Research On Health State Estimation Of Lithium-ion Battery Based On LSTM Network

Energy storage is an important part of realizing the national " double carbon" strategic goal.Among many energy storage schemes,lithium-ion battery energy storage has the highest efficiency,but the high cost and insecurity of lithium-ion battery affect its development.Accurately grasping the State of Health(SOH)of energy storage lithium battery can improve its utilization rate,improve the overall life of the battery pack and the working performance of the energy storage system.At present,the research on the health status of lithium-ion batteries needs to be improved in practicality.There are many studies on single batteries,and relatively few studies on battery packs.In addition,most of the lithium battery data used in the research are obtained in the laboratory environment,and there is a lack of research on actual data.Based on this,this paper mainly studies the SOH of lithium battery data sets and actual data obtained by experiments.Firstly,the evaluation indexes and related parameters of the health status of lithium-ion batteries are introduced.The different connection modes of lithium-ion batteries are listed and the reasons for their capacity attenuation are analyzed.Secondly,the NASA PCoE public data set,the CALCE public data set and the actual energy storage data from the field used in the research are introduced,and the above data are analyzed from the charging and discharging mode.Then,a variety of network models are built in Pycharm software,and the features of different dimensions are used as model input respectively.The Long Short Term Memory(LSTM)network is compared with different time series networks,and the relevant parameters of LSTM network are optimized based on genetic algorithm,and better prediction results are obtained.Finally,the data acquisition system of lithium-ion battery pack based on STM32 single chip microcomputer is designed.By comparing the advantages and disadvantages of different acquisition schemes,Analog Front End(AFE)is selected to collect various data.The designed acquisition system uses a master-slave structure.The real-time operating system is transplanted on the I.MX6ULL single-chip microcomputer to improve the real-time performance of the system.The STM32 single-chip microcomputer is used as the slave for data interaction.The collected data is displayed through the form application and sent to the cloud platform.After the cloud platform receives the data,the health status of each single cell in the battery pack is calculated through the trained model,and the prediction accuracy meets the design requirements.