Balance Evaluation Of Lithium Battery Pack Based On Data Driven

With the rapid development of China’s economy,the problems of environment and resources have become increasingly prominent.As the most potential "green energy" in the 21st century,lithium battery can solve the problems of environmental pollution and resource shortage to a certain extent.In addition,lithium battery has the advantages of small volume and large energy,so it is widely used in various fields.Generally speaking,the energy of a single lithium battery is limited and can not meet the conditions required for equipment operation.Usually,multiple battery cells are connected in series to form a battery pack for power supply.In fact,battery grouping is a complex process.Due to the limitation of technical level,there are inconsistencies in voltage and temperature of each battery unit,which will reduce the performance and service life of the whole battery pack,and even lead to safety problems.A combination of qualitative and quantitative analysis is adopted in this paper,and the balance analysis and evaluation of lithium battery is handled using data mining.The work of the thesis is listed as below:First,the voltage standard deviation of each cell is used as an index to measure the balance of the battery pack after comparing and analyzing the monitoring complexity,balance fit and real-time measurement.Secondly,the balance of the battery pack is analyzed by using the boxplot,and the unbalanced interval of the lithium battery is obtained by using the inner limit of the boxplot as the threshold.Furthermore,a dynamic evaluation model based on machine learning and deep learning is designed to obtain more scientific and reliable imbalance information to make up for the insufficiency of the static model.In machine learning,the SPOT and ARIMA-SVR combined models are adopted.The SPOT model continuously updates the threshold to find out the equilibrium abnormal points,and the ARIMA-SVR model compares the actual value with the fitted value to judge the inconsistency of the battery pack.In the aspect of deep learning,the LSTM model is selected,and the residual test is carried out using the 3sigma principle to determine the equilibrium level of the battery,on the basis of the overall fitting degree of 90%.Finally,using the historical data provided by a lithium battery company for analysis,it is found that the above algorithms are feasible and reasonable,and can provide an effective solution for the balance evaluation of lithium batteries.