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

Learn More →

A new power swing detection method in power systems with large‐scale wind farms based on modified empirical‐mode decomposition method

A new power swing detection method in power systems with large‐scale wind farms based on modified... INTRODUCTIONNowadays, most mankind's electricity demand is supplied via transmission and distribution power systems. These systems transmit the generated energy in the power plants to the consumer through transmission lines [1]. Since power transmission lines have a significant role in this process, protecting them is extremely crucial to maintain network stability and ensure stable energy transmission [2]. Generally, distance relays are used as the main protection of the transmission lines due to their performance simplicity and simple coordination with other relays. Despite all these advantages, these relays may malfunction and disconnect healthy lines during power swing. Power swing can be caused by occurrence of fault, large disturbances in demand load, switching, adding or disconnection large amount of loads etc [1, 3]. Healthy line disconnections during power swing can cause instability in the power network and can even lead to a global blackout [4–7]. Hence, presenting methods to determine the power swing and prevent healthy line disconnections is greatly important.In addition, there is a growing concern over the sudden increase of energy demand and environmental issues. One of the current solutions is to use wind‐based renewable energies. Since the presence of wind turbines can significantly alter the power swing features and cause problems for the current power swing detection methods [8, 9], it is extremely important to present new methods, in which the impact of wind turbine presence is alleviated.Being previously mentioned, power swing detection plays an important role in the power systems. According to this importance, several studies have been conducted on the subject in recent years that will be reviewed in this section.In [10], a method is presented based on the power change rate to determine the power swing. In [11], the resistance seen by the distance relay is used to distinguish between the power swing and fault. This reference states that the during power swing, resistance seen by the distance relay constantly changes, while it is constant during faults. This method takes plenty of time to detect the power swing. In [12] a method is suggested by combining the concentric characteristics and the continuous monitor of apparent impedance. This method so complex that makes its practical use difficult. In [13], the Swing Centre Voltage (SCV) is used to distinguish between the power swing and fault. This method is unable to detect the unstable power swings. Besides, this method faces difficulties during 1800 3‐phase fault with phase angle of 180 degrees. The Fast Fourier Transform (FFT) has several applications in protecting power systems. Many references have used this transform to determine the power swing. Ref. [14] uses FFT to obtain the DC component of the current. Also, ref. [15] presents an FFT analysis‐based method. Ref. [2] utilizes the combination of the FFT and the signal energy to determine the power swing. Although this method can detect the high‐impedance faults which are simultaneous with power swing, it failed to detect the unstable power swing. Generally, the most important challenge of the methods which employ FFT is to precisely select the threshold value. None of the FFT using references has presented a method to select the proper value of the threshold and they have only used empirical values on the test network. Additionally, there are some other shortcomings for such methods like spectral leakage and picket fence due to asynchronous sampling and recording a limited number of samples. Refs. [16‐19] have used wavelet transform (WT) to compensate for the FFT problems. The wavelet transform‐based methods can quickly determine the power swing. Furthermore, these methods can also detect faults that are simultaneous with the power swing. The most important issue of the WT‐based methods is their dependency on a high sampling rate. Ref. [20] has presented a method based on the inverse drop in DC current. In [21, 22], some methods are presented based on machine learning under the supervision of support vector machines (SVM) and ANFIS. These methods require several simulations to train various ranges of fault and power swings. The automatic regression method is presented in [23]. This method has no proper performance during multi‐mode power swing. Authors in [24] have presented a method based on moving window averaging (MWA) of the current. This method has no proper performance during multi‐mode power swing and when signal is noisy. Refs. [25, 26] have used WT‐based methods to distinguish between the power swing and fault. Although these methods have suitable performance, they are significantly limited due to their required high sampling rate. Moreover, these methods are not cost‐efficient due to their need for special hardware. Additionally, these methods malfunction during asymmetric power swing and sudden increase of load. Ref. [28] uses the instantaneous frequency change rates to determine the power swing. The shortcomings of this method include its failure to detect the asymmetric power swing and malfunction during the signal noise. Ref. [29] uses Lissajous figures to detect faults which occur simultaneously with the power swing. However, it is not possible to detect asymmetric power swing using methods that employ Lissajous figures. detection In addition, this method malfunction during multi‐mode power swing occurrence and the signal noise. In [11], the phase angle of current positive sequence component is used to distinguish between the power swing and fault.As explained in the previous sections, the presence of wind turbines affects the power swing features in the transmission level and may lead to the inappropriate performance of these methods. Therefore, in this paper, a method is presented based on modified EMD with PMU to detect the power swing in the power systems with large‐scale wind turbines. EMD method has already been used in [33, 34]. However, the main shortcoming of these articles is the improper performance in the presence of a large‐scale wind turbine. Moreover, ref. [34] has other shortcomings such as malfunction during multi‐mode power swing. According to the mentioned arguments, the main purpose of this paper can be described as following:Using modified EMD with PMU to reduce the effect of the large‐scale wind turbines for power swing detection.Detecting 3‐phase fault simultaneously with the power swing in the power systems with large‐scale wind turbines.Detecting different power swing types including stable, unstable, and multi‐mode.A proper performance during noisy signal.This article is structured as follow: The new method is introduced in the second section. The simulation results are presented in the third section. In the fourth section, the proposed method is compared with some new available methods, and finally the conclusion is drawn in the fifth section.THE PROPOSED METHODPMU is a status estimation device that can measure the current and voltage phasor. In this article, the measurement accuracy is increased using PMU. Typically, PMUs acquire the required variables with the help of global positioning satellite system and using simultaneous sampling. PMU information is stored in the central protection unit and employed in the presented method. Since implementing PMUs increases the total cost, it is not possible to install them in all substations, and an optimization algorithm is required in order to optimize the number and placement of PMUs. However, since the authors in this paper only used PMUs information, and the innovation and the main purpose of this article is not about locating and internal system of PMUs, these subjects are ignored in this article. Refs. [28, 35] can be used for information on the locating of PMUs. Moreover, ref. [28] have suggested a noble method for locating PMUs in order to be implemented with distance relays. Figure 1 shows a single‐line diagram of a two‐circuit power system with a central protection unit. S and E represent the local and remote buses.1FIGURETwo‐circuit single‐line power system diagramIn the presented method of this paper, only the currents of the two ends of the transmission line are used. So, after sampling the current signals and sending the sampled data to the central protection unit (CPU), the FFT and the main component of the signals are extracted from every 6 current phases using (1) [36].1XK=∑n=0N−1Signaln×e−i2πKn/NK=0,…,N−1$$\begin{equation} \def\eqcellsep{&}\begin{array}{l} {X_K} = \sum_{n = 0}^{N - 1} {Signa{l_n} \times {e^{ - i2\pi Kn/N}}} \\[4pt] K = 0,\ldots,N - 1 \end{array} \end{equation}$$The values of Ibase${I_{base}}$, which are obtained from the current of both ends of the transmission line, are calculated using (2) to (4).2Ibase(a)=IS(a)+IR(a)$$\begin{equation}{I_{base}}(a) = \left| {{I_S}(a) + {I_R}(a)} \right|\end{equation}$$3Ibase(b)=IS(b)+IR(b)$$\begin{equation}{I_{base}}(b) = \left| {{I_S}(b) + {I_R}(b)} \right|\end{equation}$$4Ibase(c)=IS(c)+IR(c)$$\begin{equation}{I_{base}}(c) = \left| {{I_S}(c) + {I_R}(c)} \right|\end{equation}$$Subsequently, original signal (Ibase${I_{base}}$) should be decomposed into a number of intrinsic mode functions (IMFs) using EMD. This method is presented by Huang et al. in [37]. The waveform frequency is a constant value that is easily defined. Still, the signals are not practically sinusoidal or constant. Therefore, showing multiple unstable signals as a combination of various components of the signal with a precise evaluation of an event is