Study On SOC Estimation Of Power Battery Based On IPSO-RLS Optimized RBF Network

With the lack of energy and the strict control of environment pollution by the whole society, electric vehicles and hybrid vehicles have become an important development direction in today’s automotive industry, and as a power source, lithium-ion batteries have the advantage of its favored. The performance and cycle life of the battery restricts the development of the electric vehicles and hybrid vehicles. Battery is a closed and complex electrochemical reaction system, and the battery state in the process is an important data for maintaining the efficient use and safety of the battery, and ensuring the stable operation of the vehicles. Many factors may affect the SOC estimation accuracy and the estimation algorithm has more prominent influence on predicting SOC.This paper expanded research on the purpose of improving the accuracy and the real-time of the SOC estimation algorithm. And propose an algorithm to predict the battery SOC. The detail researches are as follows:Firstly, selected the 4Li Fe PO battery as the research carrier for this paper, and introduced the battery structure, operating principle and characteristics, the SOC estimation effecting factors and commonly used estimation algorithm.Secondly, combined with the existing algorithms and intelligent algorithms to analysis and determined the prediction algorithm for the battery SOC estimation. Selected radical basis function(RBF) neural network as the main subject for the battery SOC estimation, after in-depth studying on the RBFN algorithm, found that the chosen of its network structure and parameters has a great influence on the computing ability of the RBFN algorithm. The selection of the appropriate parameters and structure is an effective means to improve the accuracy of the estimation algorithm.Thirdly, proposed to mix the improve PSO algorithm with the regularized least squares method to optimize the parameters and structure of the RBFN and improve the generalization ability of the network. The mixed optimization algorithm optimizes the regularization term of the regularized least squares method, the basis function of the RBFN, the node center and width, and other parameters, improves the SOC estimation accuracy and real-time of the RBFN algorithm.Finally, combined the optimization algorithm with the RBFN and formed the new SOC estimation algorithm. In order to test the accuracy and real-time of the algorithm in battery SOC estimation, compared the predicted results of the new algorithm with other estimation algorithms, and the results show that the mixed optimization algorithm has advantage in predicting battery SOC. Designed the system hardware and software for this new algorithm, and testing the function of the system.Given the RBFN mixed with the improved PSO algorithm and the regularized least squares method has advantage in the terms of estimating the battery SOC, and has the practical significance to meet the requirements of the battery SOC estimation precision and fast.