| Lithium-ion batteries are widely used in various fields such as aerospace,electric vehicles,smart grids,and new energy generation due to their unique advantages.Lithium-ion batteries are usually used as the key component of the power supply in the equipment.The degradation or failure of lithium-ion batteries will seriously affect the reliability and safety of the equipment.Therefore,the performance monitoring,state estimation,and prognostics of lithium-ion batteries have attracted more and more attention,and become the research hotspots in the field of prognostics and health management(PHM)technology.The advance of PHM technology for lithium-ion batteries can improve the reliability,availability,automatic diagnosis and health forecasting capabilities of the power supply system and even the whole equipment,which has significant theoretical as well as practical values.This article focuses on the key technologies of lithium-ion battery PHM,especially health prediction,state estimation and failure analysis of lithium-ion batteries.The major innovations and contributions of this article are drawn as follows:(1)Research on health prediction methods for lithium-ion batteries using electrical signalsBased on the electrical signals in the battery management system,the health prediction methods for lithium-ion batteries under practical operating conditions are studied,including capacity estimation and remaining useful life(RUL)prediction methods.1)Regarding the problem of degradation feature extraction of lithium-ion batteries under partial charge and discharge working conditions,a capacity estimation method based on semi-supervised convolutional neural network(SS-CNN)is proposed.This method can automatically extract features from battery partial charge information for capacity estimation.Besides,a semi-supervised training strategy is developed to take advantage of extra unlabeled data,which can improve the generalization of the model and the accuracy of capacity estimation.2)Aiming at the problem of predicting the RUL of lithium-ion batteries under complex and uncertain working conditions,a method based on the fusion of neural network degradation modeling and improved particle filter(PF)is proposed.The neural network,which is used to replace traditional empirical degradation models,can effectively model the battery degradation under complex operating conditions.While the particle distribution of PF is optimized based on the Bat algorithm,which solves the particle degradation and impoverishment problem.This Bat optimized PF method can accurately update the parameters of the degradation model and improve the accuracy of RUL prediction.(2)Research on state assessment methods for lithium-ion batteries using ultrasonic sensing signalsThis article studied the state-of-the-art state assessment technology of lithium-ion battery based on new sensing technology(ultrasonic sensing),including state of charge(SOC)estimation method and health assessment method.1)A SOC estimation method of lithium-ion batteries is proposed based on ultrasonic sensing and Gaussian process regression(GPR)models.By integrating the ultrasonic time of flight(TOF)feature with traditional electrical signals,the SOC estimation model achieves better accuracy under dynamic operating conditions.Furthermore,by fusing the temperature signal for input,a temperature aware model is built for SOC estimation under different environmental stresses.2)A health assessment method for lithium-ion batteries based on ultrasonic sensing and Mahalanobis distance metric is proposed.First,the relationship between ultrasonic behavior and battery health status is analyzed by the experimental study,which is used for chosen ultrasound features to characterize the health of the battery.Secondly,by combining ultrasonic features and battery surface temperature,a new battery health index is constructed to assess battery degradation and detect early failure.(3)Research on failure analysis methods for lithium-ion batteries1)Aiming at the degradation analysis of battery individual electrodes in practical applications,a method for evaluating the electrode behavior based on the external characteristics of the battery is proposed.This method is based on a three-electrode system to in-situ monitor the performance of the battery and each electrode.Then,the relationship between the electrode behaviors and the full cell performance can be established,which can be used for extracting the voltage and impedance features for electrode degradation analysis.Those features reveal the internal degradation mechanism of the lithium-ion batteries.2)Regarding the lithium-ion battery failure caused by manufacturing defects,a non-destructive failure analysis method based on computed tomography(CT)scanning technology is proposed.The practical guidances for battery analysis based on CT scanning,including CT scanning parameters selection and CT images interpretation,are detailed introduced.By investigating different lithium-ion batteries failure cases in real accidents,various manufacturing defects and structural deformation issues are studied,which can help evaluating battery quality and analyzing the failure causes of battery. |
