LSTM Based Approach for Timely Detection of Gradual Development of Electrical Fault in Power System

Power system reliability and efficiency are becoming a primary concern with the increase in load and expansion of power grids. Electrical faults in the power system should be detected and cleared immediately due to their critical impact on the reliability and stability of the system. This paper proposes an approach to predict the faults in the power system using machine learning techniques like Long Short-Term Memory (LSTM). The LSTM model is used to predict gradual faults in the system before their actual occurrence. Three-phase measurements of voltages, currents, and active power during faults and normal operating conditions are taken as data inputs to train the models. The robustness of the method is verified by simulating the fault with different parameters. The proposed method can be expanded to the distribution network of the power system. A modified IEEE 9 bus system is modelled in MATLAB/Simulink and is used to get the data for the experiment. The results from the experiment prove the feasibility of using LSTM networks for predicting the faults in the power system.