Research On State Of Charge Estimation Of Vehicle Lithium-ion Power Battery Based On BP Neural Network

Parameter monitoring of power battery is the key technology of new energy vehicles.The accuracy of battery-related parameter monitoring directly affects the overall performance of new energy vehicles.This paper takes the common 18650 power lithium battery as the main research object,A group search optimization algorithm based on Levi flight steps(Levy Group Search Optimizer-back propagation,LGSO-BP)neural network lithium battery charge state(State of Charge,SOC)Estimating method,Using the Levy group search optimization algorithm for the SOC neural network model,To improve the accuracy of SOC-related parameter estimation by neural networks.This paper first expounds the basic structure of the common lithium battery,analyzes the basic principle of the operation of the lithium battery in detail,focusing on the basic meaning and role of the lithium battery capacity,discharge platform,voltage,internal resistance and multiplier discharge performance.The discharge curve of lithium battery under various temperature conditions and discharge rate was studied.Secondly,according to the definition of lithium-ion power battery SOC,various factors affecting the estimation of lithium battery SOC are analyzed,and the advantages and disadvantages of various SOC estimation methods are analyzed in detail.Because BP neural network can show excellent functions both in comprehensively handling the contradiction problem of high-dimensional data information and in the development problem of multi-peak operation function optimization.For the high-dimensional operation function,its outstanding advantages,mainly show that simple operation,high convergence efficiency,and maintain strong accuracy,this paper determines the SOC estimation model based on BP neural network.Although BP neural network has outstanding advantages,its operation method has great difficulty and time loss.At the end of the calculation,the convergence rate decreased significantly,and even more stagnated,not conducive to the use.Therefore,this paper proposes the algorithm improvement method based on Levy flight step.The frequent multiple search can effectively improve the local search level.Expand the search space,can further improve the global search level,and avoid the local landing.The effectiveness of this algorithm is verified by comparison of the established group search algorithm(LGSO)performance based on Levi flight steps using 9 Benchmark functions.Finally,the operating current,operating voltage and temperature data of the measured input parameters of LGSO-BP neural network were used to estimate the SOC parameters.The value of model adaptation,convergence characteristics,S OC prediction curve and prediction error are analyzed.We conclude that LGSO algorithm has high optimization ability and high neural network prediction accuracy of lithium-ion battery on SOC estimation.Keeping the advantages of BP neural network itself,the optimization ability of relevant parameters is effectively improved,such as Levy flight steps to improve individual diversity,balance the balance between search speed,local optimization and global optimization,and make the estimation results more accurate.