Research On The Remaining Useful Life And SOC Estimation Strategy Of Power Batteries For Vehicles

Lithium-ion batteries have been utilized in aerospace,electronic components and other fields on a large scale,especially in the new energy automobile industry.Due to the instability of its internal structure,the battery may appear safety accidents such as aging,overcharge and overdischarge,and thermal runaway during work and storage.Therefore,in order to prevent the continued use of the aging battery and th e overcharge and overdischarge of the battery,it is of great practical significance to accurately predict the remaining useful life(RUL)and state of charge(SOC)of the battery.In this paper,firstly starting from the empirical model of battery aging,the particle filter algorithm and its improved particle filter algorithm are used to predict the RUL parameters,and the pros and cons of the algorithm are compared from the perspective of accuracy,then an improved extended particle filter with a regulari zed sampling method is put forward.Secondly,based on two data-driven models,the autoregressive integrated moving average model(ARIMA)and the support vector regression model(SVR)extract the linear information and nonlinear information of the observat ion data to predict the future trends.Then an ARIMA-SVR-IPF fusion algorithm with a datadriven model to guide particle filter prediction was constructed.For the study of SOC parameters,this paper uses the extended Kalman algorithm to estimate the param eters based on Thevenin model.The principal research content of this paper is as follows:1.When predicting the remaining useful life of lithium-ion batteries,we generally use the empirical model of capacity attenuation and the particle filter algorithm.There are problems such as the reduction of particle diversity and the low accuracy of RUL prediction.For the purpose of solving above problems,firstly,three improved particle filter algorithms are analyzed and compared,namely,the EPF algorithm,the UPF algorithm and the RPF algorithm.In the comparison of the algorithms,it is found that the UPF algorithm has the highest prediction accuracy,but due to its double sampling of the algorithm leads to long processing time.The prediction accura cy of EPF algorithm is inferior to UPF,the RPF algorithm has low prediction accuracy but improves the diversity of particles.Based on the advantages and disadvantages of the above algorithms,this article combines regularized sampling and extended partic le filter algorithms,then proposes a new fusion particle filter algorithm(IPF),and predicts and analyzes the remaining useful life.2.When the training data is insufficient and the empirical model has errors,the IPF algorithm will have the defect of l arge prediction errors.For the purpose of solving above problems,the autoregressive integral moving average model is first ly used to extract linear information from the observation data,and then the support vector regression machine model is used to ext ract nonlinear information from the residual data.Finally,the integrated prediction results of the two data-driven models are used as IPF algorithm’s observation data,thus an ARIMA-SVR-IPF fusion algorithm with higher prediction accuracy is constructed and reduces empirical model dependence.3.The accurate monitoring of the state of charge can reduce the aging rate of the battery.There is interference noise in the actual measurement of the voltage.Therefore,this paper uses the discrete wavelet transform method to reduce the noise of the terminal voltage,then SOC is estimated based on the Thevenin model and the extended Kalman algorithm.According to the results,after the noise-added terminal voltage is input into the SOC estimation system,the estimated result of the extended Kalman algorithm and the ampere-hour integration method differs nearly 10%,while the estimated result of the noise-reducing terminal voltage is less than 3%.