变温度工况下锂离子电池健康状态退化预测

State of health(SOH) degradation prediction of lithium-ion batteries under variable temperatures is crucial for their health management, safety, and reliability. Existing parametric methods with strong prior assumptions on degradation trajectories are not suitable for batteries with diverse degradation modes at different temperatures. To accurately predict SOH degradation, a non-parametric estimation method for SOH degradation trajectory and its transformation coefficient is proposed, which models the SOH degradation at different temperatures without strong assumptions. Secondly, an adaptive polynomial fitting method for temperature and transformation coefficient is proposed, whose comprehensive loss function ensures the fitting accuracy, robustness, and rationality, and can achieve SOH degradation prediction at any constant temperature. Then, an offline SOH degradation prediction method under variable temperatures is proposed, where the equivalent cycle number can consider the impact of cumulative degradation on future SOH degradation. Finally, an online update method is proposed, which incorporates the influence of historical temperature into the updated prediction through the variable temperature interaction coefficient. The effectiveness of the proposed method and its superiority over commonly used methods are verified by experimental data. The mean absolute error of SOH prediction at constant temperature is at least 32% lower than other methods, and the minimum mean absolute error under variable temperature can reach 0.002 6. This study can provide accurate predictions for the SOH degradation and lifetime of lithium-ion batteries under complex and variable temperature conditions, effectively ensuring their health management and safe and reliable operation.