microgrid respond timely to satisfy the time- entire dispatching horizon is divided into several temporal varying electrical demand. That is, they normally exhibit on off segments coupled in the chronological order. The basic negligible T (T ) and relatively large ramp-up/ramp- principle of MPC is to solve the RHOP for a current segment down rates. offline and retain the dispatching order for the starting time slot. Such optimization process proceeds dynamically based 2.3 Problem formulation of microgrid energy on the scale of the look-ahead time window. scheduling Obviously, the MPC method directly targets sub-optimal scheduling solutions and makes proactive decisions. Based on the assumptions and technical models previ- However, it may be practically flawed in a small-scale ously mentioned, we formulate the microgrid cost-mini- microgrid for the following reasons: mization scheduling as a receding horizon optimization problem (RHOP), which is expressed as t N P P incr sunk start min c P ðtÞþ c y ðtÞþ c ½y ðtÞ y ðt 1Þ þ kðtÞP ðtÞ i i i i grid i i i ð11Þ t¼t 1 i¼1 s:t: ðÞ 5 ðÞ 10 where kðtÞ is the spot price of power obtained from the 1) A limited look-ahead window prevents the scheduling main grid at time slot t. The optimization window is solution. Compared with the PD, the competitiveness of bounded within [t , t*]. We assume that the price of elec- such solutions relates to the forecasting horizon of tricity sent from the microgrid to the main grid is negli- interest. gible. This assumption allows us to underscore the 2) The MPC method strongly relies on the forecasting of effective coordination of dispatchable sources in the future information. Large forecasting errors cause the microgrid, to minimize the operation cost, which is the solutions to deviate from the optima and generate misleading or ineffective dispatch order. In a small- main focus of our study. It should be noted that the RHOP is correlated across the scale microgrid, a high forecasting accuracy of the system net load is impossible. slots, owing to the existence of a start-up cost. PD is an offline optimal solution to the RHOP, spanning the entire 3) Due to the uncertainty associated with the prediction process, the MPC method cannot provide performance dispatching period with the full knowledge of future information, including the system net load and electricity guarantees. spot prices. 4) With the increase of the look-ahead time window, the optimization process will become computationally inefficient. Dispatching horizon 4 Retroactive algorithm for real-time scheduling Dispatching order 1 The formulated RHOP is generally known to be NP- Look-ahead s teps Dispatching order 2 complete. As discussed in Section 3, proactive decision- making based on MPC significantly depends on accurate Dispatching order k forecasting, which exposes limitations in tackling unpre- dictable net loads in small-scale microgrids. Instead, the retroactive algorithm explored in this paper relies on zero Fig. 2 Basic principle of MPC 123 1612 Youwei JIA et al. or little forecasting information, while it is still capable of decisions, while a quantifiable competitive ratio can be tracking the PD. guaranteed rigorously. Furthermore, in the CHASE algorithm, it is evident that scheduling decisions can be 4.1 Fundamental case with a single local generator made several steps ahead before T . It is provided that the future information is forecasted in a look-ahead time We first consider the fundamental case of the RHOP window, i.e., ½t; t þ r. However, it is worth noting that with only one local generator. At time slot t, the opportu- such decision-making may be misleading or ineffective, nity benefit of switching on a local generator is defined as: resulting from inherent forecasting errors. To avoid this, the following novel algorithm conserves the leverage dðtÞ¼ C½yðtÞ¼0; PðtÞ¼ 0; P C½yðtÞ grid ð12Þ forecasting information. Algorithm 1 calculates the ¼1; PðtÞ; P ¼ 0 grid competitive ratio of this algorithm. The opportunity benefit cumulates from the beginning time slot to the current time slot. The cumulative opportunity benefit is defined as: start DðtÞ¼ minf0; maxfc ; Dðt 1Þþ dðtÞgg ð13Þ The PD solution can be interpreted through the following types of dispatching segments over the entire horizon. A rigorous mathematical proof can be found in [26]. As illustrated in Fig. 3, four types of dispatching segments are defined as follows. The red point in Fig. 3 indicates a look-ahead window containing / time slots. 