Fault diagnosis of High Voltage Circuit Breaker using Random Forest and PSo-KELM

Abstract

High voltage circuit breaker (HVCB) is an essential and important device of power grid, while any faults occurred in operation mechanism of HVCB will greatly jeopardize reliability and safety of power system. In order to improve fault diagnosis performance, a new fault diagnosis method using random forest (RF) and kernel extreme learning machine (KELM) is presented in this paper. At first, RF is applied to select the critical features derived from coil current. Then, the selected features are applied as inputs of KEML to establish fault diagnosis model. Next, particle swarm optimization (PSO) is introduced to turn crucial parameters of KELM to enhance diagnosis accuracy. Finally, practical samples are used to assess fault diagnosis performance of the presented method. Experimental results indicates the proposed RF-KELM method is capable of providing higher diagnosis accuracy than conventional approaches, which indicates promising future.