| With the increasing demand for electric vehicles,lithium-ion batteries are an important power source for electric vehicles,and their battery management system（BMS）has become a research hotspot at home and abroad.State of health（SOH）estimation and remaining useful life（RUL）prediction in BMS can provide a basis for battery detection and diagnosis to prolong battery life.In order to obtain accurate results of battery SOH estimation and RUL prediction,the following work is done in this paper.First,the development process of electric vehicles and lithium-ion batteries is introduced,and the domestic and foreign research status of battery SOH estimation and RUL prediction is expounded.Second,a charge-discharge cycle experiment is carried out on the lithium manganate battery.The relationship between battery capacity and the number of charge-discharge cycles is recorded and the aging characteristics of the lithium manganate batteries are analyzed.And data of Lithium Cobalt Oxide batteries at the center for advanced life cycle engineering（CALCE）at the University of Maryland is recorded and the aging characteristics of these batteries are analyzed.In the above two battery data,the health indicator（HI）for SOH estimation is extracted.Factors such as charge and discharge time,charge and discharge current,and battery capacity will all affect the battery performance.The battery capacity is usually difficult to measure directly,and the battery discharge current is determined by the specific working conditions,which is random.Therefore,this paper extracts the easily measurable charging time and charging current curve area from the charging stage as HIs.Then,considering the linear relationship between battery capacity and SOH,the correlations between HIs and battery capacity are analyzed by Spearman correlation coefficient and Pearson correlation coefficient.Third,an SOH estimation method based on whale optimization algorithm-relevance vector machine（WOA-RVM）is proposed in this paper.On the basis of traditional RVM,a multi-kernel function model is established,which combines the advantages of different kernel functions to improve the estimation accuracy.At the same time,HIs are fused to avoid the coupling relationship between different HIs affecting the estimation accuracy.The intelligent optimization algorithm can dynamically optimize the weights to improve the estimation accuracy.Therefore,the WOA algorithm in the intelligent optimization algorithm is used to optimize the weights of the kernel functions and the coefficients of different HIs.The simulation results show that the method proposed in this paper improves the SOH estimation accuracy.Fourth,in order to predict the battery RUL,a battery aging model is established.Considering the self-recovery phenomenon of the capacity,the capacity curve has the characteristics of local fluctuations,and it is difficult to accurately predict the RUL using a single method.The empirical mode decomposition method is used to divide the capacity decay data into the overall trend part and the detail fluctuation parts.The particle filter（PF）-based approach is used to predict the overall trend part.In order to alleviate the phenomenon of particle degradation,an intelligent optimization algorithm,ant lion optimization（ALO）,is embedded in the PF.With the help of the SOH estimation in the previous chapter,the particles are moved to the maximum likelihood region and the prediction ability is improved.After that,the autoregressive integrated moving average model（ARIMA）is used to predict the detail fluctuation parts.The above two prediction curves are added to realize multi-level RUL prediction.The simulation results prove that the fusion algorithm proposed in this paper combines the advantages of PF and ARIMA,which can not only realize the prediction of RUL accurately,but also reasonably consider the influence of local fluctuation of capacity in the process of prediction. |