1) Type 0: ½1; T c c c start 2) Type 1: ½T þ 1; T ,if DðT Þ¼c and i iþ1 i DðT Þ¼ 0 iþ1 c c c 3) Type 2: ½T þ 1; T ,if DðT Þ¼ 0 and i iþ1 i c start DðT Þ¼c iþ1 4) Type 3: ½T þ 1; T end As concluded in [26], the PD solution to the RHOP is given by: c c 0if t 2½T þ 1; T ; Type 0, Type 2 or Type 3 PD i iþ1 y ðtÞ¼ c c 1if t 2½T þ 1; T ; Type 1 i iþ1 ð14Þ Recall that the CHASE algorithm proposed in [26]is devised to make online decisions at critical points T . Different from MPC, this is a ‘‘look back’’ algorithm by nature. At each time slot, the cumulative opportunity benefit, i.e., cumulative cost difference, is calculated and compared by looking into the historical states. Obviously, there always exist delays (i.e., ½T ; T ) for those online i i Theoretically, the competitive ratio of CHASE in con- sideration of the forecasting error satisfies: CR 1 hCHASE up dw start min sunk 4c þ e P ðk k Þþ 4ðr þ 1Þc load max min þ ð15Þ start 2c ð1 þ cÞ 2 2 e ¼ r r r d rd 2 ð16Þ Fig. 3 Schematic ofDðtÞand classification of dispatching segments 123 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1613 max incr sunk Given that fP ðtÞjt 2½1; T gis the day-ahead fore- day P c þ c load load c ¼ ð17Þ max incr sunk casting net loading profile, which can well reflect the P ðk c Þ c max load general trend (i.e., net load variations), the partitioning up incr k ¼minfkðtÞjkðtÞ cgð18Þ min objective and constraints are: dw incr day k ¼maxfkðtÞjkðtÞ\cgð19Þ N P P max > b min costðy ðtÞ; P ðtÞ; P ðtÞÞ b b n n grid min max t¼1 n¼1 where P and P are the minimum and maximum ð22Þ load load s:t:ðÞ 5 ; ðÞ 7 ðÞ 10 ;ðÞ 20 system net loads, respectively. b ly;n : n P ðtÞþ P ðtÞ P ðtÞ n grid load 4.2 General case with multiple heterogeneous The variable is the permutation of fb ; b ; ...; b g. 1 2 N generators This optimal partitioning problem can be efficiently solved by exhaustive search when N B 6, or by intelligent The essential idea of handling multiple heterogeneous optimization methods (e.g., particle swarm optimization) generators is to partition the electrical load into N layers when N [ 6. and allocate each layer to a dedicated generator, where N is Based on the illustration above, Algorithm 2 lists the the total number of local thermal generators. In this sense, hCHASE algorithm. the formulated RHOP is decomposed into N sub-problems and each one is solved individually with Algorithm 1. Intuitively, generators with a large start-up cost cover the load at the bottom layers as these normally exhibit least frequent variations. It should be noted that the ineffective partition of the load can increase the operation cost. The combined solution to all sub-problems is optimal on the condition of the optimal partitioning of load demand. In particular, given N local dispatchable generators in a microgrid, the generation capacity is max max max fP ; P ; ...; P g. The load demand is partitioned into 1 2 N N layers by following an order fb ; b ; ...; b g (from the 1 2 N bottom to the top) based on the power capacity of indi- vidual generators. Specifically, each layer is defined as n1 ly;n max ly;j P ðtÞ¼ minfP ; P ðtÞ P ðtÞg n 2½1; N load load b load j¼1 5 Numerical simulations ð20Þ The upper limit on the competitive ratio of the proposed If fa ; a ; ...; a g corresponds to an optimal partitioning 1 2 N hCHASE algorithm is provided in (15), which corresponds order, it should satisfy: to the