A local over-thermal fault evaluation method for C5F10O insulated power equipment based on DWT and BP neural network optimized by GA

Due to the growing problem of global warming, C5F10O is promising to replace SF6 as an insulation medium in power equipment because of its low global warming potential and excellent insulation performance and thus has a wide application prospect in the electrical engineering field. Local over-thermal fault is one of the most severe faults in power equipment and has a close relationship with the characteristic decomposition components (CDCs). This paper is devoted to proposing an evaluation method for local over-thermal fault by analyzing CDCs. The discrete wavelet transformation method was adopted to recognize CDC (CF2, CF2CF2, COCF2, and CFCF3) from their variation curves, and the fault feature vector was extracted based on the analysis of frequency band energy. The back propaganda neural network optimized by the genetic algorithm was employed to evaluate the severity of local over-thermal fault with a high accuracy. This work can lay a theoretical basis for local over-thermal fault evaluation based on CDCs in environmentally friendly power equipment.