A rough set-based bio-inspired fault diagnosis method for electrical substations

Abstract

Imprecision and uncertainty in the alarm messages may significantly affect the accuracy and reliability of substation fault diagnosis results. To deal with that, a new rough set-based bio-inspired fault diagnosis method (RSBFDM) is proposed in this paper. It consists of four key components, namely the substation sub-region division method, the rough set attribute reduction algorithm, the binary reasoning spiking neural P system (BRSNPS), and the parallel reasoning algorithm. Specifically, the substation sub-region division method is used together with the rough set reduction algorithm to find the reduced fault production rule set for each sub-region. This simplifies the complexity of the problem and allows us to deal with fault alarm information uncertainty. Then, the BRSNPS and its reasoning algorithm are proposed to fulfill the fault knowledge representation and reasoning, yielding accurate fault diagnosis results. Thanks to the collaboration of rough sets and spiking neural P systems, no historical statistics and expertise are required and the scale of the problem is reduced. Experimental results carried out on realistic 110 kV and 750 kV substations show that the proposed method outperforms other alternatives.