a challenge [38]. Mainly, the signal empirical decomposition is a purification process in which various modes of swings are extracted from the main signal. The extracted modes are unipolar signals that include one band with a frequency. Hence, a signal can be decomposed to a few intrinsic modes with the following conditions: (1) The extremes of a dataset should be equal to the number of zero points or at least have a difference equal to one number. (2) Its integral should be equal to zero in a defined time interval. The EMD method comprises eight different steps to accomplish this task:First step: The signal decomposition process starts with finding the maximum and minimum points of the input signal.Second step: Creating an upper‐envelope curve through the third‐degree curve‐fitting on the maximum points.Third step: Creating a lower‐envelope curve through the third‐degree curve‐fitting on the minimum points.Fourth step: Mean subtraction of the upper‐envelope curve and lower‐envelope curve of the input signal using (5).5h1(t)=x(t)−m1(t)$$\begin{equation}{h_1}(t) = x(t) - {m_1}(t)\end{equation}$$In (4), h1(t)${h_1}(t)$is the IMF value of the beginning of the signal, x(t)$x(t)$denotes the input signal, and m1(t)${m_1}(t)$ is the mean of the upper‐envelope and lower‐envelope of the curve.Fifth step: Assessing the IMF condition and the stopping criterion using (6).6Dk=∑t=0Th1k−1(t)−h1k(t)∑t=0Th1k−1(t)2$$\begin{equation}{D_k} = \frac{{\sum_{t = 0}^T {\left| {h_1^{k - 1}(t) - h_1^k(t)} \right|} }}{{\sum_{t = 0}^T {{{\left| {h_1^{k - 1}(t)} \right|}^2}} }}\end{equation}$$Sixth step: If the fifth condition is not satisfied, the signal obtained from the fourth step is replaced by the main signal, and the rest of the process is repeated from the beginning.Seventh step: If the fifth condition is met, the screening process is finished and C1=h1k${C_1} = h_1^k$ is considered as the first IMF, which in fact is the high‐frequency component of the signal x(t)$x(t)$.Eighth step: The remaining is defined as Equation (7) and if it satisfies the IMF condition, it will be considered as an IMF, otherwise, if it meets the first condition, it will be considered as an initial signal, and the steps 1–4 are iterated until the next IMF is obtained, and if the first condition is not met, it will be considered as a remaining (r).7r1=x(t)−C1k$$\begin{equation}{r_1} = x(t) - C_1^k\end{equation}$$The main signal is the sum of the IMFs plus the remaining (8).8x(t)=r+∑n=1Nhn$$\begin{equation}x(t) = r + \sum_{n = 1}^N {{h_n}} \end{equation}$$In the final step, the value of the obtained RMS IMFs is calculated using (9) [30].9RMS=∑k=1NIk2N$$\begin{equation}RMS = \sqrt {\frac{{\sum_{k = 1}^N {I_k^2} }}{N}} \end{equation}$$The signal decomposition using EMD results in multiple signals. Therefore, one reference signal must be employed. In this paper, IMF2 will be used as the reference signal. [31] thoroughly explains the method of adopting a reference signal. In addition, using the proposed method of this paper, the power swing and fault can be distinguished by defining a threshold value and using the condition in (10). The threshold value is obtained by using various computer simulations. Studies [31, 37] can be referred to get more information about threshold value detection. The threshold value is equal to 0.1 in this paper.10IfIMF2>Threshold→Faultisdeteced$$\begin{equation}{\rm{If IMF2 > Threshold }} \to {\rm{ Fault is deteced}}\end{equation}$$According to the stated points, Figure 2 shows the algorithm of the presented method.2FIGUREThe presented method algorithmSIMULATION RESULTSTest systemIn order to implement the proposed algorithm, an IEEE standard 39‐bus network is used. The single‐line diagram of this network is illustrated in Figure 3. This network includes 39 buses and 10 generators. Multiple wind farms are connected to buses to evaluate the performance of the suggested method while having a large‐scale wind turbine in the network. The data of these wind farms are presented in Table 1. In this article the DFIG type turbine is implemented which uses a three‐phase induction generator. The implemented generator model is illustrated in the Figure 4. In this generator, the stator winding is directly connected to the grid, while the rotor winding is controlled using an inverter. As its most important advantage, the DFIG can provide electricity with constant frequency via a variable mechanical speed. The DFIG model is considered as a wind farm in the studied network.3FIGURESingle‐line diagram of IEEE 39 bus system1TABLEThe studied wind farm dataRowBusNominal apparent power (MVA)Power factorParallel units1265000.8502264000.8403295000.8504294000.8404FIGUREThe implemented wind turbine modelThe wind farms will merely be connected to the network to assess the first case, however, since the suggested method should be comprehensive, and to ensure the desired performance in the absence of the wind farms, these wind farms will be disconnected from the network while assessing the stable, unstable, and multi‐mode power. Additionally, PMUs are installed at the buses 26 and 29 in order to protect the 26–29 line.The studied network is analysed using Digsilent software and the results are transferred to the MATLAB software to evaluate the suggested protection algorithm. It should be noted that the sampling rate is considered equal to 10 kHz.Case study 1: Fault and power swing occurrence during wind farm connectionsIn this case, the effect of wind farm presence is evaluated on the suggested method. All the wind farms are connected to the network based on all the information provided in the previous section. The power switches of the 26–28 and 28–29 lines are opened for one second to create the power swing. Also, a three‐phase fault is placed on the 50 percent of the line (26–29 line) at 3 s to simulate the fault. Figure 5a shows the current changes in the event of fault and power swing occurrence. As mentioned before, the presence of wind turbines in the network changes the power swing parameter. Figure 5b shows a stable power swing in the absence of a wind turbine in the network. By comparing it to Figure 5a (until 3 s that fault enters the network) it may be observed that we are seeing a completely different signal.5FIGURECurrent waveform during power swing (a) the waveform when there is a wind turbine in the network (b) the waveform when there is no wind turbine in the networkAccording to the previously mentioned results, it is evident that the current methods are not able to detect this type of power swing. To prove this point, multiple power swing detection methods that have been presented in recent years are reviewed. The reviewed methods include the following items:The power swing detection method using chaos theory presented in 2020 [39].The power swing detection method using instantaneous frequency presented in 2020 [38].The power swing detection method based on 3‐phase RMS change rates presented in 2018 [31].The power swing detection method based on moving windows average (MWA) presented in 2014 [25].The power swing detection method is based on signal energy changes presented in 2019 [2].The power swing detection method is based on Taylor series presented in 2016 [48].Figure 6 shows the performance of the 5 mentioned methods during the power swing of Figure 5a. Clearly, all these methods face problems during power swing (the interval of 1–3 s) and incorrectly detects a fault for power swing. Moreover, these methods do not consider any difference between the power swing and fault in the networks with a wind turbine.6FIGUREThe performance of the current methods during power swing in a network with a wind turbine (a) chaos theory, (b) instantaneous frequency, (c) RMS, (d) MWA, (e) Energy (f) Taylor seriesIn addition to the five previously reviewed references, as it was mentioned in the Section 1, [33], which uses EMD like the presented paper, also cannot determine this type of power swing. Figure 7 shows the performance of the suggested method in [33] versus this case. As it is seen in Figure 7, in the event of this type of power swing occurrence, the calculated IMF value exceeds the threshold value defined by the paper. Hence, this reference also cannot determine this type of power swing. It should be noted that if the threshold value is increased to address this issue, fault detection will no longer be possible in the wind turbine‐less networks. Furthermore, when the threshold value increases, the fault detection will no longer be possible in the networks connected to the wind turbines either.7FIGUREThe performance of the suggested method in [33] during power swing in the networks connected to a wind turbineAccording to the mentioned points, presenting an effective method is vital. Figure 8 depicts the suggested method performance of this paper during power swing. Based on this figure, the calculated IMF values do not exceed the threshold value during the power swing (between 1 to 3 s). Whereas, in the event of 3‐phase fault (at 3 s), the calculated IMF values exceed the threshold value and the fault is detected approximately in 4 ms.8FIGUREThe performance of the suggested method of this paper in the networks connected to a wind turbineCase study 2: Detection of stable, unstable, and multi‐mode power swing detection simultaneous with faultAlthough the main purpose of this article is to present a method to differentiate fault and power swing of the networks connected to large‐scale wind turbines, such methods should be able to operate in various conditions. Therefore, the suggested method performance in various situations of power swing and faults are reviewed in this section.The stable power swing can be considered the simplest type of power swing. This type of power swing is not a major challenge for the detection methods. First, all the