Research On State Of Health Estimation Method For Lithium-ion Battery Of Electric Vehicle

Entering the new century, problems posed by the non-renewable energy shortage andenvironmental pollution turn increasingly prominent, development of electric vehicles isbecoming an important measure for countries around the world to accelerate thestrategic transformation of transport energy. Lithium-ion battery is the main powersource for electric vehicles, storage device for vehicle energy, of which performancestatus directly characterizes the majority of performance indexes of vehicle. However,estimation methods about battery state of health stay immature, which has been onebottleneck in the development of battery management system technology. In order tomeet the requirements of national legislation and to ensure pure electric vehicles workreliably, Lithium-ion battery state of health must be monitored.In this paper, battery neural network model is established to guarantee normalrunning of vehicle, based on26650Lithium iron phosphate battery and self-made BMS.Several aspects as follows are researched:(1)The internal structure and working principle of charging and discharging processof lithium iron phosphate battery, reasons for capacity fading during course are analyzedand effects that temperature, self-discharge rate, open circuit voltage and internalresistance of the battery has on the battery state of health are studied.(2) Four battery modeling methods are described in detail: electrochemical model,empirical model, the equivalent circuit model, artificial neural network model. Aftercomparison of three modeling methods, artificial neural network model is picked tomodel the batteries. BP neural network based on LM algorithm is used to simulatecomplex nonlinear relationship between the health status of the battery and parameters.(3) Estimation model of battery health state based on LMBP method is proposed, aneural network structure including the number of neurons in each layer and the transferfunction is designed and MATLAB simulation program is written. In order to get themodel sample data, charge-discharge cycles experiments of26650lithium ironphosphate is carried out, and battery status data is recorded in real time, parameter datainput into the model to learn. Further,30cells of batteries in preparation are divided intothree groups of different discharge cycles, the state parameters are input into the modelto obtain estimates of SOH value, compared with the actual value of SOH, theestimation error is less than5%. (4)A battery test platform is built, including hardware and software of the system, thehardware includes a master controller and peripheral modules, data acquisition modules,charge and discharge control module, communication interface module and powermodule; Software includes PLC program and PC software, and communication betweenthe PC software and MATLAB through a virtual port, and finally through the collecteddata lithium iron phosphate SOH is predicted with the relative error of less than5%, inline with the power battery SOH estimation accuracy.