High Impedance Fault Modelling and Simulation of 33kV/11kV Distribution Network Using MATLAB

The characteristic features of High Impedance Faults (HIFs) have made their detection and location an uphill task, as conventional protection systems do not detect or locate them. In the quest to develop a deep learning-based technique (which is purely data-driven), for HIF detection and location, real-world HIF data which is unavailable became a necessity. The dearth of HIF data and records necessitated this study, to obtain the fault data required for training a fully connected convolutional neural network to detect and locate HIF in the case-study network. To obtain HIF data, a well-proven HIF model was adopted, modeled, and simulated in a 33kV/11kV distribution network using MATLAB software. To reproduce practical fault waveforms and current magnitudes for diverse scenarios, the HIF model parameters were continuously varied. The results obtained are the current waveforms and magnitude for all scenarios considered, these were exported to the workspace and stored. The results showed that the fault current waveforms exhibited the characteristics of HIF, and the current magnitude was below detection level at high impedance values. The training data was obtained by simulating different fault types, a no-fault case, and different contact surfaces at different locations and phases. The impedance-current relationship obtained by varying the parameters of the HIF model has been presented. In total, the training dataset obtained contained 1400635 rows by 108 columns time series data capturing HIF features.