theoretical worst case. In this section, we aim to C½ðy ; P Þ; P C½ðy ; P Þ; Pð21Þ a a grid b b grid i i i i a th Suppose ðy ; P ; P Þ is an optimal solution for the n a a n n grid sub-problem, and fa ; a ; ...; a g indicates the optimal 1 2 N y y y partitioning order, then ½ðy ; P Þ ; P is an optimal a a n n n¼1 grid solution to the entire RHOP, i.e., y ðtÞ¼ y ðtÞ, a a y y P ðtÞ¼ P ðtÞ, P ðtÞ¼ P ðtÞ. a a n n grid grid i¼1 Obviously, load partitioning is a necessary step for tackling multiple generators. Optimal partitioning becomes a premise for obtaining the optimal combined solution of the RHOP. In particular, load partitioning can be formu- lated as a permutation optimization problem, which is expressed in the following section. Fig. 4 Solar PV generation for 3 years 123 1614 Youwei JIA et al. Fig. 6 Electricity spot price in summer and winter Fig. 5 Electrical loading profiles for 3 years investigate the practical competitive ratio based on exten- sive numerical simulations. Realistic data of the Belgium grid [32], including solar PV generation and electrical load demand during 2014–2016 is used and scaled down to a 1 MW-level microgrid. Figures 4 and 5 present the solar PV generation and loading profiles and a day is divided into 96 points by every 15 min. 5.1 Parameter settings In the case studies, the system has a 15-minute-long scheduling resolution. Table 1 reports the parameters for both single and multiple generator cases. In the case studies, we assume that the electricity spot price can be accurately forecasted, which is shown in Fig. 6. 5.2 Single-generator case with net load prediction errors In this case, we compare hCHASE and MPC with dif- ferent forecasting accuracies and different look-ahead time windows. The operation profiles of the PD of hCHASE and MPC for five days are shown in Fig. 7. Visually, the scheduling profiles for hCHASE are much closer to the PD Table 1 Parameter settings max start incr sunk Case Generator P c c ($/ (MW) ($) kW) Single- G1 0.85 0.045 0.020 0.0015 generator Multiple- G1 0.40 0.059 0.015 0.0011 generator G2 0.10 0.038 0.230 0.0015 G3 0.20 0.042 0.017 0.0017 Fig. 7 Scheduling profiles of PD of hCHASE and MPC for 5 days G4 0.15 0.048 0.019 0.0020 123 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1615 Fig. 8 Simulation results for hCHASE and MPC in single-generator case where forecasting accuracy is 0.9 Fig. 9 Simulation results for hCHASE and MPC in single-generator case where forecasting accuracy is 0.65 than the ones obtained by MPC. Figures 8 and 9 show the detailed comparison results including cumulative distribu- sufficiently long (i.e., more than 2.5 hours). However, a tion function (CDF) and box plots. satisfactory prediction accuracy for such a long period is As shown in Figs. 8 and 9, the competitive ratio improves impossible in a small-scale microgrid. monotonously for both hCHASE and MPC along with the increase of look-ahead time window. This is reasonable, 5.3 Multiple-generator case with net load prediction owing to the high forecasting accuracy. However, such errors monotone improvement of the competitive ratio vanishes in the condition of low forecasting accuracy, as shown in Similarly, we compare hCHASE and MPC for the Fig. 8. This is because inaccurate forecasting information multiple-generator case, for different forecasting accura- can mislead scheduling decisions and will result in a large cies and different look-ahead time windows. Figures 10 operation cost. Overall, hCHASE outperforms MPC on the and 11 show the simulation results. The single-generator condition of no or little forecasting information. The box case reaches the same conclusions. Obviously, the com- plots also show that the competitive ratio of MPC is better petitive ratio CDF of hCHASE completely dominates the than the one of hCHASE, when the look-ahead window is 123 1616 Youwei JIA et al. Fig. 11 Simulation results for hCHASE and MPC in multiple- generator case