Research On Power Battery SOC Estimation Based On Fusion Calculation Considering Temperature Influence

Power lithium-ion battery is one of the key components of electric vehicles,and plays a vital role in many aspects of vehicle performance.The ternary lithium-ion battery occupies a very large proportion in the practical application of power batteries by virtue of its excellent characteristics.The accuracy of the estimation accuracy of the remaining power(State of Charge,SOC)of the battery at the current moment directly determines the performance of the battery,and can reduce the occurrence of battery overcharge and overdischarge,which can maximize the battery usage process.Cycles.However,in practical applications,the state of charge of the power battery cannot be directly obtained,and the estimation error fluctuates greatly at some temperatures.In view of these issues,this article will focus on the following aspects:Firstly,a battery characteristics is obtained by conducting a charge and discharge experiment and analysising the result of the experiment on the NCR18650 battery under different influencing factors.The study found that temperature has a certain influence on battery capacity,which will further affect the accuracy of battery SOC estimation.A temperature-dependent ampere-hour integration method is proposed.Aiming at the shortcomings of different algorithms in SOC estimation,an improved algorithm combining BP neural network and square root no-track Kalman filter is proposed to study the SOC estimation algorithm of ternary lithium-ion batteries considering the influence of temperature.Secondly,in order to identify the parameters of the battery model,the battery model is studied,and the experimental data is collected.Then the identification algorithm is studied,and then the parameters of the battery model are identified based on experiments.Finally,the battery model is established and the output results are analyzed and studied.The results provide a foundation for the hardware in the loop test.Thirdly,the structure principle of neural network is analyzed,and the neural network algorithm improved by genetic algorithm is designed.To obtain algorithm training data,current voltage and temperature data under different temperatures and Highway Fuel Economy Driving Schedule,LA92 Dynamometer Driving Schedule,and Supplemental Federal Test Procedure Driving Schedule data is collected.The BP neural network improved by genetic algorithm is used to obtain the weight and threshold that can make the output result in the closest state to the true value.In order to improve the operation speed and accuracy,the adaptive learning rate is introduced to improve the network structure.After the model is trained,Urban Dynamometer Driving Schedule data is used to test and verify the trained improved neural network.The model output is excellent,but there is a certain deviation at some sampling points.Using the square root no-trajectory Kalman filter to optimize the output of the design algorithm,the effect is significantly improved,and the estimation result is more accurate.Finally,based on the previous research of the research group,a set of hardware-in-the-loop test platform based on NI PXIe is designed for the algorithm designed in this paper.The battery pack model built by Simulink is imported into the NI Veristand simulation environment.The algorithm designed in the previous article is burned into the BMS controller,and the hardware-in-the-loop simulation platform is built to realize the hardware-in-the-loop verification.The results can meet the expected requirements and verify the feasibility of the algorithm designed in this paper at different temperatures.