Intelligent fault diagnosis of air conditioning system in electric bus

Abstract

Air conditioning (AC) system becomes the second largest energy consumption part of electric bus, which will greatly affect recharge mileage and driving performance. It is very important to set up effective and reliable diagnostic model for AC system in electric bus. In this paper, various types of feature parameters are extracted and analyzed for related soft-fault firstly; and then new fault prediction model of BP neural network is established with improved particle swarm algorithm. There are three steps of prediction data modeling to determine the type and trend of AC system failure soft-fault quickly: characteristic parameters extraction; data preprocessing; the best particles calculation. However, the neural network model of the soft-fault diagnosis in high-dimensional complex problems is easy to fall into the local extreme point, named as “dimension disaster”. Thus, the improved support vector machine (SVM) model is proposed to deal with the high-dimensional problems. Penalty parameter C and Gaussian kernel function parameters of SVM, which are optimized by improved particle swarm algorithm in this paper, will influence the predicted results greatly. From soft-fault diagnosis results of different prediction models, we can see that the fault diagnosis model of optimized SVM based on improved particle swarm algorithm spends less training time and has higher predictive accuracy with optimized SVM parameters through the improved particle swarm algorithm.