Study On SOC Estimation Of Power Lithium-ion Battery Based On Support Vector Machine

At present, to achieve sustainable economic development, the traditional autoindustry is moving toward a new way of energy saving and emission reduction like thedevelopment of new energy electric vehicle (EV). As the representative of new energyvehicles, EV and their advantages are highlighted. Power lithium-ion battery, as themain power sources of EV, is also one of the main factors restricting its development,and has attracted more and more attention from the world. The battery SOC (State ofCharge) is a very important parameter in the process of running for EV. The batteryitself is a complex and closed electrochemical reaction system, and there are manyinfluence factors in SOC estimation. Therefore, the accurate estimation for SOC ismuch difficult. As a result, extracting a reasonable and effective SOC estimationalgorithm, which can improve the efficiency of the battery use in theory, increaseendurance mileage, prolong service life and reduce running cost, is of importantmeaning.This paper studies the subject of the rapid and accurate SOC estimation of powerlithium-ion battery.Ⅰ. The article first briefly introduces the background of selected topic, thedevelopment of EV at home and abroad and the current situation of power battery andits estimation method, and then, the structure of lithium-ion battery and illustrates indetail the process of charging and discharging, and explains the definition of SOC,analysis and main factors affecting the existing lithium-ion battery SOC estimationmethods and their advantages and disadvantages.Ⅱ. Then comparing the strengths and weaknesses of the traditional SOCestimation method, the Support Vector Machine (SVM) is put forward to estimate SOC,and SVM based on hybrid kernel function of algorithm is adopted; SOC estimationresults through simulation experiment, compared with a single nuclear SVM results,proves that the mixed kernel function has better generalization ability and generalizationability.Ⅲ. To find the optimal parameter sets of SVM algorithm, on the basis of thecomparison of several current common algorithm, QPSO in combination with quantumevolutionary algorithm and PSO(particle swarm optimization) algorithm of the quantumparticle swarm optimization (pso) algorithm is propose, for SVM parameters setoptimization work theory.Ⅳ. The results, by using joint algorithm of QPSO algorithm combine with SVMand after comparing it with a single kernel function SVM algorithm, show that it greatlyimproves the estimation accuracy of SOC, and also has a great improvement onconvergence speed. After comparing it with basic PSO algorithm combined with SVMmodel, the results show that average errors of QPSO algorithm greatly reduce.In addition, build lithium-ion battery measurement platform and complete the design of hardware and software of the voltage, the current, the temperaturemeasurement module, main control system and CAN communication module. In thevehicle simulation software environment, the results from the test of the hardware in theloop test show that the platform is practicable.In view of excellent generalization ability of the algorithm for QPSO-CSVM, theSOC estimation has higher precision and faster convergence speed. The algorithm willhave very good development and practical application value.