| With the development of the economy and the improvement of people’s living standards,the problems of environmental pollution and shortage of energy exhaustion have seriously threatens people’s lives,and the emission of automobile gas has a particularly serious impact on people’s lives.New energy vehicles are increasingly favored by people because of their low energy consumption,pollution-free,low noise and low cost.Battery Management System(BMS)is a key component of electric vehicles.Battery peak power estimation as an important estimator for battery management systems.It provides a reference for the power distribution and energy control of the electric vehicle system 。 It can measure the acceleration climbing performance of the vehicle and the braking energy recovery function during braking,and can effectively prevent the battery pack from overcharging and over-discharging,and improve the battery` life.In this paper,the ternary lithium battery used in pure electric vehicles is taken as the research object,and the Hybrid Pulse Power Characteristic(HPPC)is used as the test method.The specific research is as follows.(1)By collecting the literature related to battery power characteristics,and understanding the current research status of battery power characteristics.It is concluded that the current peak power research on batteries mostly focuses on the influence of a single factor on the peak power of the battery or on the battery model.So this paper proposes a neural network battery model based on data analysis.(2)Analyze the common test methods for battery peak power research.If the peak power is used directly as the test output value,the difference in ohmic internal resistance may be caused by different battery manufacturing processes,so test battery has higher estimation accuracy.The estimated accuracy of the battery that is not tested is lower,and the HPPC test obtains the peak current value,which is substituted into the corresponding formula to obtain the peak power,which avoids this situation happening.(3)The Matlab toolbox is used to establish the battery neural network model,and the error back propagation characteristics of the BP algorithm are used to train the model.The maximum relative error of the estimated battery peak power is greater than 5%.Therefore,consider using simulated annealing algorithm(SA)or Particle Swarm Optimization(PSO)improves the BP algorithm.(4)The SA-BP algorithm and the PSO-BP algorithm were written in Matlab scripting language as the training method of the model.The accuracy of the test samples based on BP,SA-BP and PSO-BP algorithms was compared.It is proved that the improved BP algorithm can enhance the estimation accuracy of the model,the trained model-based on SA-BP algorithm has the highest accuracy. |
