The Analysis Of Hybrid Electric Vehicle Battery Management System

Due to the physicochemical constraints of storage battery and the requirements of actual conditions, the battery management system (BMS) is necessary to implement the function of monitoring, management, alarm and protection in hybrid electric vehicle. In the functions battery management system needs to realize, charge of state (SOC) estimation is a very crucial problem. Accurate estimation of SOC is beneficial to use battery reasonably, avoid battery overcharge and over discharge and prolong the service life of power battery.For a long time, although researchers pay a lot of hard effort, SOC estimation accuracy is not ideal, which mainly because the highly nonlinear between battery SOC internal influence factors.All kinds of method of estimation have some shortcomings:load voltage method is not suitable for the current fluctuation use conditions; estimation error of ampere-hour method would have been accumulated; neural network depends on the choice of the appropriate sample; The fuzzy control method depends on the rich project experience; Kalman filtering depends on accurate cell model; Electrochemical impedance spectroscopy need additional electrochemical impedance equipment, increased the estimated complexity, etc. At present, combined algorithm is generally to be used in SOC estimation; therefore studying of real-time accurate SOC estimation algorithm is very necessary.In this paper, the hardware and software are designed to implement the function of monitoring, management of SOC. BMS controller is high integrated performance PIC18single chip microcomputer. Current acquisition circuit, voltage acquisition circuit, the second order filtering and sampling keeping circuit, temperature acquisition circuit, balance circuit, CAN bus communication circuit are designed. Battery temperature acquisition subroutine, battery voltage, current acquisition subroutine, balanced electrical pathway program, SOC estimation procedure, battery status judgment and battery protection procedure and CAN send and receive subroutine is designed to realize the voltage, temperature, electric current information collection, the SOC estimation, battery equalization and protection function. The algorithm of SOC is the core of BMS.As to the SOC estimation, this paper analysis the affection of SOC on equal circuit parameters and find the monotone relations between SOC and resistance. The method of SOC estimation by BP neural network is discussed. According to the battery working voltage, current, temperature, internal resistance and the former time SOC, the SOC real-time numerical value can be estimated. In order to improve the estimation accuracy, neural network weights and threshold value are optimized through the genetic algorithm. Results show that the optimized neural network prediction error significantly reduced.