Early Detection of Abnormal Grid Condition of an Integrated Transmission and Distribution System Using PMU Data

نوع: Type: thesis

مقطع: Segment: PHD

عنوان: Title: Early Detection of Abnormal Grid Condition of an Integrated Transmission and Distribution System Using PMU Data

ارائه دهنده: Provider: Iraj Gan Khani

اساتید راهنما: Supervisors: Alireza Hatami

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Mohammad Hassan Moradi , Hadi Delavari, Abbas Fattahi Meyabadi

زمان و تاریخ ارائه: Time and date of presentation: November 25, 2023 - 16-18

مکان ارائه: Place of presentation: Faculty of Engineering Amphitheater Hall

چکیده: Abstract: Voltage violation and flow violation are the most important security constraints in power systems. The long-term voltage stability, which covers the voltage violation constraint, is one of the sub-branches of voltage stability, which occurs in a period of several seconds to several minutes. Several factors are involved in long-term voltage instability. The ability of generators to provide active and reactive power required by consumers, On-load Tap Changers, load model, operating point and the studied network play an important role in voltage stability studies. One of the main reasons for blackouts around the world is the user's lack of awareness that the operating point is approaching the collapse point, and the blackouts in different countries, which are rooted in long-term voltage instability, prove the importance of this issue. in such a circumstance, the presence of an auxiliary tool in addition to the SCADA system in the control centers that tracks the Phasor Measurement Unit Data online and informs the operator of abnormal conditions as soon as possible is very valuable. Long-term voltage instability usually occurs slowly along with load growth or loss of generation. As the load grows, reactive power is compensated by generators. The ongoing process of these conditions along with the operation of On-load Tap Changers reduces the power reserve of reactive generators. When any generator hits its reactive limit (reactive power limit violated), it loses any control, whatever, of its terminal voltage and the reactive output. above mentioned circumstances can be the beginning of long-term voltage instability. Long-term voltage stability in transmission and distribution networks are usually investigated separately. To evaluate instability, simplified models are used in the simulation, which may slightly exaggerate the margin of instability. Authors are usually satisfied with providing an index that shows the proximity of the network's working point to the collapse point. Among the problems of existing methods, we can mention long response time, lack of accurate modeling and low accuracy. Using the capabilities of deep recurrent neural networks in solving time series problems, this thesis presents its applications in the field of long-term voltage instability. The neural network monitors the input changes online and notifies the network operator if the operating point of the network approaches the collapse point. The neural network training process requires information in both stable and unstable states. The required data is rare in an unstable state, that's why simulation is used in this thesis to create training and test data. The extended Nordic test system has been used to evaluate the proposed method. To implement the distribution network, the integrated load of the transmission network has been replaced by standard distribution systems. The simulation has been done as a continuous load flow using Digsilent software. Various scenarios have been generated by considering step load growth in areas vulnerable to power system voltage instability (central), along with (N-1) and (N-1-1) contingencies. After data mining, the outputs were compiled in Excel format and imported by MATLAB software. In the following, the proposed neural network is trained by the data obtained from the simulation and evaluated using the test data. The accuracy of the neural network is presented in the fifth chapter. In order to estimate the voltage instability with the least possible time delay after the contingency and also the most possible accuracy, a structure consisting of several parallel neural networks that are individually trained with different parameters is presented. The results related to the evaluation of this structure are also presented in the fifth chapter. Finally, using the clustering method, the inputs of the neural network are reduced, while the output does not change significantly. The results show that the LSTM neural network can reveal long-term voltage instability with acceptable accuracy a few seconds after the contingency. Finally, the presented method has been compared with the results of the methods available in the sources

فایل: ّFile: