Battery cycle life prediction

Due to the long time and high cost of battery cycle life detection, the establishment of life model and life evaluation and prediction have become the research hotspots of scholars at home and abroad. Lithium battery life prediction methods can be divided into three categories according to information sources: prediction based on capacity decline mechanism, prediction based on characteristic parameters, and prediction based on data-driven. 1 Prediction based on capacity decay mechanism The prediction based on the mechanism is based on the aging and decay mechanism of the internal structure and materials of the battery during the cycle to estimate the battery life. This method requires the use of basic models to describe the physical and chemical reaction processes that occur inside the battery, such as Ohm's law, electrochemical polarization, concentration polarization, and internal diffusion of electrode materials. Based on the loss of active lithium ions during the battery cycle, Ning et al. used first principles to simulate the capacity decline model of lithium cobalt oxide batteries. The influencing parameters include exchange current density, DOD, interface membrane impedance, and charge cut-off voltage. The author compares the life prediction model with the measured data and finds that the model is very close to the actual test results. Virkar proposed a battery degradation model based on non-equilibrium thermodynamics, taking into account the influence of chemical potential and SEI film on capacity degradation, and pointed out that there will be unbalanced monomers in the series battery pack, and the interface between the positive electrode and the electrolyte SEI film may also be produced, leading to increased capacity degradation. 2 Prediction based on characteristic parameters Prediction based on characteristic parameters refers to the use of changes in certain characteristic factors of the battery during the aging process to predict battery life. At present, researchers pay most attention to the relationship between EIS and cycle life. Li et al. studied the changes in impedance spectra of commercial lithium cobalt oxide batteries during 1C charge and discharge cycles, and observed the changes in electrode materials by XRD, TEM and SEM. They found that the Nyquist curves of the positive and negative electrodes of lithium batteries correspond to The size of the semicircle in the low-frequency region of the interface membrane impedance increases with the increase of the number of cycles, and the cycle life of the battery can be inferred from this. EIS can give a finer description of battery impedance, but the detection instrument is susceptible to external interference and it is difficult to perform effective analysis on complex spectra. In contrast, the measurement of pulse impedance is simple and easy, and online monitoring can be quickly realized. 3 Data-driven prediction The data-driven approach refers to the direct analysis of test data to mine the law without considering the physical and chemical reactions and mechanisms inside the battery. It is a simulation method based on experience. The more common ones are time series model (AR), artificial neural network model (ANN) and correlation vector method (RVM). The AR model is based on the measured data at some previous time points to infer the predicted value in the current state, and has linear characteristics. Taking into account the non-linear relationship between battery capacity attenuation and the number of cycles, Luo Yue proposed an improved non-linear AR model, which introduced accelerated degradation factors in the later stage of the prediction to improve the accuracy of prediction. The ANN model is an artificial intelligence network system composed of multiple neurons according to certain rules, and is a typical nonlinear model. The RVM model is a data regression analysis method, which can flexibly control over-fitting and under-fitting by adjusting parameters, and has the characteristics of probabilistic prediction. The prediction method based on the internal mechanism has better theoretical support and better accuracy, but the complexity is large. The advantage of the data-driven method is that it is simple and practical, but because the acquired data cannot cover all parameters, it also has certain limitations. Sex. This article mainly analyzes the influencing factors of power lithium-ion battery cycle life and the research of life prediction model. It can be seen that there are many factors that affect the cycle life of power lithium batteries, and for lithium batteries of different materials and structures, the influencing factors are also different. From the analysis in the article, we can see that we can extend the battery life by controlling the parameters, such as allowing the battery to work under the appropriate temperature, rate, and charging and discharging conditions. Relatively speaking, the influencing factors of the cycle life of the battery pack are more complicated, because these factors will have mutual coupling effects, and the monomer consistency problem will cause the performance of the battery pack to not be fully exerted, and seriously shorten the cycle of the battery pack. life. When predicting the cycle life of the battery, it can be based on the internal mechanism of the battery, a certain characteristic parameter or a large amount of data that has been measured, and the establishment of an accurate, reasonable and simple and operable model can accurately evaluate the cycle life of the battery and further optimize its performance. It is of great significance. Disclaimer: Some pictures and content of articles published on this site are from the Internet. If there is any infringement, please contact to delete. Previous: All solid polymer electrolytes for lithium batteries?