wind farms are disconnected from the network to evaluate the effect of this type of power swing. Then, the circuit breakers of the 26–28 and 28–29 lines are opened in 0.8 s to create the power swing. A 3‐phase fault is located at 50 percent of the protected lines in 2.1 s to create a fault simultaneous with the power swing. Figure 9 shows the performance of the suggested method during stable power swing with simultaneous fault. As it is seen in this figure, the suggested method has managed to detect the fault almost within 2.32 ms.9FIGUREThe performance of the suggested method of this paper during the stable power swing and 3‐phase fault simultaneous with the power swingThe unstable power swing usually is developed due to extreme turbulence like sudden adding or removing a major part of a load, 3‐phase fault etc. Unlike stable power swing, this type of swing does not end after a few cycles. Although it is needed to disconnect the unstable synchronous generator from the network to end this type of power swing, the protected line should remain in the circuit so that it can supply the connected loads through other available resources. To simulate this type of power swing, first, a 1000 MW and a 200 MVAR loads are connected to the network in 0.8 s. Then, a 3‐phase fault is placed on the 50 percent of the protected lines in 2.5 s to simulate a simultaneous fault with the power swing. Figure 10 shows the performance of the suggested method when the unstable power swing occurs with simultaneous 3‐phase fault. As it is seen in this figure, the suggested method has managed to detect the fault almost within 1.95 ms.10FIGUREThe performance of the suggested method of this paper during unstable power swing and 3‐phase fault simultaneous with power swingA major challenge for power swing detection methods is the multi‐mode power swing detection. Unlike the stable and unstable power swing, the multi‐mode power swing has no constant swing frequency. Additionally, in some parts of this type of power swing, the current signal is distorted and is not completely sinusoidal. This type of power swing occurs when multiple synchronous generators swing relative to each other. The protected line of 17–18 is considered to evaluate the effect of this type of power swing. A 3‐phase fault is placed at 50 percent of the 4–14 line to create a multi‐mode power swing. The power switches of the 4–14 line are opened in 300 ms. This helps the protection system of the 17–18 line to detect the multi‐mode power swing. Figure 11 depicts the suggested method performance of this paper during multi‐mode power swing. According to this figure, even during 3‐phase fault simultaneous with the power swing, the suggested method still can detect the fault within 6.02 ms.11FIGUREThe performance of the suggested method of this paper during the multi‐mode power swing and 3‐phase fault simultaneous with the power swingThe main advantage of the suggested method is the multi‐mode power swing and fault detection time reduction. Table 2 shows the comparison results of the suggested method of this paper with the references presented in Sections 2 and 3. The references are compared with the suggested method in this paper in terms of fault detection time and multi‐mode power swing detection capability. As it is shown in this table, most reviewed methods cannot determine the multi‐mode power swing. Only the methods presented in [33, 39] can determine this type of power swing. Additionally, by comparing fault detection time, it can be concluded that the suggested method can enhance the fault detection time as well.2TABLEComparison of the suggested method in this paper with similar methods based on fault detection time and multi‐mode power swing detectionReferenceMethodFault detection time (ms)Multi‐mode power swing[39]Chaos7–8Yes[40]Instantaneous frequency31–33No[30]RMS100No[25]MWA16No[2]Energy10–11No[33]EMD and Hilbert7–9YesProposed methodEMD and PMU2–6YesCase study 3: Signal noiseNoise is an inseparable part of any actual system. It is almost impossible to build a noise‐free system. Therefore, it is better to evaluate the noise tolerance of the power swing detection methods . Most protection studies conducted in recent years have used white Gaussian noise to explore the effect of noise on the performance of their methods. For example, ref. [42] has used a power transformer to assess the effect of noise on the protection method, and ref. [43] has used white Gaussian noise to explore the impact of noise on the fault locating method. Several references have also used white Gaussian noise to assess the noise effect in the field of power swing detection such as [7, 31, 37]. Figure 12 shows the performance of the suggested method when the noise occurs in the signal with various signal to ratios (SNRs). As it is shown in Figure 12, the suggested method of this paper has a desirable performance in white Gaussian noise with SNR = 40 dB and SNR = 30 dB, however, when white Gaussian noise is added to the main signal with an SNR of 20 dB, the suggested method will face problems.12FIGUREThe performance of the suggested method of this paper during white Gaussian noise (a) SNR = 40 dB, (b) SNR = 30 dB, (c)SNR = 20 dBAmong the currently reviewed references, the presented methods in [31, 37] are resistant against white Gaussian noises up to SNR = 30 dB Whereas, the refs. [2, 24, 30, 38] are not noise‐resistant and are extremely affected by the noise.Case study 4: Asymmetric power swingThe power swings are not always symmetrical and during some events such as single pole switching or automatic reclosing of a single pole an asymmetrical power swing may occur [39]. As mentioned in the introduction section, asymmetric power swing can cause problems for existing methods. On this basis, the performance of the presented method during asymmetric power swing is investigated. In order to create asymmetric power swing, phase A switches of lines 26–28 and 28–29 are opened in 1.2 s. This causes asymmetric power swing (only phase A starts to oscillate) on the protected line (line 26–29). In order to investigate the subject thoroughly, a three‐phase fault is placed on 50% of line 26–29 in 3 s. Figure 13 presents the performance of the proposed method during asymmetric power swing which is simultaneous with a three‐phase fault. According to this figure, the proposed method has operated appropriately during asymmetric power swing.13FIGUREThe performance of the presented method during the asymmetric power swingCase study 5: The fault detection accuracyIn order to investigate this case, 500 different events are randomly generated in DigSilent software and applied to the algorithm in MATLAB software. From these 500 events, 100 events cause stable power swing, 100 events cause unstable power swing, and the other 300 events cause various types of faults (single‐phase, two‐phase, and three‐phase). The results of this test are presented in the Table 3. According to this table, in the analysis of stable and unstable power swing, 100% of the events have been detected by the algorithm. In case of single‐phase and two‐phase faults, the fault detection accuracy percentage is 99% and for three‐phase faults it is 98%. In general, it can be concluded that the presented algorithm has been operated correctly in 99.2% of the time.3TABLEThe fault detection accuracyEventsFault detection accuracy (%)Stable power swing100Unstable power swing100Single‐phase fault99Two‐phase fault99Three‐phase fault98All events99.2COMPARING THE PROPOSED METHOD WITH SIMILAR METHODSIn recent years, various methods have been suggested to determine power swing. Each method has its specific signals, signal process tools, sampling frequency etc. Although the suggested method in this paper is compared with the current methods in the simulation section, a general comparison is presented in this section based on different factors. Table 4 shows a comprehensive comparison between the proposed method with other current methods for power swing detection.4TABLEThe comprehensive comparison of the proposed method and the other methods for power swing detectionReferenceParameterMethodSampling rate (kHz)Symmetrical fault detectionFaster than one cycleEasy to implementWind turbine[44]Active and reactive powerRate of change of power2YesNoYesNo[10]Current and voltageAutomatic regression1YesNoYesNo[36]Active powerFFT10YesYesYesNo[21]CurrentProny3YesYesYesNo[45]CurrentWavelet40.96YesYesYesNo[18]CurrentTravelling wave10YesYesNoNo[42]CurrentPhaselet10YesYesYesNo[39]CurrentGMDH10YesYesYesNoProposed methodCurrentEMD‐PMU10YesYesYesYesCONCLUSIONToday, technological development has increased the electric energy consumption in the world. The increased electric energy consumption, concerns over natural resources, and carbon emission has caused the distributed generation to be considered as a useful solution. Among the different kinds of distributed generations, the use of large‐scale wind turbines is increasing in the level of the power system. The presence of such wind turbines in combination with the synchronous generators can cause changes in the power swing parameters of the power system. These changes may cause malfunction of the power swing detection methods and lead to undesirable tripping of the healthy power lines. Therefore, a modified EMD method was presented in this paper using PMU. The suggested method can easily differentiate power swing from fault in the systems connected to the large‐scale wind turbines. Moreover, the suggested method is also applicable in systems without wind turbine. The simulation results and the comparisons that are made with similar methods indicate the efficiency and applicability of the proposed method. In the following, using a statistical analysis including 500 different events, the efficiency and the accuracy of the presented method is tested and is calculated to be 99.6%.In the following, using a statistical analysis including 500 different events, the efficiency and the accuracy of the presented method is tested and is calculated to be 99.6%. Obviously, the proposed approach of this article and most of the reviewed references use a threshold value to distinguish between power swing and fault. Using a threshold value can cause difficulties and challenges for operators and researchers. In addition, determining the threshold value can be a sophisticated and time‐consuming task. For this reason, the authors suggest the research on methods of determining faults without the need of a threshold value or methods to determine the threshold value for future works.AUTHOR CONTRIBUTIONSA.A.N.: Investigation; Methodology; Software; Validation; Writing – original draft; Writing – review & editing. F.R.: Methodology; Project administration; Software; Writing – original draft; Writing – review & editing. A.F.: Formal analysis; Methodology; Validation; Writing – review & editing.CONFLICT OF INTERESTThe authors declare no conflict of interest.DATA AVAILABILITY STATEMENTData sharing not applicable to this article as no datasets were generated or analysed during the current study.REFERENCESJannati M., Mohammadi M.: A novel fast power swing blocking strategy for distance relay based on ADALINE and moving window averaging technique. IET Gener. Transm. Distrib. 