where forecasting accuracy is 0.65 Fig. 10 Simulation results for hCHASE and MPC in multiple- generator case where forecasting accuracy is 0.9 one of MPC for both the conditions of high and low presence of certain degrees of forecasting errors with forecasting accuracies. The simulation results show guarantee of theoretical performance. Extensive case that, compared with the PD, hCHASE gives rise to less studies for the proposed algorithm and classical MPC are than 11% cost difference. carried out, using a 3-year-long realistic system data. Simulation results show that the cost difference between hCHASE and PD is less than 11% on average, which 6 Conclusion suggests better performance than the conventional MPC approach with the cost difference at 30% compared to PD. This paper proposes a retroactive algorithm hCHASE Acknowledgements This work was partially supported by Hong for microgrid real-time scheduling. The advantages of the Kong RGC Theme-based Research Scheme (No. T23-407/13N and proposed algorithm are its forecasting independence and No. T23-701/14N) and SUSTech Faculty Startup Funding (No. high computational efficiency. The proposed approach can Y01236135 and No. Y01236235). effectively tackle multiple heterogeneous generators in the 123 A retroactive approach to microgrid real-time scheduling in quest of perfect dispatch solution 1617 Open Access This article is distributed under the terms of the [18] Luo Z, Gu W, Wu Z et al (2018) A robust optimization method Creative Commons Attribution 4.0 International License (http:// for energy management of CCHP microgrid. J Mod Power Syst creativecommons.org/licenses/by/4.0/), which permits unrestricted Clean Energy 6(1):132–144 use, distribution, and reproduction in any medium, provided you give [19] Leung JY (2004) Handbook of scheduling: algorithms, models, appropriate credit to the original author(s) and the source, provide a and performance analysis. CRC Press, Boca Raton link to the Creative Commons license, and indicate if changes were [20] Mohan V, Singh JG, Ongsakul W et al (2016) Performance made. enhancement of online energy scheduling in a radial utility distribution microgrid. Int J Electr Power Energy Syst 79:98–107 [21] Shi W, Li N, Chu C-C et al (2017) Real-time energy manage- ment in microgrids. IEEE Trans Smart Grid 8(1):228–238 References [22] Du Y, Pei W, Chen N et al (2017) Real-time microgrid eco- nomic dispatch based on model predictive control strategy. [1] Katiraei F, Iravani R, Hatziargyriou N et al (2008) Microgrids J Mod Power Syst Clean Energy 5(5):787–796 management. IEEE Power Energy Mag 6(3):54–65 [23] Garcia-Torres F, Bordons C (2015) Optimal economical [2] Papadimitriou C, Kleftakis V, Hatziargyriou N (2017) Control schedule of hydrogen-based microgrids with hybrid storage strategy for seamless transition from islanded to interconnected using model predictive control. IEEE Trans Ind Electron operation mode of microgrids. J Mod Power Syst Clean Energy 62(8):5195–5207 5(2):169–176 [24] Gu W, Wang Z, Wu Z et al (2017) An online optimal dispatch [3] Haider S, Li G, Wang K (2018) A dual control strategy for schedule for CCHP microgrids based on model predictive power sharing improvement in islanded mode of AC microgrid. control. IEEE Trans Smart Grid 8(5):2332–2342 Protect Control Mod Power Syst 3:1–8 [25] Gisin B, Gu Q, Mitsche JV et al (2010) ‘‘Perfect dispatch’’—as [4] Li P, Han P, He S et al (2017) Double-uncertainty optimal the measure of PJM real time grid operational performance. In: operation of hybrid AC/DC microgrids with high proportion of Proceedings of IEEE PES general meeting, Providence, USA, intermittent energy sources. J Mod Power Syst Clean Energy 25–29 July 2010, 8 pp 5(6):838–849 [26] Lu L, Tu J, Chau C-K et al (2013) Online energy generation [5] Liu W, Gu W, Yuan X et al (2018) Fully distributed control to scheduling for microgrids with intermittent energy sources and coordinate charging efficiencies