A Multi-Layer Perceptron Neural Network for Fault Type Identification for Transmission Lines

Electric Transmission Lines (TLs) frequently experience faults. Electric faults not only impose extremely adverse stress on grid apparatus, but they also disrupt power flow in grids which may result even in blackouts. Due to the growing demand for electric power all over the world, quick grid restoration is utmost necessary. To this end, identification of the faulty phases is remarkably important. In this article, a Multi-Layer Perceptron Artificial Neural Network (MLP-ANN)-based fault type identification approach is proposed. This approach employs the Discrete Wavelet Transformation (DWT) for time-frequency analysis of three-phase single-ended voltage and current measurements. The DWT also filters out the noises present in the measurements to some extent. The extracted features of the filtered measurements are then utilized to calculate the energy which provides necessary input features to the MLP-ANN for training. Then, the trained ANN identifies the fault types. The power system is modelled in Matlab/Simulink and the approach is implemented in Matlab. The performance of the proposed approach is evaluated, and the outcomes show that the MLP-ANN is able to identify the type of faults with high precision.