مکان‌یابی خطای بهبودیافته با به‌کارگیری نشانگرهای خطا و مدل احتمالاتی بار

امروزه هوشمندسازی شبکه‌های توزیع از موضوعات چالشی و جدید محسوب می‌شود. یکی از مشخصه‌های هوشمندسازی در این شبکه‌ها قابلیت مکان‌یابی سریع و دقیق خطا در آن‌هاست. این موضوع موجب کاهش بخشی از زمان خاموشی‌هایی می‌شود که گروه‌های تعمیرات صرف جست‌وجوی محل خطا به روش سنتی می‌کنند. اما مکان‌یابی خطا در شبکه‌های توزیع با توجه به ذات توپولوژیک آن‌ها (انشعابات متعدد، بارهای میانی، خطا در تخمین میزان بارها، نا همگن بودن خطوط و...) کار بسیار دشوار و متفاوت از شبکه‌های انتقال برق است. در این تحقیق، برای کاهش تعداد جواب‌های مکان‌یاب خطا از اطلاعات نشانگرهای خطا استفاده شده است. در این مقاله، برای مکان‌یابی خطا، روشی مبتنی بر جاروب پس‌رو پیش‌رو معرفی شده است. در این راستا با توجه به تخمینی بودن میزان بارهای موجود در شبکه، به‌طبع با پارامترهای دارای عدم قطعیت سروکار خواهیم داشت. به‌دلیل شرایط آب‌وهوایی تقریباً یکسان برای فیدرهای توزیع و رفتارهای گاه مشابه مشترکین، بین این پارامترها همبستگی نیز وجود خواهد داشت. لذا برای مدل‌سازی مسئله و به‌دست‌آوردن تاثیر پارامترهای احتمالاتی بر روی محل محاسبه‌شدۀ خطا از دو روش مرسوم مونت کارلو و تخمین دونقطه‌ای استفاده شده است. در نهایت، برای ارزیابی روش پیشنهادی از شبکۀ تست 11 باس، و شبکۀ واقعی 306 باسه فیدر شرف‌آباد در شهر کرمان استفاده شده است.

