Real‐time recognition of power quality disturbance‐based deep belief network using embedded parallel computing platform

Abstract

This paper proposes a novel integrated solution for real‐time power quality (PQ) disturbance analysis. Traditionally, the recognitions are based on the feature extraction and are implemented offline or on the advanced reduced instruction set computer machine/digital signal processor platforms. In this paper, the optimized deep belief network (DBN) analyzes the PQ signals by learning knowledge from raw data of signals directly, which maximizes the features of original signals, and runs on an embedded parallel computing platform (EPCP). In simulation studies, eight types of common PQ disturbances are divided into 17 classes of data frame models according to the PQ disturbances that may occur during a fixed period of time. These samples of data frame models are utilized to train and optimize the DBNs on a central server. Compared with the existing classifiers, the simulation results demonstrate that the proposed approach has higher accuracy and stronger robustness. Then, the optimized DBN is sent to EPCP, and the previous DBN model is updated. The EPCPs are used to recognize the signals of PQ disturbances from a real time digital system (RTDS). The proposed integrated solution has excellent performance regarding accuracy but is time consuming. © 2020 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.