Parameter Identification And SOC Estimation Of Battery Based On Intelligent Optimization Algorithm

The state of charge(SOC)is an important parameter in battery,which characterizes how much capacity the battery still has remained.Establishing an accurate battery equivalent model and identifying precise model parameters are of great significance to the accurate estimation of the state of charge.In this paper,the battery is taken as the research object,and the battery SOC is estimated by the MS-AUKF(master slave adaptive unscented kalman filter)algorithm,the improved robust extreme learning machine algorithm and the intelligent optimization algorithm.The specific method is as follows.For the problems that in traditional extended kalman filter and unscented kalman filter method,the noise variance is fixed so the error precision is not high,a kind of SOC estimation method based on neural networks and master slave adaptive unscented kalman filter(MS-AUKF)algorithm is proposed.Firstly,a second-order Thevenin model of the battery is built,for the nonlinear relationship between open circuit voltage and the SOC of battery,neural network is used to improve the fitting accuracy,instead of polynomial model.Several key parameters of the battery model which affect the SOC of the battery are determined combining with experimental data.MS-AUKF algorithm is adopted to estimate the SOC of the battery.The main filter is used to estimate the state of the system,and the slave filter is used to estimate the variance matrix of the noise.At each iteration of the algorithm the noise variance of the system model is updated,which overcomes the shortcomings of divergence caused by initial value of noise variance generally set by experience in traditional kalman filter.In the previous chapter,a SOC estimation method based on neural network and MS-AUKF algorithm is proposed.A third-order Thevenin equivalent circuit model of the battery is established to improve the accuracy of the equivalent circuit model.An improved BSO(beetle swarm optimization)method is proposed to identify battery parameters in this paper in order to effectively improve the performance of the BSO algorithm.The chaotic initialization increases the probability of finding the optimal value of the BSO algorithm by making the initial values spread throughout the solution space.Gaussian perturbation is beneficial to the BSO algorithm to jump out of the local optimalsolution by leading the next speed of the particle with the average of the sum of the optimal values of all particles which contains a Gaussian perturbation factor.The test result which aims at the optimization performance of the algorithm with five test functions show that: the improved algorithm has faster convergence speed and higher estimation accuracy compared with the original BSO algorithm.Identified the battery parameters using the improved algorithms.The results show that the improved algorithm has good convergence and high prediction accuracy.Based on the first two chapters,a method based on bird swarm algorithm optimizing robust extreme learning machine is proposed to estimate the charge state of the battery.Robust extreme learning machine overcomes the shortcomings that extreme learning machine can not deal with the abnormal value,so the prediction accuracy of the network is improved.The parameters such as the number of hidden node and the adjustment factor of robust extreme learning machine are optimized by bird swarm algorithm,so the problems that the parameters such as the number of hidden nodes and the adjustment factors are difficult to be determined can be solved,which can further improve the convergence speed of the network and help to find the global optimal value.Several key parameters including current,voltage,temperature and internal resistance,which affect the SOC characteristics of the battery,are collected to model and test by ADVISOR software.Simulation results show that compared with other algorithms such as BPNN,RBFNN,ELM and FNN,BSA-RELM has a smaller error and higher prediction accuracy.