Abstract: Aiming at the issue of discharge pulse extraction for multi-source and massive PD signals, a novel parallel method based on Message Passing Interface was proposed. The proposed method applied a parallel mode called manager-worker-writer. In this method, a manager dynamically assigned task to several workers, and these workers executed tasks in parallel and a writer received results from workers in real time, so data management was separated from task execution. In addition, the manager identified sources of PD signals and sent them to workers as the keys for analyzing different data files and setting algorithm parameters, so multi-source and heterogeneous PD signals can be processed in parallel. Experimental results show that the proposed parallel method is efficient, feasible and possesses good speed-up ratio. It achieves total extraction accuracy of above 90%, which can satisfy most engineering applications.

Key words: partial discharge, message passing interface(MPI), parallel computing, big data, pulses extraction, data processing