A Flexible Protection Approach for DC Microgrid Using Wavelet based Machine Learning Technique

In this paper, a protection scheme for a DC micro-grid is proposed. The proposed protection scheme’s novelty lies in its usage of concepts like K-means clustering, wavelet analysis, and machine learning techniques for fault detection. Current sample data is considered for different fault events. Using the K-means clustering, a low impedance fault is detected. For the discrimination of high impedance fault and sudden change in load condition, the data is processed with discrete wavelet transform and then fed into machine learning classifiers to get the fault information. The datasets are trained using an SVM classifier to obtain the results. The speed of operation of the classifier suits the requirement. A comparative analysis is done using other classifiers like KNN and NB for the best accuracy results. The proposed algorithm is tested on 5 bus ring-type DC microgrid using MATLAB/SIMULINK.