Research On Prediction Of Remaining Useful Life Of Energy Storage Battery Considering Fusion Type Data-Driven

Lithium-ion batteries have the advantages of high energy density,high average output voltage,high output power,low self-discharge rate,no memory effect,fast charging,wide operating temperature range,no maintenance,environmental friendliness,and long service life.They are widely used in consumer electronics,electric vehicles,the national defense military,and other fields,and even future energy storage in the field,its market capacity is likely to exceed that of electric vehicles.However,lithium-ion batteries also have certain shortcomings,such as high cost,special protection circuits must be adopted to prevent overcharge or over-discharge,etc.If the lithium-ion batteries run for a long time,it will cause the internal function of the battery to age and reduce the capacity.It can cause damage to the company’s property,or endanger people’s lives.Therefore,it is very important to accurately predict the remaining useful life(RUL)of lithium-ion batteries.Accurate RUL prediction can improve the stability and safety of the battery-powered system,thereby identifying the occurrence of failures and responding promptly.This paper takes lithium-ion batteries as the research object and studies the state of charge(SOC)estimation and RUL prediction methods of lithium-ion batteries based on a fusion data-driven method.The main research work is as follows:1.Aiming at the second-order resistor-capacitor(RC)equivalent circuit model of lithium-ion batteries,online identification of parameters through the forgetting factor recursive least squares(FFRLS)method is the key and difficult point in the application of lithium-ion batteries.A method based on the fusion of the square-root high-degree extended kalman filter(SHEKF)and grey prediction model(GPM)is proposed to estimate the SOC of the lithium-ion battery,based on the second-order RC equivalent model to characterize the working characteristics of the lithium-ion battery.Online parameter identification and modification of battery model parameters are carried out through FFRLS;The SHEKF-GPM fusion model is proposed to estimate the SOC of lithium-ion battery online.2.To improve the particle depletion phenomenon and insufficient diversity when predicting the RUL of lithium-ion batteries by a particle filter(PF),a method of unscented kalman filter(UKF)and neural network(NN)fusion is proposed to estimate the RUL of lithium-ion batteries.Use UKF to estimate the degradation capacity data of lithium-ion batteries,and then use NN to recursively update the residual value of the UKF predicted battery capacity,and further predict the RUL of the lithium-ion battery.Select four different cycle periods of lithium-ion battery capacity degradation data,and design a comparative experiment based on UKF-NN fusion prediction to quantify the accuracy of the proposed method.3.The accuracy of lithium-ion battery RUL prediction depends on the degradation model,so a method based on time recurrent neural network(TRNN)and firefly algorithm(FA)to optimize PF fusion is proposed to predict the lithium-ion battery RUL.First,compared with traditional empirical models,the TRNN model is easy to capture the long-distance dependence between capacity degradations,to obtain better performance;Secondly,it is benefited from the unique mechanism of the FA algorithm,introduces FA optimization PF technology and recursively updates the TRNN degradation parameters,so that the particle population moves to high like the region to improve PF particle distribution,avoiding particles to replenish the phenomenon;Finally,for each group of data sets with different periods,the battery prediction results under different methods are quantitatively evaluated.