Improved Wavelet Based Artificial Neural Networks for Robust DC Fault Detection of MTDC Grid

In the transient protection of multi-terminal direct current (MTDC) grids based on modular multilevel converters (MMC), fand a fast and reliable fault detection method is a significant technical challenge. Existing fault detection schemes have issues such as reliance on threshold selection, susceptibility to high-impedance faults, and long detection times. This paper proposes an intelligent fault detection method for MTDC grids based on artificial neural networks (ANN). According to the boundary conditions of the grid, the improved discrete wavelet transform (DWT) and karenbauer transformation are utilized to extract the features of the fault signal for neural network training, achieving fault detection in the DC grid. Simulations are conducted using PSCAD/EMTDC and MATLAB. The robustness of the trained network is tested under various conditions through many samples. The experimental results show that this method can accurately and quickly detect DC faults and has good robustness.