DSP Based Online Power Quality Events Detection and Classification Using Hilbert Huang Transform and Random Forest Method

Abstract

In this paper, empirical mode decomposition (EMD), Hilbert transform (HT), and random forest technique are combined to detect and classify the power quality events (PQEs) in real-time. Empirical mode decomposition (EMD) is used to decompose the non-stationary PQEs into the mono-component mode of oscillations, known as intrinsic mode functions (IMF). The most efficacious features are extracted by applying Hilbert transform on individual IMF. Four features such as standard deviation of magnitude, Hilbert energy, Tsallis entropy, and standard deviation of phase are computed from the AM-FM signal of Hilbert transform to train the random forest classifier. Random forest (RF) classifier, a congregated form of numerous self-determining decision trees that has a unique ability to deal well with uneven data sets with missing variables and faster learning speed compared with other machine learning methods. The main objective of this paper is to assess the worth of the proposed HHT-RF method implemented in the hardware prototype using a digital signal processor (DSP) in the power quality events detection and classification in real-time.