15(1), 97–107 (2021)Taheri B., Salehimehr S., Razavi F.: A new method for fast power swing detection using the rate of change of energy in the current signal. In: 2019 27th Iranian Conference on Electrical Engineering (ICEE). Yazd, Iran (2019)Taheri B., Faghihlou M., Salehimehr S., Razavi F.: Symmetrical fault detection during power swing using mean value of sampled data from the current signal. IETE J. Res. 1–13 (2020)Bakar A.H.A., Yatim F.M., Yusof S., Othman M.R.: Analysis of overload conditions in distance relay under severe system contingencies. Int. J. Electr. Power Energy Syst. 32(5), 345–350 (2010)Gunasegaran M., Tan C., Bakar A.H.A., Mokhlis H., Illias H.A.: Progress on power swing blocking schemes and the impact of renewable energy on power swing characteristics: A review. Renewable Sustainable Energy Rev. 52, 280–288 (2015)Salehimehr S., Taheri B., Faghihlou M.: Detection of power swing and blocking the distance relay using the variance calculation of the current sampled data. Electr. Eng. 104, 913–927 (2021)Taheri B., Sedighizadeh M.: Detection of power swing and prevention of mal‐operation of distance relay using compressed sensing theory. IET Gener. Transm. Distrib. 14(23), 5558–5570 (2020)Juttu T.R., Bhalja B.R., Andreev M., Malik O.P.: Synchrophasor assisted power swing detection scheme for wind integrated transmission network. IEEE Trans. Power Delivery. 37(3), 1952–1962 (2021)Haddadi A., Kocar I., Karaagac U., Gras H., Farantatos E.: Impact of wind generation on power swing protection. IEEE Trans. Power Delivery 34(3), 1118–1128 (2019)Rao J.G., Pradhan A.K.: Differential power‐based symmetrical fault detection during power swing. IEEE Trans. Power Delivery 27(3), 1557–1564 (2012)Khoradshadi‐Zadeh H.: Evaluation and performance comparison of power swing detection algorithms. In: IEEE Power Engineering Society General Meeting. San Francisco, USA. pp. 1842–1848 (2005)Martuscello L., Krizauskas E., Holbach J., Lu Y.: Tests of distance relay performance on stable and unstable power swings reported using simulated data of the August 14 th 2003 system disturbance. In: 2009 Power Systems Conference. Clemson, USA. pp. 1–21 (2009)Benmouyal G., Hou D., Tziouvaras D.: Zero‐setting power‐swing blocking protection. In: 31st Annual Western Protective Relay Conference. pp. 19–21 (2004)Karegar H.K., Mohamedi B.: A new method for fault detection during power swing in distance protection. In: 2009 6th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology. Chonburi, Thailand. pp. 230–233 (2009)Mahamedi B.: A very fast unblocking scheme for distance protection to detect symmetrical faults during power swings. In: 2010 Conference Proceedings IPEC. Singapore. pp. 378–383 (2010)Brahma S.: Use of wavelets for out of step blocking function of distance relays. In: 2006 IEEE Power Engineering Society General Meeting. Montreal, Canada. p. 5 (2006)Pang C., Kezunovic M.: Fast distance relay scheme for detecting symmetrical fault during power swing. IEEE Trans. Power Delivery 25(4), 2205–2212 (2010)Mahamedi B.: A new power swing blocking function based on wavelet transform. In: 2011 2nd International Conference on Electric Power and Energy Conversion Systems (EPECS). Sharjah, UAE. pp. 1–6 (2011)Dubey R., Samantaray S., Tripathy A., Babu B.C., Ehtesham M.: Wavelet based energy function for symmetrical fault detection during power swing. In: 2012 Students Conference on Engineering and Systems. Allahabad, India. pp. 1–6 (2012)Lotfifard S., Faiz J., Kezunovic M.: Detection of symmetrical faults by distance relays during power swings. IEEE Trans. Power Delivery 25(1), 81–87 (2009)Seethalekshmi K., Singh S., Srivastava S.: A classification approach using support vector machines to prevent distance relay maloperation under power swing and voltage instability. IEEE Trans. Power Delivery 27(3), 1124–1133 (2012)Zadeh H.K., Li Z.: A novel power swing blocking scheme using adaptive neuro‐fuzzy inference system. Electr. Power Syst. Res. 78(7), 1138–1146 (2008)Rao A.K., Ahmad A.: Power swing blocking (PSB) function for distance relay using prediction technique. Int. J. Syst. Assur. Eng. Manage. 8(2), 301–307 (2017)Rao J.G., Pradhan A.K.: Power‐swing detection using moving window averaging of current signals. IEEE Trans. Power Delivery 30(1), 368–376 (2014)Alsyoufi Y.R., Hajjar A.A.: A high‐speed algorithm to discriminate between power swing and faults in distance relays based on a fast wavelet. Electr. Power Syst. Res. 172, 269–276 (2019)Patel B., Bera P.: Fast fault detection during power swing on a hybrid transmission line using WPT. IET Gener. Transm. Distrib. 13(10), 1811–1820 (2019)Rao J.T., Bhalja B.: Prevention of maloperation of distance relay under severe stressed conditions for series compensated transmission line considering optimal placement of phasor measurement units. IET Gener. Transm. Distrib. 14(11), 2148–2159 (2020)Taheri B., Salehimehr S., Razavi F., Parpaei M.: Detection of power swing and fault occurring simultaneously with power swing using instantaneous frequency. Energy Syst. 11(2) 491–514 (2020)Patel B.: A new technique for detection and classification of faults during power swing. Electr. Power Syst. Res. 175, 105920 (2019)Taheri B., Razavi F.: Power swing detection using rms current measurements. J. Electr. Eng. Technol. 13(5), 1831–1840 (2018)Taheri B., Hosseini S.A., Abyaneh H.A., Razavi F.: Power swing detection and blocking of the third zone of distance relays by the combined use of empirical‐mode decomposition and Hilbert transform. IET Gener. Transm. Distrib. 14(6), 1062–1076 (2020)Biswal S., Biswal M.: Fault‐swing discrimination using Hilbert–Huang transform integrated discrete teager energy operator. IET Sci. Meas. Technol. 12(7), 829–837 (2018)Parpaei M., Ghaffarzadeh N., Tayar M.Z.: Rational random walk‐based optimal placement of phasor measurement units to enhance the initializing and guiding the optimization processes. Int. Trans. Electr. Energy Syst. 27(7), e2323 (2017)Taheri B., Faghihlou M., Salehimehr S., Razavi F.: A fast Fourier transform‐based method for power swing detection and distance relay malfunction prevention. J. Control Autom. Electr. Syst. 31(6), 1458–1468 (2020)Huang N.E., et al.: The empirical mode decomposition and the Hilbert spectrum for non‐linear and non‐stationary time series analysis. Proc. R. Soc. London, Ser. A 454(1971), 903–995 (1998)Shukla S., Mishra S., Singh B.: Empirical‐mode decomposition with Hilbert transform for power‐quality assessment. IEEE Trans. Power Delivery 24(4), 2159–2165 (2009)Salehimehr S., Taheri B., Razavi F., Parpaei M., Faghihlou M.: A new power swing detection method based on chaos theory. Electr. Eng. 102(2), 663–681 (2020)Taheri B., Salehimehr S., Razavi F., Parpaei M.: Detection of power swing and fault occurring simultaneously with power swing using instantaneous frequency. Energy Syst. 11(2), 491–514 (2020)Hosseini S.A., Taheri B., Abyaneh H.A., Razavi F.: Comprehensive power swing detection by current signal modeling and prediction using the GMDH method. Prot. Control Mod. Power Syst. 6(1), 1–11 (2021)Ghalesefidi M.M., Ghaffarzadeh N.: A new phaselet‐based method for detecting the power swing in order to prevent the malfunction of distance relays in transmission lines. Energy Syst. 12(2), 491–515 (2021)Hashemi S.M., Sanaye‐Pasand M., Shahidehpour M.: Fault detection during power swings using the properties of fundamental frequency phasors. IEEE Trans. Smart Grid 10(2), 1385–1394 (2017)Taheri B., Sedighizadeh M.: A moving window average method for internal fault detection of power transformers. Cleaner Eng. Technol. 4, 100195 (2021)Taheri B., Salehimehr S., Sedighizadeh M.: A fault location algorithm for parallel line based on the long short‐term memory model using the distributed parameter line model. Int. Trans. Electr. Energy Syst. e13032 (2021)Lin X., Gao Y., Liu P.: A novel scheme to identify symmetrical faults occurring during power swings. IEEE Trans. Power Delivery 23(1), 73–78 (2007)Brahma S.M.: Distance relay with out‐of‐step blocking function using wavelet transform. IEEE Trans. Power Delivery 22(3), 1360–1366 (2007)Tekdemir, I.G., Alboyaci B.: A novel approach for improvement of power swing blocking and deblocking functions in distance relays. IEEE Trans. Power Delivery 32(4), 1986–1994 (2016) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png IET Generation Transmission & Distribution Wiley

