| In order to effectively respond to global climate change and achieve the "double carbon" goal,new energy vehicles are an important driving force for environmental protection and energy conservation.Due to the excellent performance of the Lithium iron phosphate battery,it has become an important power source for electric vehicles.The performance and service life of battery systems are receiving more attention from society.However,some details in the application process of the battery pack have become increasingly prominent.The same type of single battery presents different aging rates under the same working conditions.The performance of the single battery determines the overall life of the battery system.State of Health(SOH)is a key indicator of battery life and an important part of BMS(Battery Management System).Therefore,accurate estimation of the health status of batteries is of great significance to improve the overall performance of the battery pack and the safety level of new energy vehicles.At the same time,it is helpful to promote the healthy and stable development of the new energy vehicle industry.With the wide application of big data and cloud technology,the data of electric vehicles can be monitored.These raw data can be used to evaluate the consistency of battery packs,estimate battery life and conduct battery safety warnings.In addition,the inconsistency of the individual cells in the battery pack is also one of the problems that cannot be ignored.In the aging process,although the overall working conditions are the same,there are still some differences between the individual cells in terms of the external characteristics of the voltage and current of the battery and the capacity decay rate;Moreover,with the aging of the battery,the degree of this difference will gradually increase,which also increases the difficulty of accurately estimating the battery SOH.Therefore,it is necessary to deeply analyze the aging difference characteristics of different monomers of the same brand and model in the whole life cycle,and then design a SOH estimation scheme that can adapt to this difference.In this paper,the aging difference characteristics and SOH estimation of lithium iron phosphate batteries are studied.The main work is as follows:(1)Study on aging characteristics and aging differences of lithium iron phosphate battery under different charge discharge rates.According to the capacity decay characteristics of lithium iron phosphate battery,the battery samples were tested in cycles based on the standard rate and the acceleration rate,and the decline characteristics of the battery were analyzed from three factors of voltage,current and capacity.The results indicate that the larger the charge discharge rate,the faster the battery decay.Meanwhile,a large amount of raw data and parameters related to battery SOH were obtained through cyclic testing.For multiple single battery samples of the same brand and model,independent experimental tests were carried out under the same working stress conditions.It was found that the difference of lithium iron phosphate battery runs through the entire life cycle,and there is no specific linear relationship between the life of lithium iron phosphate battery and the number of cycles,and the difference tends to increase in the late cycle aging.Under different magnification,aging differences also vary.The general rules of aging difference characteristics and multi rate charging characteristics of lithium iron phosphate battery are summarized to lay a foundation for battery SOH estimation.(2)SOH estimation of battery based on voltage probability density.The idea of probability density function(PDF)is introduced.Under different magnifications,the cumulative value of voltage sampling frequency near the peak of the voltage probability density curve is taken as the health feature.The relationship between characteristic parameters and SOH is obtained by the fitting method,and the aging characteristic table under different magnifications is obtained.On this basis,the fixed voltage window and sliding window optimization methods are used to determine the voltage characteristic interval.Finally,the SOH estimation accuracy obtained by the two schemes is compared and analyzed.(3)SOH estimation of battery based on Sparse Auto Encoder(SAE)and Back Propagation Neural Network(BPNN).A comprehensive learning network structure based on SAE-BPNN is constructed.First,SAE is used to extract hidden compression features related to battery aging status from the voltage data of battery charging later stage;Then,the nonlinear mapping between the aging characteristics and battery SOH is established by BPNN;Finally,using the optimized SAE-BPNN network and the voltage data at the end of battery charging,the battery SOH can be estimated.In addition,this paper also analyzes the accuracy of SOH estimation based on SAE-BPNN and its adaptability to the aging difference between cells.(4)SOH estimation of battery based on generated model.Combining the Va DE model and WGANGP model,established a Va DE-WGANGP comprehensive model.Using open source data sets to generate external battery characteristic curves and construct comprehensive data resources;Finally,the battery SOH is estimated using a convolutional neural network combined with comprehensive data resources. |
