Short Circuit Current Prediction of Large Power Grid Based on Continual Spatio Temporal Graph Convolutional Network

In recent years, prediction of short circuit current prediction is important for the safety and reliability of electrical power systems, enabling the proper equipment sizing and protective measures. The accurate prediction helps in preventing equipment damage and ensures effective fault management. Traditional approaches for short circuit current prediction had faced several challenges which include inaccurate modeling of power system components and complexity of power system. Therefore, this research proposes Continual Spatio Temporal Graph Convolutional Network (CSTGCN) to accurately predict short circuit current of large power grid. Initially, the data is taken from Institute of Electrical and Electronics Engineers (IEEE) 39-bus system which consists of short circuit data, transmission line data and generator data. Then, this data is preprocessed by using Short Time Fourier Transform (STFT) which reduce noise in the signal by filtering out unwanted frequencies. After that, Hilbert Transform is used to extract features like amplitude, phase and frequency. Finally, prediction is by using proposed CSTGCN which captures spatio-temporal dependencies between power system components by allowing more accurate modeling of their interactions and behaviors. The proposed CSTGCN achieved better results in Mean Absolute Percentage Error (MAPE) (4.21) and Root Mean Square Error (RMSE) (0.06) when compared with existing LeNetwork5.