Improved differential evolution algorithm of model‐based diagnosis in traction substation fault diagnosis of high‐speed railway

Abstract

The traction substation plays an important role in high-speed railway (HSR) as it can provide electric energy for trains, whose fault may threaten the safe and stable operation of HSR. Compared with common diagnosis methods, such as expert systems that heavily depend on professional experience, model-based diagnosis (MBD) has some distinct advantages. However, the inefficiency and incompleteness of calculating minimal hitting sets (MHSs) limit the performance of MBD. To reduce these limitations, the binary differential evolution with secondary population algorithm is proposed to calculate the MHSs. This algorithm can take advantage of differential evolution algorithm to improve the computational efficiency. The secondary population is used to enhance the convergence rate. In addition, the MHSs ensured strategy is proposed to improve the computational accuracy. Experiments are carried out on an actual traction substation in Hefei–Nanning HSR, and the results show that the MHSs can be solved accurately to finish the fault diagnosis of the traction substation in a short time.