Online State Monitoring of Distribution Cable Faults Based Empirical Mode Decomposition with Convolutional Neural Network and Spiking Neural Network

In recent years, composited distribution cable terminals are important component of supplying power, which is extensively used in traction power supply systems. Traditional approaches for online state monitoring of distribution cable faults had faced several challenges which include more noise in data and contains only time domain features which is difficult to monitor cable faults. Therefore, this paper proposes Empirical Mode Decomposition-Convolutional Neural Network-Spiking Neural Network (EMD-CNN-SNN) for online state monitoring of distribution cable faults which is employed on electric fault line detection dataset, available in Kaggle. This dataset provides real time monitored three phase current signals and fault related parameters of distribution cables. The data is taken and preprocessed by using Fast Fourier Transform (FFT) which is used to renovate time domain signals into frequency domain signals. Then, these preprocessed signals i.e. optimal features were selected by using EMD-CNN which is suitable for real time data, robust to noise and interference in signals. Finally, SSN is used for classification of optimal features and effectively diagnosed the distribution cable faults. The proposed EMD-CNN-SNN achieved accuracy and precision of (99.8<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) and (98.3<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) when compared with existing models like Artificial Neural Networks-Support Vector Machine (ANN-SVM).