A new power swing detection method in power systems with large‐scale wind farms based on modified empirical‐mode decomposition method

Download PDF
12 pages

Loading next page...
 
/lp/wiley/a-new-power-swing-detection-method-in-power-systems-with-large-scale-gNo27rbl9y

References (45)

Publisher
Wiley
Copyright
© 2023 The Institution of Engineering and Technology.
eISSN
1751-8695
DOI
10.1049/gtd2.12727
Publisher site
See Article on Publisher Site

Abstract

INTRODUCTIONNowadays, most mankind's electricity demand is supplied via transmission and distribution power systems. These systems transmit the generated energy in the power plants to the consumer through transmission lines [1]. Since power transmission lines have a significant role in this process, protecting them is extremely crucial to maintain network stability and ensure stable energy transmission [2]. Generally, distance relays are used as the main protection of the transmission lines due to their performance simplicity and simple coordination with other relays. Despite all these advantages, these relays may malfunction and disconnect healthy lines during power swing. Power swing can be caused by occurrence of fault, large disturbances in demand load, switching, adding or disconnection large amount of loads etc [1, 3]. Healthy line disconnections during power swing can cause instability in the power network and can even lead to a global blackout [4–7]. Hence, presenting methods to determine the power swing and prevent healthy line disconnections is greatly important.In addition, there is a growing concern over the sudden increase of energy demand and environmental issues. One of the current solutions is to use wind‐based renewable energies. Since the presence of wind turbines can significantly alter the power swing features and cause problems for the current power swing detection methods [8, 9], it is extremely important to present new methods, in which the impact of wind turbine presence is alleviated.Being previously mentioned, power swing detection plays an important role in the power systems. According to this importance, several studies have been conducted on the subject in recent years that will be reviewed in this section.In [10], a method is presented based on the power change rate to determine the power swing. In [11], the resistance seen by the distance relay is used to distinguish between the power swing and fault. This reference states that the during power swing, resistance seen by the distance relay constantly changes, while it is constant during faults. This method takes plenty of time to detect the power swing. In [12] a method is suggested by combining the concentric characteristics and the continuous monitor of apparent impedance. This method so complex that makes its practical use difficult. In [13], the Swing Centre Voltage (SCV) is used to distinguish between the power swing and fault. This method is unable to detect the unstable power swings. Besides, this method faces difficulties during 1800 3‐phase fault with phase angle of 180 degrees. The Fast Fourier Transform (FFT) has several applications in protecting power systems. Many references have used this transform to determine the power swing. Ref. [14] uses FFT to obtain the DC component of the current. Also, ref. [15] presents an FFT analysis‐based method. Ref. [2] utilizes the combination of the FFT and the signal energy to determine the power swing. Although this method can detect the high‐impedance faults which are simultaneous with power swing, it failed to detect the unstable power swing. Generally, the most important challenge of the methods which employ FFT is to precisely select the threshold value. None of the FFT using references has presented a method to select the proper value of the threshold and they have only used empirical values on the test network. Additionally, there are some other shortcomings for such methods like spectral leakage and picket fence due to asynchronous sampling and recording a limited number of samples. Refs. [16‐19] have used wavelet transform (WT) to compensate for the FFT problems. The wavelet transform‐based methods can quickly determine the power swing. Furthermore, these methods can also detect faults that are simultaneous with the power swing. The most important issue of the WT‐based methods is their dependency on a high sampling rate. Ref. [20] has presented a method based on the inverse drop in DC current. In [21, 22], some methods are presented based on machine learning under the supervision of support vector machines (SVM) and ANFIS. These methods require several simulations to train various ranges of fault and power swings. The automatic regression method is presented in [23]. This method has no proper performance during multi‐mode power swing. Authors in [24] have presented a method based on moving window averaging (MWA) of the current. This method has no proper performance during multi‐mode power swing and when signal is noisy. Refs. [25, 26] have used WT‐based methods to distinguish between the power swing and fault. Although these methods have suitable performance, they are significantly limited due to their required high sampling rate. Moreover, these methods are not cost‐efficient due to their need for special hardware. Additionally, these methods malfunction during asymmetric power swing and sudden increase of load. Ref. [28] uses the instantaneous frequency change rates to determine the power swing. The shortcomings of this method include its failure to detect the asymmetric power swing and malfunction during the signal noise. Ref. [29] uses Lissajous figures to detect faults which occur simultaneously with the power swing. However, it is not possible to detect asymmetric power swing using methods that employ Lissajous figures. detection In addition, this method malfunction during multi‐mode power swing occurrence and the signal noise. In [11], the phase angle of current positive sequence component is used to distinguish between the power swing and fault.As explained in the previous sections, the presence of wind turbines affects the power swing features in the transmission level and may lead to the inappropriate performance of these methods. Therefore, in this paper, a method is presented based on modified EMD with PMU to detect the power swing in the power systems with large‐scale wind turbines. EMD method has already been used in [33, 34]. However, the main shortcoming of these articles is the improper performance in the presence of a large‐scale wind turbine. Moreover, ref. [34] has other shortcomings such as malfunction during multi‐mode power swing. According to the mentioned arguments, the main purpose of this paper can be described as following:Using modified EMD with PMU to reduce the effect of the large‐scale wind turbines for power swing detection.Detecting 3‐phase fault simultaneously with the power swing in the power systems with large‐scale wind turbines.Detecting different power swing types including stable, unstable, and multi‐mode.A proper performance during noisy signal.This article is structured as follow: The new method is introduced in the second section. The simulation results are presented in the third section. In the fourth section, the proposed method is compared with some new available methods, and finally the conclusion is drawn in the fifth section.THE PROPOSED METHODPMU is a status estimation device that can measure the current and voltage phasor. In this article, the measurement accuracy is increased using PMU. Typically, PMUs acquire the required variables with the help of global positioning satellite system and using simultaneous sampling. PMU information is stored in the central protection unit and employed in the presented method. Since implementing PMUs increases the total cost, it is not possible to install them in all substations, and an optimization algorithm is required in order to optimize the number and placement of PMUs. However, since the authors in this paper only used PMUs information, and the innovation and the main purpose of this article is not about locating and internal system of PMUs, these subjects are ignored in this article. Refs. [28, 35] can be used for information on the locating of PMUs. Moreover, ref. [28] have suggested a noble method for locating PMUs in order to be implemented with distance relays. Figure 1 shows a single‐line diagram of a two‐circuit power system with a central protection unit. S and E represent the local and remote buses.1FIGURETwo‐circuit single‐line power system diagramIn the presented method of this paper, only the currents of the two ends of the transmission line are used. So, after sampling the current signals and sending the sampled data to the central protection unit (CPU), the FFT and the main component of the signals are extracted from every 6 current phases using (1) [36].1XK=∑n=0N−1Signaln×e−i2πKn/NK=0,…,N−1$$\begin{equation} \def\eqcellsep{&}\begin{array}{l} {X_K} = \sum_{n = 0}^{N - 1} {Signa{l_n} \times {e^{ - i2\pi Kn/N}}} \\[4pt] K = 0,\ldots,N - 1 \end{array} \end{equation}$$The values of Ibase${I_{base}}$, which are obtained from the current of both ends of the transmission line, are calculated using (2) to (4).2Ibase(a)=IS(a)+IR(a)$$\begin{equation}{I_{base}}(a) = \left| {{I_S}(a) + {I_R}(a)} \right|\end{equation}$$3Ibase(b)=IS(b)+IR(b)$$\begin{equation}{I_{base}}(b) = \left| {{I_S}(b) + {I_R}(b)} \right|\end{equation}$$4Ibase(c)=IS(c)+IR(c)$$\begin{equation}{I_{base}}(c) = \left| {{I_S}(c) + {I_R}(c)} \right|\end{equation}$$Subsequently, original signal (Ibase${I_{base}}$) should be decomposed into a number of intrinsic mode functions (IMFs) using EMD. This method is presented by Huang et al. in [37]. The waveform frequency is a constant value that is easily defined. Still, the signals are not practically sinusoidal or constant. Therefore, showing multiple unstable signals as