Technical gazette, Vol. 30 No. 6, 2023.
Original scientific paper
https://doi.org/10.17559/TV-20230712000799
Embedded Parallel Computing Platform for Real-Time Recognition of Power Quality Disturbance Based on Deep Network
Dewan Feng
; Department of Network and Information Security, Chongqing Vocational Institute of Safety & Technology, Chongqing, China
*
* Corresponding author.
Abstract
Systems powered by scattered sustainable power sources are highly susceptible to disturbances in the quality of power. Power Quality Disturbances (PQD) signals can degrade the functionality of grid-powered appliances. The older techniques for recognizing thePQD signals involve feature extraction. Manual analysis needs to set up a digital signal processor platform, which may lead to a time-complex process and errors in accuracy. Real-time PQD (RPQD) recognition techniques have advanced Embedded Parallel Computing Platform(EPCP), various signal processing methods, artificial intelligence, and Deep Network (DN) methodologies to recognize RPQD signals successfully in real-time scenarios using EPCP-RPQD-DN. Initially, the proposed algorithm implements hybridized Deep Belief Network and Long Short-Term Memory (DBN-LSTM) to accurately recognize the real-time PQD signals. Secondly, the DBN module maximizes the input signal features for the generation of PQD in a fixed period by training phase directly from raw PQD input signals and forwards it to the LSTM module. Third, in LSTM, the time series nature of PQD signals is easily analyzed using three layers, allowing it to run on the EPCP model. The PQD sample signals are employed to train the DBN in a central monitoring server. A series of PQD signals generated by the EPCP simulation environment is carried out to validate the effectiveness of the EPCP-RPQD-DN approach. Real-time simulation of electromagnetic fault conditions in the power system by Real Time Digital Simulator (RTDS) hardware. Experimental evaluation shows that DN learning improves accuracy rate, reduces computational overhead, and minimizes error rate compared to existing approaches.
Keywords
deep belief network; embedded parallel computing; long-short term memory; power quality disturbance
Hrčak ID:
309241
URI
Publication date:
25.10.2023.
Visits: 566 *