Research On Model-based Lithium-ion Battery SOC Estimation

With the rapid development of the automobile industry, the shortage of oilresources and environmental pollution problems have become increasingly prominent.Electric vehicles have some advantages of zero emissions and low noise, which willbecome the only way for the development of automobile industry. However, electricvehicles are still some problems, such as short driving range, the initial high cost andpoor security. Traction battery system technology has become a major bottleneck inthe development and industrialization of electric vehicles.Firstly, Four traction batteries are listed in the paper. By way of comparison, thelithium-ion battery is an ideal traction battery and has good prospects. The tractionbattery is selected, the battery management technology is needed to achieve theenergy of the electric vehicle can be used efficiently and safely. However, an accurateestimate of the battery SOC is the premise and key of the battery management systemrunning well. This article is based on the battery model and parameter identification,dedicated to the study of the lithium-ion battery SOC estimation algorithm.Secondly, The accurate estimation of SOC needs good battery model. In order toestablish a battery model for engineering, we introduce electrochemical impedancespectroscopy techniques for modeling study of the lithium-ion battery. Lithium-ionbattery electrochemical impedance is measured under potentiostatic mode, differentfrequency impedance data is plotted as electrochemical impedance spectroscopy.Analysis software reuse fitting impedance spectroscopy to obtain a lithium-ion batteryimpedance model. Through theoretical analysis, impedance model need to besimplified, and form the final second-order RC equivalent circuit model.After the establishment of the second-order RC model, we need to identify theparameters of the model. Two ways: off-line parameter identification method andon-line parameter identification method. The off-line parameter identification obtainvoltage hysteresis curve of pulse intermittent by pulse charging and dischargingexperiments, and then use Origin to fit the hysteresis voltage curve to obtain themodel parameters. The model parameters are affected by temperature, the current rateand SOC, so design pulse charging and discharging experiments under differenttemperatures, different current rate and different SOC. By comparison, we can get themodel parameters with the variation of several influencing factors, and create modelparameters offline database preliminarily. Online parameter identification collected current and voltage data by DST dynamic conditions experiments, with forgettingfactor recursive least squares algorithm to identify model parameters. The recognitionresult is verified by the comparison of the terminal voltage. The error analysis isshown that recognition result is good, and the second-order RC model is reasonable.Online identification of parameters can be obtained the corresponding SOC through alook-up table (offline database), and we can estimate the battery SOC by this method.Further, this method can estimate open-circuit voltage of the battery, combined withthe open-circuit voltage method, the SOC of the battery can be roughly estimated.Finally, based on second-order RC model and online parameter identification, theextended Kalman filter algorithm is used to estimate SOC of the battery. As a statevariable, SOC is written state space equation of the battery system. The algorithmcombines the Ah method and the open-circuit voltage method, the open-circuitvoltage method overcome the shortcomings of the Ah method error accumulation, theclosed-loop estimate SOC. Comparison with the actual SOC can be drawn: EKFalgorithm has a good estimation, which can be used as an effective method of onlineestimate SOC under dynamic conditions.