Search

CN-121978540-A - Capacity decay trend prediction method for low-temperature lithium battery

CN121978540ACN 121978540 ACN121978540 ACN 121978540ACN-121978540-A

Abstract

The invention provides a capacity fading trend prediction method of a low-temperature lithium battery, and relates to the technical field of battery health management and prediction. The capacity attenuation trend prediction method of the low-temperature lithium battery specifically comprises the following steps of S1, initial data acquisition, S2, feature extraction, S3, construction of a hybrid prediction model, S4, model training and offline verification, S5, online deployment and dynamic correction, and S6.SOH evaluation and early warning output. By integrating an electrochemical mechanism and deep learning to construct a hybrid prediction model, high-precision prediction of lithium battery capacity attenuation and full life cycle intelligent health management under a low-temperature environment are realized, the problems of accelerated battery aging and prominent safety risk in alpine regions are effectively solved, early warning and intelligent management of battery health states are facilitated, and the service life of the battery under low-temperature application is prolonged.

Inventors

  • Cui Shuhe

Assignees

  • 崔书赫

Dates

Publication Date
20260505
Application Date
20260318

Claims (8)

  1. 1. The capacity fading trend prediction method for the low-temperature lithium battery is characterized by comprising the following steps of: s1, initial data acquisition Setting a plurality of temperature gradients within a range of-80 ℃ to 10 ℃ and performing an accelerated cycle charge-discharge test to simulate an actual low-temperature use scene, and acquiring capacity attenuation, internal resistance change, voltage curve, temperature distribution and cycle times in the test process to form a complete aging data set under a multi-temperature condition; s2, feature extraction Based on an electrochemical principle, analyzing a main aging mechanism at low temperature, screening out a core input variable through correlation analysis, and establishing a model to construct a feature set; s3, constructing a mixed prediction model The electrochemical experience model and the LSTM deep learning model are fused to construct a hybrid prediction model, the electrochemical experience model adopts an Arrhenius equation to describe the influence of temperature on the aging rate, physical constraint and priori knowledge are provided, the LSTM deep learning model processes multi-dimensional time sequence characteristics, a long time sequence dependency relationship and a nonlinear attenuation mode are captured, and the two models are complemented through weighted fusion or residual error correction; s4, model training and offline verification Dividing experimental data into a training set, a verification set and a test set, training an LSTM (least squares) network by using early-stage circulating data and fitting Arrhenius equation parameters, evaluating a prediction error on the test set, and controlling an average absolute percentage error to be within 5% by adjusting the number of layers of the network, the learning rate and a time window so as to ensure that the model has reliable prediction performance; s5, on-line deployment and dynamic correction The trained model is accessed into a battery management system, rolling prediction is carried out based on temperature, SOC and current data acquired in real time, an online correction mechanism is established, and when the deviation between the actually measured capacity and the predicted value exceeds a threshold value, new data is utilized to finely adjust model parameters; S6.SOH evaluation and early warning output And calculating the SOH of the health state in real time, visually displaying the capacity fading trend and the residual life interval, simultaneously establishing a grading early warning system, wherein the SOH is higher than 80% and is in a normal running state, the interval between 60% and 80% is recommended to pay attention to and optimize the use strategy, and the replacement reminding is triggered when the SOH is lower than 60% or short-term failure is predicted.
  2. 2. The method of predicting a capacity fade tendency of a low temperature lithium battery as set forth in claim 1, wherein the plurality of temperature gradients in S1 comprises-80 ℃, -70 ℃, -60 ℃, -50 ℃, -40 ℃, -30 ℃, -20 ℃, -10 ℃, 0 ℃, 10 ℃.
  3. 3. The method for predicting a capacity fade trend in a low temperature lithium battery as defined in claim 1, wherein in S2, the main aging mechanism comprises lithium ion diffusion limitation, electrolyte performance degradation, SEI film thickening and lithium precipitation.
  4. 4. The method for predicting a capacity fade trend in a low temperature lithium battery as set forth in claim 1, wherein in S2, the core input variables include real-time temperature, state of charge, current stress, cumulative cycle number, and historical capacity retention.
  5. 5. The method for predicting the capacity fade trend in a low temperature lithium battery as set forth in claim 1, wherein in S3, the Arrhenius equation is given by , wherein, In order to be able to adapt the ageing rate constant, In order for the activation energy to be sufficient, Is a gas constant which is a function of the gas, Absolute temperature.
  6. 6. The method for predicting capacity fade trend in a low temperature lithium battery as set forth in claim 1, wherein in S3, the LSTM deep learning model comprises an input layer for inputting multi-dimensional time series characteristics including temperature, SOC, current and historical capacity, a hidden layer for capturing long time series dependencies for the LSTM network, processing a nonlinear fade mode, and an output layer for outputting a future capacity retention curve and a remaining cycle life.
  7. 7. The method for predicting capacity fade trend in a low temperature lithium battery as set forth in claim 1, wherein in S4, the training set, the validation set and the test set are in a proportion of 70% of the training set, 15% of the validation set and 15% of the test set.
  8. 8. The method for predicting capacity fade trend in a low temperature lithium battery as set forth in claim 1, wherein in S6, the calculation formula of the state of health SOH is as follows , wherein, For the currently available capacity it is possible, Is rated capacity.

