Study On Capacity Degradation And Life Prediction Of Lithium-ion Battery Under Thermal Gradient Condition

Lithium-ion battery has become the main power source of new energy vehicles due to its advantages of good safety performance,high specific energy,high voltage platform,low self-discharge rate and good cycle performance.When batteries are assembled,the temperatures of individual batteries in a battery pack are uneven due to uneven packaging,box design,and heat generation and heat dissipation.As a result,thermal gradients are inevitably generated.Although temperature is closely related to battery aging,there is academic debate about the effect of thermal gradient on aging rate.The study of battery capacity attenuation and life prediction under thermal gradient conditions is of great significance to the study of battery thermal effect and life.Firstly,in order to ensure the consistency of the performance of the batteries used,the consistency screening method was used to screen the batteries used in the experiment.In order to solve the problem that the traditional fuzzy C-means(FCM)algorithm is difficult to converge and cannot realize the function of "selecting more and choosing less",an improved FCM algorithm is proposed.Principal component analysis was used to reduce the dimension of the sample characteristic parameters,k-means algorithm was used to optimize the initial clustering center of FCM algorithm,and the membership matrix of algorithm results was processed to achieve the function of "more selection and less selection".Compared with the traditional FCM algorithm,the improved algorithm has fewer iterations.The results of the improved FCM algorithm show that the standard deviation of each characteristic parameter decreases by a larger percentage before and after the screening,indicating that the screening effect is good.Secondly,the effect of thermal gradient on the aging rate of lithium battery was analyzed.In this paper,the capacity of 12 Ah lithium iron phosphate battery as the research object,set up the thermal gradient condition experiment platform.The effects of thermal gradient on the aging characteristic parameters(capacity,charging-discharge curve and internal resistance)of six batteries were analyzed by cyclic aging experiment,capacity calibration experiment and mixed pulse power characteristic test(HPPC).The temperature gradient set in this paper has no obvious effect on the battery aging characteristics.When the thermal gradient exists,the battery capacity decay rate is accelerated when the average temperature is too high or too low.When the average temperature is too high,the battery voltage platform decreases and internal resistance decreases;when the average temperature is too low,the battery capacity decreases and internal resistance increases slightly.Then,the influence of thermal gradient on the capacity decay of lithium battery was analyzed from the microscopic point of view and the electrochemical reaction mechanism inside the battery.The principles of P2 D model,heat transfer model and capacity decay model were described.The electrochemical-thermal coupling model was established according to the internal chemical reaction and heat transfer law of lithium battery.The electrochemical-thermal coupling model was combined with the capacity decay model of lithium battery to establish the electrochemical-thermal aging model.The simulation results show that the average temperature of the battery is the main factor influencing the capacity decay rate of the battery,and the average temperature is too high or too low to accelerate the aging of the battery.The conclusion is consistent with the aging experiment under thermal gradient condition and mutually verified.Finally,in view of the thermal gradient battery capacity attenuation under the condition of data’s volatility,and traditional extreme learning machine(ELM)instability and prediction accuracy problems,put forward sparrow based search algorithm(SSA)to optimize the ELM network weights and threshold of battery life prediction model,and USES the chapter 3 ageing experimental data for validation.The results show that compared with traditional ELM,the prediction results of ELM optimized by SSA have lower mean absolute error percentage and root mean square error,indicating that SSA-ELM model has higher prediction accuracy and better stability,and has obvious advantages in processing data with high nonlinear degree.