Improved Fault Location Using Fault Indicators and Loads Probabilistic Model

Introduction: The SAIDI is one of the principal indices in today #39;s restructured electricity network. Locating the fault occurred in the distribution feeders will remarkably increase the SAIDI and decrease commercial losses. Therefore, it is required to have a system that compute the fault location in the shortest possible time. Due to complexity and excessive number of buses in distribution networks, fault location cannot be handled by human operators alone. Today #39;s advanced networks involve smart grid infrastructure. Meanwhile, fault location is one of the basic components in smart grids. Therefore, if the fault point is computed by fault locator, it is practicable to remove automatically the destructed equipment and also recover the healthy part of the distribution network with remote control switches. Materials and methods: A backward forward sweep impedancebased fault location considering probabilistic correlated loads for real distribution systems is used. The backward forward sweep method is introduced to model the intermediate loads and several laterals along the feeder during the fault. To apply the fault location algorithm in the proposed method, initially the fault is detected and classified. This can deal with the protection relays in the substation. In the absence of protection relay outputs, some methods are used to detect and classify faults. After identifying the time of occurred fault and its type, the fault location algorithm needs the loads values along the feeder. However, the loads are not generally available by online measurement equipments. Therefore, the value of loads must be determined from the load forecasting algorithms; and the values of predicted loads always involve error. Therefore, these values are modified by measuring the total load at the substation. Then, a hybrid method is presented to reduce the number of multiple results of fault location. In the proposed algorithm, the fault detectors data and the voltage and current signals in the substation are used together. Result: To show the efficiency of the proposed method, two networks with different complexity (the 11bus test network and the real 306bus network of KermanIran) are considered for simulations. The predicted loads are often correlated in power grids. Therefore, the effect of the correlation between the loads on the standard deviation of the output error is one of the most important and attractive cases for researchers. The correlation is increased from 0 to 0.9 to show this effect. The standard deviation parameter of the Gaussian normal distribution function for probabilistic loads is considered to be 0.35. The standard deviation of the output error for the different states of this scenario is given in the following table. In this regard, the fault resistance is assumed to be 100 ohms and the type of fault is singlephase to ground. The standard deviation of the output error in meters for different fault distances and different correlations between loads in 11 busDis (m)Correlation00.10.20.30.40.50.60.70.80.93000.930.940.840.900.990.830.790.800.780.726000.950.930.800.850.910.720.650.650.610.539000.980.940.780.800.840.620.530.500.440.3312001.020.960.770.780.790.550.440.390.310.1615001.081.000.790.780.760.510.400.350.290.1718002.632.542.482.582.833.013.383.904.495.1621009.468.978.698.869.6110.7712.4714.6217.0819.91240016.5415.6015.0515.2816.5318.6421.6725.4529.7934.79270038.2337.3938.3541.7047.1453.9461.5369.6778.8188.28300064.4362.7564.3969.9478.5488.2398.35108.08116.57122.913300107.57104.53106.87114.94127.04140.80154.90167.56177.92184.25A comparison is made between the results of the proposed method and those obtained from the methods introduced in the papers of Mr. Salim, Nouri, and Gabr to show the superiority of the proposed method over the previous ones. The real306busnetwork of KermanIran has been used for this evaluation. In this regard, the standard deviation parameter in the Gaussian normal distribution function of the loads is 0.3; the fault resistance is 100 ohms; and the correlation rate between loads is 0.1 for area A, 0.3 for area B, and 0.2 for area C. According to the results of the simulation in following table, it is clear that the output obtained from the proposed method does not diverge at the end of the feeder, using the probabilistic model. The results of the previous methods, due to their definite nature, diverged for faults located at the end points of the feeder. On the other hand, the number of multiple results is reduced in the proposed method because the fault detector data is used to determine the faulty branches. As a result, a distribution dispatcher can uninterruptedly issue the necessary commands to disconnect the faulty part and recover the other healthy parts of the network in minimum time. This significantly improves SAIDI in the distribution network. The comparison of the proposed method with previous methods in 306 bus Salim 2009Nouri 2011Gabr 2017Proposed methodDis (km)maximum error from real fault point (m)maximum multiple resultsmaximum error from real fault point (m)maximum multiple resultsmaximum error from real fault point (m)maximum multiple resultsmaximum error from real fault point (m)maximum multiple results13.111.111.310.91225.1217.3217.9215.61451.2327.9329.6326.116181.6675.3670.3673.818251.25187.85182.15184.2110divergeddiverged287.56280.36273.9212divergeddiverged340.45361.25356.4114divergeddivergeddivergeddiverged180.14177.6216divergeddivergeddivergeddivergeddivergeddiverged428.71Discussion and Conclusion: Fault locations are of particular importance in today #39;s distribution networks. Accurate performance of the fault locator will have a significant impact in improving SAIDI in distribution networks. Therefore, due to the nature of distribution networks, there are several important challenges in networks. The problem of multiple results in distribution networks is minimized by receiving signals from the fault detectors that are optimally placed along the feeder. As another issue, consumer loads are not under the control of distribution companies. On the other hand, the discussion of load forecasting is always faced with deviations from the actual value. In this regard, probabilistic fault location in distribution networks was evaluated as a suitable solution to this problem. The probabilistic fault location determines the search length for the Event Group. The search length is sent to the distribution dispatcher with its probability. The probabilistic fault location method was implemented for the real 306bus network in KermanIran. The results indicate that the proposed method is operational. In probabilistic fault location for high impedance faults that occur at the end points of the feeder, divergence does not occur in the search algorithm. Therefore, probabilistic fault location is a better method with an advantage of a greater convergence compared to the definite methods.