Adaptive prediction of remaining useful life for lithium battery based on high-order extended Kalman filter
Remaining Useful Life(RUL)prediction of lithium batteries is crucial for preventive maintenance of battery health management systems,and degradation models based on stochastic processes play a key role in this field,especially the use of two-implied-state nonlinear Wiener process,which can describe the degradation trend of lithium batteries more flexibly and specifically.In the lithium batteries RUL adaptive prediction,the degradation model parameters are often updated online using the Extended Kalman Filter(EKF)method,which is a nonlinear Gaussian filter with an approximate accuracy of only one order,and the filtering accuracy is low for strongly nonlinear systems.This paper is based on the nonlinear Wiener degradation model of lithium batteries with double hidden states.It designs a High-order Extended Kalman Filter(HEKF)to reduce truncation error by utilizing information from high-order terms and obtains optimal online parameter estimates.It further derives the probability density function of RUL and achieves adaptive prediction of RUL for lithium batteries.Finally,the example validation by NASA's lithium battery degradation data shows that the accuracy of this paper's method is improved by 83.9%and 53.3%,respectively,compared with the prediction results of the other two methods.
lithium batteryremaining useful lifehigher-order extended Kalman filternonlinear wiener processdouble hidden states
万方数据
维普
