Research On SOC Estimation Method For Lithium-ion Battery Based On Attention Mechanism Enhancement Algorithm

In recent years,the energy crisis is getting worse,and the exhaust emissions of traditional cars have caused serious air pollution,so electric vehicles have received extraordinary attention.At present,most electric vehicles use lithium-ion batteries as the power source.Accurate estimation of SOC not only provides the current and remaining energy performance of the battery,but also relates to the driving safety of electric vehicles.However,the SOC cannot be obtained directly through measurement,can only be estimated through other measurable battery characteristic data.Nowadays,some methods for SOC estimation of lithium-ion batteries based on recurrent neural network algorithm have achieved good estimation accuracy,but there is still a problem of insufficient learning of battery data,which leads to slow convergence rate of the algorithm and low estimation accuracy.In order to solve the above problems,this paper proposes a lithiumion battery SOC estimation method based on the attention mechanism enhanced recurrent neural network algorithm,which can further improve the accuracy of the SOC estimation result.The main work of this paper is as follows:Firstly,the data of lithium-ion battery is collected and processed,and the data test system of lithium-ion battery is built.Through theoretical and experimental studies,the battery data structure that meets the requirements of the SOC estimation method is designed.Then,the battery test experiments are designed under three standard driving conditions: Dynamic Stress Test(DST),Federal Urban Driving Schedule(FUDS)and Supplemental Federal Test Procedure Driving Schedule-US06(US06)to collect the required battery data,and the obtained battery data is preprocessed.A battery database is designed to store all the battery data,which provides data for the subsequent battery SOC estimation experiments.Secondly,the SOC estimation model of lithium-ion battery based on the attention mechanism enhanced recurrent neural network algorithm is established.After the training and optimization of the model,the optimal parameters of the model are determined.The mean absolute error range of SOC estimation results for self-measured data and public data is between [0.20%,0.42%] and [0.32%,0.76%] respectively,which proves that the model can accurately track the SOC state of the battery under different driving conditions and different temperatures.By comparing the model estimation results before and after the enhancement of the attention mechanism,it is verified that the attention mechanism can indeed enhance the performance of the recurrent neural network,which can reduce the mean absolute error of the SOC estimation results of the model from 0.61% to 0.40%,and accelerate the convergence rate of the model.By visualizing the model’s attention to different battery characteristic data,the model’s attention ability is verified.Finally,through the design of hardware and software,including the main controller module,data acquisition module and analog-to-digital conversion module,the online SOC estimation system of lithium-ion battery is built.On this system,the proposed SOC estimation model of lithium-ion battery based on the attention mechanism enhanced recurrent neural network algorithm is realized.Then,the online SOC estimation experiments of lithium-ion battery under three driving conditions are carried out on the system.The results show that the system can accurately estimate the battery SOC in seconds,which verifies the feasibility of the proposed method.