Description

Capacity decay trend prediction method for low-temperature lithium battery Technical Field The invention relates to the technical field of battery health management and prediction, in particular to a capacity decay trend prediction method of a low-temperature lithium battery. Background The capacity fade of a lithium battery refers to the phenomenon that the storable and releasable electric quantity of the battery gradually decreases due to irreversible chemical reaction and physical change such as structural change of an electrode material, decomposition of an electrolyte, growth of lithium dendrite, thickening of an SEI film, loss of active lithium and the like in repeated charge and discharge cycles or long-term storage. The process is influenced by factors such as temperature, charge-discharge multiplying power, cut-off voltage, use habit and the like, and is a core problem for limiting the service life and performance of the battery. The prior patent (publication number: CN 120870899A) discloses a method and a system for predicting the capacity attenuation trend of a lithium battery, which relate to the technical field of battery health management and prediction, and specifically comprise the steps of generating deep fusion characteristics of data driving characteristics and electrochemical characteristics through a dual calibration mechanism, inputting the characteristic sequences into an electrochemical process perception model, encoding the characteristic sequences into potential state vectors for representing personalized degradation, inputting the vectors into a neural differential equation model as initial conditions, learning degradation dynamics rules and integrating and solving, and finally generating continuous health state attenuation tracks and obtaining a prediction result. The invention overcomes the defect of weak generalization capability of the traditional black box model by deeply fusing the physical mechanism and the data driving model, can realize long-term, high-precision and continuous prediction of full life cycle fading trend by only using early weak signals of the battery, and has high reliability and practical value. The inventor finds that the following problems exist in the prior art in the process of realizing the application, the existing capacity attenuation trend prediction method of the lithium battery mostly only focuses on the problem of battery capacity attenuation at normal temperature, and in some alpine regions, capacity attenuation of the lithium battery can be remarkably increased at low temperature, on one hand, the viscosity of electrolyte is increased at low temperature, the ion conductivity is reduced, the internal resistance of the battery is increased, the discharge capacity is temporarily reduced, on the other hand, the lithium ion intercalation speed at a negative electrode is reduced during low-temperature charging, and metal lithium precipitation is easily caused, so that irreversible active lithium loss and potential safety hazard are caused. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a capacity decay trend prediction method of a low-temperature lithium battery, which is beneficial to realizing early warning and intelligent management of the health state of the battery and prolonging the service life of the battery under low-temperature application. In order to achieve the purpose, the invention is realized by the following technical scheme that the capacity fading trend prediction method of the low-temperature lithium battery specifically comprises the following steps: s1, initial data acquisition Setting a plurality of temperature gradients within a range of-80 ℃ to 10 ℃ and performing an accelerated cycle charge-discharge test to simulate an actual low-temperature use scene, and acquiring capacity attenuation, internal resistance change, voltage curve, temperature distribution and cycle times in the test process to form a complete aging data set under a multi-temperature condition; s2, feature extraction Based on an electrochemical principle, analyzing a main aging mechanism at low temperature, screening out a core input variable through correlation analysis, and establishing a model to construct a feature set; s3, constructing a mixed prediction model The electrochemical experience model and the LSTM deep learning model are fused to construct a hybrid prediction model, the electrochemical experience model adopts an Arrhenius equation to describe the influence of temperature on the aging rate, physical constraint and priori knowledge are provided, the LSTM deep learning model processes multi-dimensional time sequence characteristics, a long time sequence dependency relationship and a nonlinear attenuation mode are captured, and the two models are complemented through weighted fusion or residual error correction; s4, model training and offline verification Dividing experimental data into a training set, a