Research On SOH Of Low Voltage Li-ion Battery Pack

Because current sources of fuel,such as oil,are harmful to the environment and their reserves are declining,the electrification of some automobiles is very necessary.In order to make electric vehicles run efficiently and safely,high-efficiency,high-environment-friendly,the portable power sources are required.The main power sources include batteries,super capacitors and fuel cells.With the increasing development of electric vehicles,batteries gradually play an important role in the process of motorized vehicles.In the actual work process,the battery performance as an energy source depends on some internal and external factors,such as life,temperature,charge and discharge cycle and its chemical composition.Among the different types of batteries,lithium-ion battery is used most widely.Lithium-ion batteries have the characteristics of fast charging,high power density and high energy efficiency,wide operating temperature range,low self-discharge rate,light weight and small size.Therefore,lithium-ion battery is very suitable for using in electric vehicles.In electric vehicles,it is important to accurately estimate the battery SOH.SOH can be used to predict battery health,characterize the measurement of battery degradation,provide a basis for BMS safety protection,and ensure the normal operation of electric vehicles.SOH cannot be directly measured.Inaccurate estimates will harm the use of batteries and increase the risk of electric vehicles.Therefore,accurate estimates are indispensable.The SOH estimation can be implemented by various online and offline methods.In this thesis,online estimation of the battery SOH can be achieved through an offline data training prediction model.In this thesis,firstly,four lithium-ion battery cells with different life stages are randomly selected and tested for cycle life through standard charge-discharge cycling conditions until the full-life data of the lithium-ion battery are obtained.Based on the acquired test data,data such as voltage,current,temperature,and charge/discharge capacity in the charge-discharge cycle are extracted.In the actual use of lithium-ion batteries,the charging cycle generally uses a fixed working condition,and the discharge cycle changes according to the actual power demand.Therefore,this article uses the charging cycle data as a support to extract on-line key eigenvalues to improve the practicality of the prediction model.After obtaining the online characteristics,the current capacity of the battery is predicted and analyzed by linear regression model,BP neural network model and SVR prediction model respectively.The result shows that the SVR prediction model has better prediction accuracy and robustness,and the prediction error is within 2.5%.In order to predict the health of the battery pack,this thesis calculates the battery pack capacity.According most electric vehicles' battery pack adopt parallel serial coupling manner.This thesis builds the equivalent model of a single lithium ion battery,and verifies its effectiveness.Then,this thesis builds a parallel battery module.Finally,this thesis adopt parallel serial coupling manner to build low-voltage lithium ions pack.According to the SVR prediction framework,the capacity of the parallel battery module is predicted first,and then the battery capacity is predicted.Finally,the health status of the battery pack is predicted,and the error of the prediction result is kept within 3%.