List of Articles Hamid Reza Safa

• Open Access Article
  • Abstract Page
  • Full-Text
  
  1 - Broken Conductor Fault Detection in Transmission Lines Connected to Renewable Energy-Based Microgrids
  Hamid Reza Safa Ali Asghar Ghadimi
  The connection of renewable energy-based microgrids in transmission lines has significantly increased recently. The presence of REMs, along with the advantages they provide, also leads to problems from different aspects of operation, control, and protection in transmiss More

  The connection of renewable energy-based microgrids in transmission lines has significantly increased recently. The presence of REMs, along with the advantages they provide, also leads to problems from different aspects of operation, control, and protection in transmission lines. The direct connection of REMs in the form of T-off in the transmission lines and without the construction of a substation, causes a severe disturbance in the performance of the protection algorithms of the line protection relays. This paper presents a fault detection method in transmission lines connected to REMs for early detection of Broken Conductor Fault (BCF) based on the information of one side of the line (the sending terminal) and using the teaching-learning artificial neural networks (ANNs). The neural network considered in this study is a combination of convolutional neural network and long short-term memory (CNN-LSTM). The hybrid model includes a Conv1D layer with 64 filters and a kernel size of 3, a MaxPooling1D layer, two LSTM layers with 32 units, a Dropout layer and a Dense layer with one unit and sigmoid activation. The necessary data for training the desired ANN have been extracted from the simulation of the main network and the implementation of various fault scenarios in MATLAB/Simulink software, and finally the considered ANN model has been programmed and modeled in the Python software environment. According to the simulation results, the accuracy of the extracted model in detecting the BCF in this proposed topology is estimated to be about 99.73%. The successful results presented in the test and evaluation results section confirm the optimal performance of the proposed algorithm. Manuscript profile