for energy storage systems. co-generation. ACM. https://doi.org/10.1145/2494232.2465551 J Mod Power Syst Clean Energy 6(5):1015–1024 [27] Lu L, Tu J, Chau C-K et al (2013) Towards real-time energy [6] Yang J, Guo B, Qu B et al (2018) Economic optimization on generation scheduling in microgrids with performance guaran- two time scales for a hybrid energy system based on virtual tee. In: Proceedings of IEEE PES general meeting, Vancouver, storage. J Mod Power Syst Clean Energy 6(5):1004–1014 Canada, 21–25 July 2013, 5 pp [7] Kroposki B (2017) Integrating high levels of variable renewable [28] Jia Y, He Y, Lyu X et al (2018) Hardware-in-the-loop imple- energy into electric power systems. J Mod Power Syst Clean mentation of residential intelligent microgrid. In: Proceedings of Energy 5(6):831–837 IEEE PES general meeting, Portland, USA, 5–10 August 2018, [8] Li Q, Xu Z, Yang L (2014) Recent advancements on the 5pp development of microgrids. J Mod Power Syst Clean Energy [29] Hajiesmaili MH, Chau C-K, Chen M et al (2016) Online 2(3):206–211 microgrid energy generation scheduling revisited: the benefits of [9] Kong X, Bai L, Hu Q et al (2016) Day-ahead optimal scheduling randomization and interval prediction. In: Proceedings of the method for grid-connected microgrid based on energy storage international conference on future energy systems, Waterloo, control strategy. J Mod Power Syst Clean Energy 4(4):648–658 Canada, 21–24 June 2016, 4 pp [10] Su W, Wang J, Roh J (2014) Stochastic energy scheduling in [30] Kong W, Dong Z, Jia Y et al (2019) Short-term residential load microgrids with intermittent renewable energy resources. IEEE forecasting based on LSTM recurrent neural network. IEEE Trans Smart Grid 5(4):1876–1883 Trans Smart Grid 10(1):841–851 [11] Farzan F, Jafari MA, Masiello R et al (2015) Toward optimal [31] Wan C, Zhao J, Song Y et al (2015) Photovoltaic and solar day-ahead scheduling and operation control of microgrids under power forecasting for smart grid energy management. CSEE J uncertainty. IEEE Trans Smart Grid 6(2):499–507 Power Energy Syst 1(4):38–46 [12] Khodaei A (2014) Resiliency-oriented microgrid optimal [32] Elia Group (2016) Solar power generation. http://www.elia.be/ scheduling. IEEE Trans Smart Grid 5(4):1584–1591 en/grid-data/power-generation/Solar-power-generation-data/ [13] Gholami A, Shekari T, Aminifar F et al (2016) Microgrid Graph scheduling with uncertainty: the quest for resilience. IEEE Trans Smart Grid 7(6):2849–2858 [14] Cai H, Chen Q, Guan Z et al (2018) Day-ahead optimal Youwei JIA received the B.Eng. degree from Sichuan University, charging/discharging scheduling for electric vehicles in micro- Chengdu, China, in 2011, and the Ph.D. degree from The Hong Kong grids. Protect Control Mod Power Syst 3:1–15 Polytechnic University, Hong Kong, China, in 2015, respectively. [15] Zhang Y, Gatsis N, Giannakis GB (2013) Robust energy man- From 2015 to 2018, he was a postdoctoral fellow at The Hong Kong agement for microgrids with high-penetration renewables. IEEE Polytechnic University. He is currently an Assistant Professor with Trans Sustain Energy 4(4):944–953 the Department of Electrical and Electronic Engineering, Southern [16] Xiang Y, Liu J, Liu Y (2016) Robust energy management of University of Science and Technology, Shenzhen, China. His research microgrid with uncertain renewable generation and load. IEEE interests include microgrid, renewable energy modeling and control, Trans Smart Grid 7(2):1034–1043 power system security analysis, complex network and artificial [17] Liu G, Starke M, Xiao B et al (2017) Robust optimisation-based intelligence in power engineering. microgrid scheduling with islanding constraints. IET Gener Transm Distrib 11(7):1820–1828 Xue LYU received the B.Eng. degree from Qingdao University of Technology, Qingdao, China, in 2013, the M.Eng. degree from Shanghai University of Electric Power, Shanghai, China, in 2016, and 123 1618 Youwei JIA et