a combination of various components of the signal with a precise evaluation of an event is a challenge [38]. Mainly, the signal empirical decomposition is a purification process in which various modes of swings are extracted from the main signal. The extracted modes are unipolar signals that include one band with a frequency. Hence, a signal can be decomposed to a few intrinsic modes with the following conditions: (1) The extremes of a dataset should be equal to the number of zero points or at least have a difference equal to one number. (2) Its integral should be equal to zero in a defined time interval. The EMD method comprises eight different steps to accomplish this task:First step: The signal decomposition process starts with finding the maximum and minimum points of the input signal.Second step: Creating an upper‐envelope curve through the third‐degree curve‐fitting on the maximum points.Third step: Creating a lower‐envelope curve through the third‐degree curve‐fitting on the minimum points.Fourth step: Mean subtraction of the upper‐envelope curve and lower‐envelope curve of the input signal using (5).5h1(t)=x(t)−m1(t)$$\begin{equation}{h_1}(t) = x(t) - {m_1}(t)\end{equation}$$In (4), h1(t)${h_1}(t)$is the IMF value of the beginning of the signal, x(t)$x(t)$denotes the input signal, and m1(t)${m_1}(t)$ is the mean of the upper‐envelope and lower‐envelope of the curve.Fifth step: Assessing the IMF condition and the stopping criterion using (6).6Dk=∑t=0Th1k−1(t)−h1k(t)∑t=0Th1k−1(t)2$$\begin{equation}{D_k} = \frac{{\sum_{t = 0}^T {\left| {h_1^{k - 1}(t) - h_1^k(t)} \right|} }}{{\sum_{t = 0}^T {{{\left| {h_1^{k - 1}(t)} \right|}^2}} }}\end{equation}$$Sixth step: If the fifth condition is not satisfied, the signal obtained from the fourth step is replaced by the main signal, and the rest of the process is repeated from the beginning.Seventh step: If the fifth condition is met, the screening process is finished and C1=h1k${C_1} = h_1^k$ is considered as the first IMF, which in fact is the high‐frequency component of the signal x(t)$x(t)$.Eighth step: The remaining is defined as Equation (7) and if it satisfies the IMF condition, it will be considered as an IMF, otherwise, if it meets the first condition, it will be considered as an initial signal, and the steps 1–4 are iterated until the next IMF is obtained, and if the first condition is not met, it will be considered as a remaining (r).7r1=x(t)−C1k$$\begin{equation}{r_1} = x(t) - C_1^k\end{equation}$$The main signal is the sum of the IMFs plus the remaining (8).8x(t)=r+∑n=1Nhn$$\begin{equation}x(t) = r + \sum_{n = 1}^N {{h_n}} \end{equation}$$In the final step, the value of the obtained RMS IMFs is calculated using (9) [30].9RMS=∑k=1NIk2N$$\begin{equation}RMS = \sqrt {\frac{{\sum_{k = 1}^N {I_k^2} }}{N}} \end{equation}$$The signal decomposition using EMD results in multiple signals. Therefore, one reference signal must be employed. In this paper, IMF2 will be used as the reference signal. [31] thoroughly explains the method of adopting a reference signal. In addition, using the proposed method of this paper, the power swing and fault can be distinguished by defining a threshold value and using the condition in (10). The threshold value is obtained by using various computer simulations. Studies [31, 37] can be referred to get more information about threshold value detection. The threshold value is equal to 0.1 in this paper.10IfIMF2>Threshold→Faultisdeteced$$\begin{equation}{\rm{If IMF2 > Threshold }} \to {\rm{ Fault is deteced}}\end{equation}$$According to the stated points, Figure 2 shows the algorithm of the presented method.2FIGUREThe presented method algorithmSIMULATION RESULTSTest systemIn order to implement the proposed algorithm, an IEEE standard 39‐bus network is used. The single‐line diagram of this network is illustrated in Figure 3. This network includes 39 buses and 10 generators. Multiple wind farms are connected to buses to evaluate the performance of the suggested method while having a large‐scale wind turbine in the network. The data of these wind farms are presented in Table 1. In this article the DFIG type turbine is implemented which uses a three‐phase induction generator. The implemented generator model is illustrated in the Figure 4. In this generator, the stator winding is directly connected to the grid, while the rotor winding is controlled using an inverter. As its most important advantage, the DFIG can provide electricity with constant frequency via a variable mechanical speed. The DFIG model is considered as a wind farm in the studied network.3FIGURESingle‐line diagram of IEEE 39 bus system1TABLEThe studied wind farm dataRowBusNominal apparent power (MVA)Power factorParallel units1265000.8502264000.8403295000.8504294000.8404FIGUREThe implemented wind turbine modelThe wind farms will merely be connected to the network to assess the first case, however, since the suggested method should be comprehensive, and to ensure the desired performance in the absence of the wind farms, these wind farms will be disconnected from the network while assessing the stable, unstable, and multi‐mode power. Additionally, PMUs are installed at the buses 26 and 29 in order to protect the 26–29 line.The studied network is analysed using Digsilent software and the results are transferred to the MATLAB software to evaluate the suggested protection algorithm. It should be noted that the sampling rate is considered equal to 10 kHz.Case study 1: Fault and power swing occurrence during wind farm connectionsIn this case, the effect of wind farm presence is evaluated on the suggested method. All the wind farms are connected to the network based on all the information provided in the previous section. The power switches of the 26–28 and 28–29 lines are opened for one second to create the power swing. Also, a three‐phase fault is placed on the 50 percent of the line (26–29 line) at 3 s to simulate the fault. Figure 5a shows the current changes in the event of fault and power swing occurrence. As mentioned before, the presence of wind turbines in the network changes the power swing parameter. Figure 5b shows a stable power swing in the absence of a wind turbine in the network. By comparing it to Figure 5a (until 3 s that fault enters the network) it may be observed that we are seeing a completely different signal.5FIGURECurrent waveform during power swing (a) the waveform when there is a wind turbine in the network (b) the waveform when there is no wind turbine in the networkAccording to the previously mentioned results, it is evident that the current methods are not able to detect this type of power swing. To prove this point, multiple power swing detection methods that have been presented in recent years are reviewed. The reviewed methods include the following items:The power swing detection method using chaos theory presented in 2020 [39].The power swing detection method using instantaneous frequency presented in 2020 [38].The power swing detection method based on 3‐phase RMS change rates presented in 2018 [31].The power swing detection method based on moving windows average (MWA) presented in 2014 [25].The power swing detection method is based on signal energy changes presented in 2019 [2].The power swing detection method is based on Taylor series presented in 2016 [48].Figure 6 shows the performance of the 5 mentioned methods during the power swing of Figure 5a. Clearly, all these methods face problems during power swing (the interval of 1–3 s) and incorrectly detects a fault for power swing. Moreover, these methods do not consider any difference between the power swing and fault in the networks with a wind turbine.6FIGUREThe performance of the current methods during power swing in a network with a wind turbine (a) chaos theory, (b) instantaneous frequency, (c) RMS, (d) MWA, (e) Energy (f) Taylor seriesIn addition to the five previously reviewed references, as it was mentioned in the Section 1, [33], which uses EMD like the presented paper, also cannot determine this type of power swing. Figure 7 shows the performance of the suggested method in [33] versus this case. As it is seen in Figure 7, in the event of this type of power swing occurrence, the calculated IMF value exceeds the threshold value defined by the paper. Hence, this reference also cannot determine this type of power swing. It should be noted that if the threshold value is increased to address this issue, fault detection will no longer be possible in the wind turbine‐less networks. Furthermore, when the threshold value increases, the fault detection will no longer be possible in the networks connected to the wind turbines either.7FIGUREThe performance of the suggested method in [33] during power swing in the networks connected to a wind turbineAccording to the mentioned points, presenting an effective method is vital. Figure 8 depicts the suggested method performance of this paper during power swing. Based on this figure, the calculated IMF values do not exceed the threshold value during the power swing (between 1 to 3 s). Whereas, in the event of 3‐phase fault (at 3 s), the calculated IMF values exceed the threshold value and the fault is detected approximately in 4 ms.8FIGUREThe performance of the suggested method of this paper in the networks connected to a wind turbineCase study 2: Detection of stable, unstable, and multi‐mode power swing detection simultaneous with faultAlthough the main purpose of this article is to present a method to differentiate fault and power swing of the networks connected to large‐scale wind turbines, such methods should be able to operate in various conditions. Therefore, the suggested method performance in various situations of power swing and faults are reviewed in this section.The stable power swing can be considered the simplest type of power swing. This type of power swing is not a major challenge for the detection