Research On SOC State Estimation Of Lithium Battery Management System

Lithium batteries have occupied the market rapidly with the advantage of long cycle life,high energy density,no environmental pollution,cost-effective and other unique advantages.It has been applied widely in the energy storage system so that people put forward higher requirements to the accuracy of the state estimation of lithium battery and the reliability of the remote management system.In this paper,the Sate of Charge(SOC)estimation algorithm for large capacity lithium battery is studied,and then combined with the actual situation of the project,two kinds of SOC estimation methods are proposed,after that a distributed battery management system based on GPRS wireless network is provided.The main research work and results are as follows:In order to solve the problem of the loss of particle diversity in SOC estimation based on the particle filter algorithm which is resulted by the introduction of resampling technique,an improved particle filter algorithm for SOC estimation is proposed.The proposed algorithm reduces the sampling variance and improves the diversity of particles by introducing the error back propagation(BP)neural network into the particle filter algorithm to update the weights.The weight updating process increases the weight of the particles located at the tail of the probability distribution,so that it can enter the high weight area.The experimental results show that,compared with the particle filter algorithm,the SOC estimation value of the improved algorithm is closer to its actual value,and the proposed algorithm has a significant improvement both in root mean square error and maximum absolute error of the SOC estimation.Considering the complexity and accuracy of establishing the state space mathematical model for lithium battery,a SOC estimation algorithm for lithium batteries based on the parameter optimization support vector machine(SVM)is proposed in this paper.It takes advantage of the powerful self-learning ability of SVM to realize the modeling and state estimation of large capacity lithium batteries.And from the perspective of selecting the optimal parameters of SVM,the simple and fast particle swarm optimization algorithm is applied to realize the parameters' global optimization to find the best parameter combination and shorten the training time of support vector machine.The experimental results show that,compared with the SVM based on grid search parameter optimization,the SVM based on the particle swarm optimization algorithm can achieve the global optimization of the model parameters in a shorter time under the same conditions of the SOC estimation.The particle swarm optimization algorithm speeds up the model training process in SVM for the SOC estimation,the training time is reduced by 13.9%on average,and the overall performance is better than the SVM based on grid search.A design of the battery management system is further presented in this paper,based on the research results of SOC estimation.It is a distributed energy storage lithium battery management system based on GPRS wireless network,which includes two parts:battery collection and management server.The acquisition and transmission of the battery status information is completed by the remote lithium battery alone.And the front and rear sub system design ideas is selected in the battery management server.In the background system,the multi task parallel mode communication interface is used to receive,analyze and store the data of the distributed battery.And the foreground system realizes the users'access to the battery data through the man-machine interface.