al. the Ph.D.degree from The Hong Kong Polytechnic University, Hong interests include demand side, grid integration of wind and solar Kong, China, in 2019, respectively. She is also a visiting Ph.D. power, electricity market planning and management, and artificial student at the Department of Electrical and Electronic Engineering, intelligence (AI) applications. Southern University of Science and Technology, Shenzhen, China. She is currently with The University of Hong Kong as a Postdoctoral Minghua CHEN received the B.Eng. and M.S. degrees from the Fellow. Her research interests include advanced modeling and control Department of Electronic Engineering, Tsinghua University, Beijing, for grid-integration of renewable energy systems. China, in 1999 and 2001, respectively, and the Ph.D. degree from the Department of Electrical Engineering and Computer Sciences, Chun Sing LAI received the B.Eng. degree (1st Hons.) in Electrical University of California at Berkeley, Berkeley, USA, in 2006. He and Electronic Engineering from Brunel University London, London, spent one year visiting Microsoft Research Redmond as a PostDoc- UK and D.Phil. in Engineering Science from University of Oxford, toral Researcher. He joined the Department of Information Engineer- Oxford, UK, in 2013 and 2019, respectively. He is currently an ing, The Chinese University of Hong Kong, Hong Kong, China, in Engineering And Physical Sciences Research Council (EPSRC) 2007, where he is currently an Associate Professor. He is also an Research Fellow with the School of Civil Engineering, University Adjunct Associate Professor with the Institute of Interdisciplinary of Leeds, Leeds, UK, and also a Visiting Research Fellow with the Information Sciences, Tsinghua University. He received the Eli Jury Department of Electrical Engineering, School of Automation, Award from UC Berkeley in 2007 (presented to a graduate student or Guangdong University of Technology, Guangzhou, China. His recent alumnus for outstanding achievement in the area of systems, current interests include power system optimization, artificial neural communications, control, or signal processing) and The Chinese networks, energy system modelling, data analytics, and energy University of Hong Kong Young Researcher Award in 2013. He also economics for renewable energy and storage systems. received several best paper awards, including the IEEE ICME Best Paper Award in 2009, the IEEE Transactions on Multimedia Prize Paper Award in 2009, and the ACM Multimedia Best Paper Award in Zhao XU received the B.Eng., M.Eng. and Ph.D. degrees from 2012. He is currently an Associate Editor of the IEEE/ACM Zhejiang University, Hangzhou, China, National University of Transactions on Networking. He serves as the TPC Co-Chair of Singapore, Singapore, and The University of Queensland, Brisbane, ACM e-Energy in 2016 and the General Chair of ACM e-Energy in Australia, in 1996, 2002 and 2006, respectively. From 2006 to 2009, 2017. He receives the ACM Recognition of Service Award in 2017 he was an Assistant and later Associate Professor with the Centre for for service contribution to the research community. His current Electric Technology, Technical University of Denmark, Lyngby, research interests include smart societal systems (e.g., smart power Denmark. Since 2010, he has been with The Hong Kong Polytechnic grids, energy-efficient data centers, and intelligent transportation University, Hong Kong, China, where he is currently a Professor in system), networked system, online competitive optimization, dis- the Department of Electrical Engineering and Leader of Smart Grid tributed optimization, and delay-constrained network coding. Research Area. He is also a foreign Associate Staff of Centre for Electric Technology, Technical University of Denmark. His research
Journal of Modern Power Systems and Clean Energy – Springer Journals
Published: Nov 16, 2019
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.