A Combined Wavelet and Data-Mining Based Intelligent Protection Scheme for Microgrid

Abstract

This paper presents an intelligent protection scheme for microgrid using combined wavelet transform and decision tree. The process starts at retrieving current signals at the relaying point and preprocessing through wavelet transform to derive effective features such as change in energy, entropy, and standard deviation using wavelet coefficients. Once the features are extracted against faulted and unfaulted situations for each-phase, the data set is built to train the decision tree (DT), which is validated on the unseen data set for fault detection in the microgrid. Further, the fault classification task is carried out by including the wavelet based features derived from sequence components along with the features derived from the current signals. The new data set is used to build the DT for fault detection and classification. Both the DTs are extensively tested on a large data set of 3860 samples and the test results indicate that the proposed relaying scheme can effectively protect the microgrid against faulty situations, including wide variations in operating conditions.