Wavelet transform‐based feature extraction for detection and classification of disturbances in an islanded micro‐grid

Abstract

As one of the most classical digital signal processing (DSP) techniques, the wavelet transformation (WT) has been applied to detect and classify power system disturbances for many years. However, its performance is easily affected by harmonics. This makes it difficult to detect and classify disturbances occurring in an islanded micro-grid, especially when non-linear loads are involved. To improve the performance of WT, in this article, the Renyi entropy is used with WT to detect and classify seven types of disturbance. To demonstrate the efficacy of Renyi entropy in the wavelet domain, a comparison is made by detecting and classifying disturbances using the raw data with the time-domain-based Renyi entropy. Other comparisons are also performed to show how the performance of Renyi is affected by different order values of Renyi and the non-linear loading level. Detection and classification are made by using the support vector machine (SVM) and K-nearest neighbour (KNN) classifiers. The results demonstrate the effectiveness of WT-based Renyi entropy and show that the performance accuracy improves with the increase in the percentage of non-linear loads.