Research On Remaining Useful Life Prediction Method Of Lithium-ion Battery Based On ELM

With the rapid development of electrochemical energy storage technology,lithium-ion batteries have been widely used in power system energy storage due to their excellent performance.However,due to factors such as internal electrochemical reaction and external working conditions,the performance of lithium-ion battery is gradually declining as the number of cycles increases.If the deterioration of battery performance can be predicted and the battery can be replaced in time,system faults or accidents caused by battery failure will be avoided,thereby reducing casualties and property losses.Therefore,effective prediction of the remaining useful life of lithium-ion batteries has become one of the focuses of attention both inside and outside the industry.In this paper,the research is carried out in three aspects,including the acquisition of indirect health factors,the parameters optimization method of kernel extreme learning machine,and the prediction method of multiple kernel extreme learning machine based on parameters optimization.The main work is as follows:(1)Aiming at the problem that direct health factors are difficult to obtain directly,it is proposed to use indirect health factors to characterize battery life attenuation characteristics.Through the analysis of three experimental datasets,four indirect health factors are extracted as the inputs of the prediction algorithm,namely,equal voltage rise charge time,equal voltage drop discharge time,time for charge temperature to reach peak value,and time for discharge temperature to reach peak value.And the effectiveness of the extracted indirect health factors is verified by correlation analysis.(2)To solve the parameters value problem of the kernel extreme learning machine algorithm,an improved grey wolf optimization is proposed for optimizing the parameters of kernel extreme learning machine.Firstly,strategies such as Tent chaotic mapping,nonlinear parameter control and weight-based position update are introduced and improved the grey wolf optimization algorithm.Then,the parameters of the single kernel extreme learning machine algorithm based on linear kernel function,polynomial kernel function and gaussian radial basis kernel function are optimized by the improved grey wolf algorithm,respectively,and the parameter optimized single kernel extreme learning machine is used for lithium-ion battery remaining useful life prediction.Finally,the experimental results demonstrate the feasibility of the single kernel extreme learning machine algorithm using improved grey wolf algorithm.(3)In order to further improve the predictive ability of single kernel extreme learning machine,this paper proposes multiple kernel extreme learning machine prediction method using parameter optimization.Firstly,by introducing the linear combination of polynomial kernel function and gaussian radial basis kernel function into kernel extreme learning machine,a prediction method based on an improved grey wolf algorithm optimized dual kernel extreme learning machine is proposed.Then,the linear kernel function is introduced to dual kernel extreme learning machine to propose a prediction method based on the improved gray wolf algorithm optimized triple kernel extreme learning machine.Finally,the experimental results show the proposed method can effectively reduce the prediction error and accurately predict the remaining useful life of lithium-ion batteries.