methods. First, all the wind farms are disconnected from the network to evaluate the effect of this type of power swing. Then, the circuit breakers of the 26–28 and 28–29 lines are opened in 0.8 s to create the power swing. A 3‐phase fault is located at 50 percent of the protected lines in 2.1 s to create a fault simultaneous with the power swing. Figure 9 shows the performance of the suggested method during stable power swing with simultaneous fault. As it is seen in this figure, the suggested method has managed to detect the fault almost within 2.32 ms.9FIGUREThe performance of the suggested method of this paper during the stable power swing and 3‐phase fault simultaneous with the power swingThe unstable power swing usually is developed due to extreme turbulence like sudden adding or removing a major part of a load, 3‐phase fault etc. Unlike stable power swing, this type of swing does not end after a few cycles. Although it is needed to disconnect the unstable synchronous generator from the network to end this type of power swing, the protected line should remain in the circuit so that it can supply the connected loads through other available resources. To simulate this type of power swing, first, a 1000 MW and a 200 MVAR loads are connected to the network in 0.8 s. Then, a 3‐phase fault is placed on the 50 percent of the protected lines in 2.5 s to simulate a simultaneous fault with the power swing. Figure 10 shows the performance of the suggested method when the unstable power swing occurs with simultaneous 3‐phase fault. As it is seen in this figure, the suggested method has managed to detect the fault almost within 1.95 ms.10FIGUREThe performance of the suggested method of this paper during unstable power swing and 3‐phase fault simultaneous with power swingA major challenge for power swing detection methods is the multi‐mode power swing detection. Unlike the stable and unstable power swing, the multi‐mode power swing has no constant swing frequency. Additionally, in some parts of this type of power swing, the current signal is distorted and is not completely sinusoidal. This type of power swing occurs when multiple synchronous generators swing relative to each other. The protected line of 17–18 is considered to evaluate the effect of this type of power swing. A 3‐phase fault is placed at 50 percent of the 4–14 line to create a multi‐mode power swing. The power switches of the 4–14 line are opened in 300 ms. This helps the protection system of the 17–18 line to detect the multi‐mode power swing. Figure 11 depicts the suggested method performance of this paper during multi‐mode power swing. According to this figure, even during 3‐phase fault simultaneous with the power swing, the suggested method still can detect the fault within 6.02 ms.11FIGUREThe performance of the suggested method of this paper during the multi‐mode power swing and 3‐phase fault simultaneous with the power swingThe main advantage of the suggested method is the multi‐mode power swing and fault detection time reduction. Table 2 shows the comparison results of the suggested method of this paper with the references presented in Sections 2 and 3. The references are compared with the suggested method in this paper in terms of fault detection time and multi‐mode power swing detection capability. As it is shown in this table, most reviewed methods cannot determine the multi‐mode power swing. Only the methods presented in [33, 39] can determine this type of power swing. Additionally, by comparing fault detection time, it can be concluded that the suggested method can enhance the fault detection time as well.2TABLEComparison of the suggested method in this paper with similar methods based on fault detection time and multi‐mode power swing detectionReferenceMethodFault detection time (ms)Multi‐mode power swing[39]Chaos7–8Yes[40]Instantaneous frequency31–33No[30]RMS100No[25]MWA16No[2]Energy10–11No[33]EMD and Hilbert7–9YesProposed methodEMD and PMU2–6YesCase study 3: Signal noiseNoise is an inseparable part of any actual system. It is almost impossible to build a noise‐free system. Therefore, it is better to evaluate the noise tolerance of the power swing detection methods . Most protection studies conducted in recent years have used white Gaussian noise to explore the effect of noise on the performance of their methods. For example, ref. [42] has used a power transformer to assess the effect of noise on the protection method, and ref. [43] has used white Gaussian noise to explore the impact of noise on the fault locating method. Several references have also used white Gaussian noise to assess the noise effect in the field of power swing detection such as [7, 31, 37]. Figure 12 shows the performance of the suggested method when the noise occurs in the signal with various signal to ratios (SNRs). As it is shown in Figure 12, the suggested method of this paper has a desirable performance in white Gaussian noise with SNR = 40 dB and SNR = 30 dB, however, when white Gaussian noise is added to the main signal with an SNR of 20 dB, the suggested method will face problems.12FIGUREThe performance of the suggested method of this paper during white Gaussian noise (a) SNR = 40 dB, (b) SNR = 30 dB, (c)SNR = 20 dBAmong the currently reviewed references, the presented methods in [31, 37] are resistant against white Gaussian noises up to SNR = 30 dB Whereas, the refs. [2, 24, 30, 38] are not noise‐resistant and are extremely affected by the noise.Case study 4: Asymmetric power swingThe power swings are not always symmetrical and during some events such as single pole switching or automatic reclosing of a single pole an asymmetrical power swing may occur [39]. As mentioned in the introduction section, asymmetric power swing can cause problems for existing methods. On this basis, the performance of the presented method during asymmetric power swing is investigated. In order to create asymmetric power swing, phase A switches of lines 26–28 and 28–29 are opened in 1.2 s. This causes asymmetric power swing (only phase A starts to oscillate) on the protected line (line 26–29). In order to investigate the subject thoroughly, a three‐phase fault is placed on 50% of line 26–29 in 3 s. Figure 13 presents the performance of the proposed method during asymmetric power swing which is simultaneous with a three‐phase fault. According to this figure, the proposed method has operated appropriately during asymmetric power swing.13FIGUREThe performance of the presented method during the asymmetric power swingCase study 5: The fault detection accuracyIn order to investigate this case, 500 different events are randomly generated in DigSilent software and applied to the algorithm in MATLAB software. From these 500 events, 100 events cause stable power swing, 100 events cause unstable power swing, and the other 300 events cause various types of faults (single‐phase, two‐phase, and three‐phase). The results of this test are presented in the Table 3. According to this table, in the analysis of stable and unstable power swing, 100% of the events have been detected by the algorithm. In case of single‐phase and two‐phase faults, the fault detection accuracy percentage is 99% and for three‐phase faults it is 98%. In general, it can be concluded that the presented algorithm has been operated correctly in 99.2% of the time.3TABLEThe fault detection accuracyEventsFault detection accuracy (%)Stable power swing100Unstable power swing100Single‐phase fault99Two‐phase fault99Three‐phase fault98All events99.2COMPARING THE PROPOSED METHOD WITH SIMILAR METHODSIn recent years, various methods have been suggested to determine power swing. Each method has its specific signals, signal process tools, sampling frequency etc. Although the suggested method in this paper is compared with the current methods in the simulation section, a general comparison is presented in this section based on different factors. Table 4 shows a comprehensive comparison between the proposed method with other current methods for power swing detection.4TABLEThe comprehensive comparison of the proposed method and the other methods for power swing detectionReferenceParameterMethodSampling rate (kHz)Symmetrical fault detectionFaster than one cycleEasy to implementWind turbine[44]Active and reactive powerRate of change of power2YesNoYesNo[10]Current and voltageAutomatic regression1YesNoYesNo[36]Active powerFFT10YesYesYesNo[21]CurrentProny3YesYesYesNo[45]CurrentWavelet40.96YesYesYesNo[18]CurrentTravelling wave10YesYesNoNo[42]CurrentPhaselet10YesYesYesNo[39]CurrentGMDH10YesYesYesNoProposed methodCurrentEMD‐PMU10YesYesYesYesCONCLUSIONToday, technological development has increased the electric energy consumption in the world. The increased electric energy consumption, concerns over natural resources, and carbon emission has caused the distributed generation to be considered as a useful solution. Among the different kinds of distributed generations, the use of large‐scale wind turbines is increasing in the level of the power system. The presence of such wind turbines in combination with the synchronous generators can cause changes in the power swing parameters of the power system. These changes may cause malfunction of the power swing detection methods and lead to undesirable tripping of the healthy power lines. Therefore, a modified EMD method was presented in this paper using PMU. The suggested method can easily differentiate power swing from fault in the systems connected to the large‐scale wind turbines. Moreover, the suggested method is also applicable in systems without wind turbine. The simulation results and the comparisons that are made with similar methods indicate the efficiency and applicability of the proposed method. In the following, using a statistical analysis including 500 different events, the efficiency and the accuracy of the presented method is tested and is calculated to be 99.6%.In the following, using a statistical analysis including 500 different events, the efficiency and the accuracy of the presented method is tested and is calculated to be 99.6%. Obviously, the proposed approach of this article and most of the reviewed references use a threshold value to distinguish between power swing and fault. Using a threshold value can cause difficulties and challenges for operators and researchers. In addition, determining the threshold value can be a sophisticated and time‐consuming task. For this reason, the authors suggest the research on methods of determining faults without the need of a threshold value or methods to determine the threshold value for future works.AUTHOR CONTRIBUTIONSA.A.N.: Investigation; Methodology; Software; Validation; Writing – original draft; Writing – review & editing. F.R.: Methodology; Project administration; Software; Writing – original draft; Writing – review & editing. A.F.: Formal analysis; Methodology; Validation; Writing – review & editing.CONFLICT OF INTERESTThe authors declare no conflict of interest.DATA AVAILABILITY STATEMENTData sharing not applicable to this article as no datasets were generated or analysed during the current study.REFERENCESJannati M., Mohammadi M.: A novel fast power swing blocking strategy for distance relay based on ADALINE and moving window averaging technique. IET Gener. Transm. Distrib. 