Vehicle power system condition prediction based on weighted hidden Markov model

Abstract

A new condition prediction approach based on weighed Hidden Markov Model(HMM) was presented in order to solve the problem of trend prediction for vehicle power system.Through the establishment of the power system HMM,the hidden state of HMM was predicted by weighted prediction method,and the observed state of the model was calculated by the observation probability density of the maximum probability hidden state.The approach was applied to the state prediction of rate turn of the system voltage adjusting pulse width signals,and was compared to that of using BP(Back Propagation) neural network and Auto-Regression(AR) prediction model with the same sequence.The results show that the method has better prediction on the system state change.