| Lithium-ion batteries have large power density,low self-discharge rate,no memory effect,and long cycle life,thus they have attracted much attention in modern electronic devices,communication equipment,medical equipment.With lithium-ion batteries being vigorously developed,they are also gradually becoming the energy source of choice in the field of new energy resources,especially in the new energy vehicle industry.The state of health and remaining useful life of Lithium-ion batteries are of great importance to the secure and steady running of the whole system.Thus,the battery management system needs to monitor the batteries in real time,realize the prediction of state of health and remaining life,further guide the working and maintenance of Lithium-ion batteries,which can guarantee the security and dependability of the system.In view of the time-varying,dynamic and nonlinear characteristics of lithium-ion battery in application,as well as the capacity regeneration phenomenon,this paper carries out research on the prediction method of the state of health and remaining life of lithium-ion battery by using the fused health indicator and improved Gaussian process regression model.First of all,in this paper,many literature researches have been conducted on the methods of predicting the remaining life of lithium-ion batteries and extracting health indicators,and determines to use the Gaussian process regression method to predict the remaining life of lithium-ion batteries.The operating theory of lithium-ion battery is analyzed,and the aging process of battery is analyzed from the external characteristic parameters,which provides a theoretical basis for extracting health indicators in the subsequent stage.Secondly,according to the analysis of the change law of the data of the lithium-ion battery during the charging period,from the voltage,current and temperature curves,the charging time for equal voltage interval,the charging time of constant current,capacity increment peak value,capacity increment peak position,and the time of temperature peak are extracted as the five health indicators of lithium-ion batteries.Without losing the original information,the stack autoencoder is used to fuse the five health indicators to eliminate redundant information,which improves the generalization ability of the model.Then,the algorithm theory of Gaussian process regression is investigated deeply,and the effect of covariance function on model is discussed.The state of health estimation with different steps ahead is realized by fused health indicator and Gaussian process regression model.Experimental validation is performed using two battery datasets,and the results show that the method has good precision for achieving state of health estimation,as well as uncertainty expression capability.Finally,the remaining life prediction model is established by combining variational mode decomposition,Gaussian process regression and dynamic adaptive immune particle swarm optimization.In view of the capacity regeneration phenomenon of batteries,decomposing the fused health indicator by using variational mode decomposition to mine the internal information of the data and reduce the data complexity.For different components,the Gaussian process regression prediction model is established using different covariance functions,which can effectively capture the long-term declining trend and short-term regeneration phenomenon.The dynamic adaptive immune particle swarm algorithm is obtained by improving the particle swarm algorithm,which optimizes the Gaussian process regression model,realizes the determination of the hyperparameters of the kernel function,establishes a more accurate degradation relationship model,and finally realizes the accurate prediction of the remaining life,as well as the uncertainty characterization.The proposed fusion type algorithm is used to predict the remaining life at different starting prediction points using NASA battery and CALCE battery data,and the offline prediction results show that the proposed method can solve the problem of large prediction error caused by capacity regeneration problem.Setting up comparison experiment,the results show that the fusion type algorithm has better generalization performance and higher prediction accuracy compared with other models,which proves the effectiveness of the fusion type algorithm in achieving accurate remaining life prediction of lithium-ion batteries. |