15(1), 97–107 (2021)Taheri B., Salehimehr S., Razavi F.: A new method for fast power swing detection using the rate of change of energy in the current signal. In: 2019 27th Iranian Conference on Electrical Engineering (ICEE). Yazd, Iran (2019)Taheri B., Faghihlou M., Salehimehr S., Razavi F.: Symmetrical fault detection during power swing using mean value of sampled data from the current signal. IETE J. Res. 1–13 (2020)Bakar A.H.A., Yatim F.M., Yusof S., Othman M.R.: Analysis of overload conditions in distance relay under severe system contingencies. Int. J. Electr. Power Energy Syst. 32(5), 345–350 (2010)Gunasegaran M., Tan C., Bakar A.H.A., Mokhlis H., Illias H.A.: Progress on power swing blocking schemes and the impact of renewable energy on power swing characteristics: A review. Renewable Sustainable Energy Rev. 52, 280–288 (2015)Salehimehr S., Taheri B., Faghihlou M.: Detection of power swing and blocking the distance relay using the variance calculation of the current sampled data. Electr. Eng. 104, 913–927 (2021)Taheri B., Sedighizadeh M.: Detection of power swing and prevention of mal‐operation of distance relay using compressed sensing theory. IET Gener. Transm. Distrib. 14(23), 5558–5570 (2020)Juttu T.R., Bhalja B.R., Andreev M., Malik O.P.: Synchrophasor assisted power swing detection scheme for wind integrated transmission network. IEEE Trans. Power Delivery. 37(3), 1952–1962 (2021)Haddadi A., Kocar I., Karaagac U., Gras H., Farantatos E.: Impact of wind generation on power swing protection. IEEE Trans. Power Delivery 34(3), 1118–1128 (2019)Rao J.G., Pradhan A.K.: Differential power‐based symmetrical fault detection during power swing. IEEE Trans. Power Delivery 27(3), 1557–1564 (2012)Khoradshadi‐Zadeh H.: Evaluation and performance comparison of power swing detection algorithms. In: IEEE Power Engineering Society General Meeting. San Francisco, USA. pp. 1842–1848 (2005)Martuscello L., Krizauskas E., Holbach J., Lu Y.: Tests of distance relay performance on stable and unstable power swings reported using simulated data of the August 14 th 2003 system disturbance. In: 2009 Power Systems Conference. Clemson, USA. pp. 1–21 (2009)Benmouyal G., Hou D., Tziouvaras D.: Zero‐setting power‐swing blocking protection. In: 31st Annual Western Protective Relay Conference. pp. 19–21 (2004)Karegar H.K., Mohamedi B.: A new method for fault detection during power swing in distance protection. In: 2009 6th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology. Chonburi, Thailand. pp. 230–233 (2009)Mahamedi B.: A very fast unblocking scheme for distance protection to detect symmetrical faults during power swings. In: 2010 Conference Proceedings IPEC. Singapore. pp. 378–383 (2010)Brahma S.: Use of wavelets for out of step blocking function of distance relays. In: 2006 IEEE Power Engineering Society General Meeting. Montreal, Canada. p. 5 (2006)Pang C., Kezunovic M.: Fast distance relay scheme for detecting symmetrical fault during power swing. IEEE Trans. Power Delivery 25(4), 2205–2212 (2010)Mahamedi B.: A new power swing blocking function based on wavelet transform. In: 2011 2nd International Conference on Electric Power and Energy Conversion Systems (EPECS). Sharjah, UAE. pp. 1–6 (2011)Dubey R., Samantaray S., Tripathy A., Babu B.C., Ehtesham M.: Wavelet based energy function for symmetrical fault detection during power swing. In: 2012 Students Conference on Engineering and Systems. Allahabad, India. pp. 1–6 (2012)Lotfifard S., Faiz J., Kezunovic M.: Detection of symmetrical faults by distance relays during power swings. IEEE Trans. Power Delivery 25(1), 81–87 (2009)Seethalekshmi K., Singh S., Srivastava S.: A classification approach using support vector machines to prevent distance relay maloperation under power swing and voltage instability. IEEE Trans. Power Delivery 27(3), 1124–1133 (2012)Zadeh H.K., Li Z.: A novel power swing blocking scheme using adaptive neuro‐fuzzy inference system. Electr. Power Syst. Res. 78(7), 1138–1146 (2008)Rao A.K., Ahmad A.: Power swing blocking (PSB) function for distance relay using prediction technique. Int. J. Syst. Assur. Eng. Manage. 8(2), 301–307 (2017)Rao J.G., Pradhan A.K.: Power‐swing detection using moving window averaging of current signals. IEEE Trans. Power Delivery 30(1), 368–376 (2014)Alsyoufi Y.R., Hajjar A.A.: A high‐speed algorithm to discriminate between power swing and faults in distance relays based on a fast wavelet. Electr. Power Syst. Res. 172, 269–276 (2019)Patel B., Bera P.: Fast fault detection during power swing on a hybrid transmission line using WPT. IET Gener. Transm. Distrib. 13(10), 1811–1820 (2019)Rao J.T., Bhalja B.: Prevention of maloperation of distance relay under severe stressed conditions for series compensated transmission line considering optimal placement of phasor measurement units. IET Gener. Transm. Distrib. 14(11), 2148–2159 (2020)Taheri B., Salehimehr S., Razavi F., Parpaei M.: Detection of power swing and fault occurring simultaneously with power swing using instantaneous frequency. Energy Syst. 11(2) 491–514 (2020)Patel B.: A new technique for detection and classification of faults during power swing. Electr. Power Syst. Res. 175, 105920 (2019)Taheri B., Razavi F.: Power swing detection using rms current measurements. J. Electr. Eng. Technol. 13(5), 1831–1840 (2018)Taheri B., Hosseini S.A., Abyaneh H.A., Razavi F.: Power swing detection and blocking of the third zone of distance relays by the combined use of empirical‐mode decomposition and Hilbert transform. IET Gener. Transm. Distrib. 14(6), 1062–1076 (2020)Biswal S., Biswal M.: Fault‐swing discrimination using Hilbert–Huang transform integrated discrete teager energy operator. IET Sci. Meas. Technol. 12(7), 829–837 (2018)Parpaei M., Ghaffarzadeh N., Tayar M.Z.: Rational random walk‐based optimal placement of phasor measurement units to enhance the initializing and guiding the optimization processes. Int. Trans. Electr. Energy Syst. 27(7), e2323 (2017)Taheri B., Faghihlou M., Salehimehr S., Razavi F.: A fast Fourier transform‐based method for power swing detection and distance relay malfunction prevention. J. Control Autom. Electr. Syst. 31(6), 1458–1468 (2020)Huang N.E., et al.: The empirical mode decomposition and the Hilbert spectrum for non‐linear and non‐stationary time series analysis. Proc. R. Soc. London, Ser. A 454(1971), 903–995 (1998)Shukla S., Mishra S., Singh B.: Empirical‐mode decomposition with Hilbert transform for power‐quality assessment. IEEE Trans. Power Delivery 24(4), 2159–2165 (2009)Salehimehr S., Taheri B., Razavi F., Parpaei M., Faghihlou M.: A new power swing detection method based on chaos theory. Electr. Eng. 102(2), 663–681 (2020)Taheri B., Salehimehr S., Razavi F., Parpaei M.: Detection of power swing and fault occurring simultaneously with power swing using instantaneous frequency. Energy Syst. 11(2), 491–514 (2020)Hosseini S.A., Taheri B., Abyaneh H.A., Razavi F.: Comprehensive power swing detection by current signal modeling and prediction using the GMDH method. Prot. Control Mod. Power Syst. 6(1), 1–11 (2021)Ghalesefidi M.M., Ghaffarzadeh N.: A new phaselet‐based method for detecting the power swing in order to prevent the malfunction of distance relays in transmission lines. Energy Syst. 12(2), 491–515 (2021)Hashemi S.M., Sanaye‐Pasand M., Shahidehpour M.: Fault detection during power swings using the properties of fundamental frequency phasors. IEEE Trans. Smart Grid 10(2), 1385–1394 (2017)Taheri B., Sedighizadeh M.: A moving window average method for internal fault detection of power transformers. Cleaner Eng. Technol. 4, 100195 (2021)Taheri B., Salehimehr S., Sedighizadeh M.: A fault location algorithm for parallel line based on the long short‐term memory model using the distributed parameter line model. Int. Trans. Electr. Energy Syst. e13032 (2021)Lin X., Gao Y., Liu P.: A novel scheme to identify symmetrical faults occurring during power swings. IEEE Trans. Power Delivery 23(1), 73–78 (2007)Brahma S.M.: Distance relay with out‐of‐step blocking function using wavelet transform. IEEE Trans. Power Delivery 22(3), 1360–1366 (2007)Tekdemir, I.G., Alboyaci B.: A novel approach for improvement of power swing blocking and deblocking functions in distance relays. IEEE Trans. Power Delivery 32(4), 1986–1994 (2016)

Journal

IET Generation Transmission & Distribution – Wiley

Published: Mar 1, 2023

Keywords: distance relay; large‐scale wind farms; power swing; power system protection

There are no references for this article.