Remaining useful life prediction of lithium-ion batteries based on POA-SVR method in experimental platforms

In experimental platforms with energy storage units, the performance degradation of energy storage batteries has become a key issue restricting the operational effectiveness of the platforms. To improve the accuracy in predicting the remaining useful life (RUL) of batteries in such scenarios, this paper proposes a lightweight, interpretable and small-sample-friendly hybrid prediction method based on the pelican optimization algorithm support vector regression (POA-SVR). Firstly, by analysing the full lifecycle charging and discharging data of energy storage batteries, a multi-dimensional degradation feature set is constructed. Combined with Pearson correlation analysis, feature dimensionality reduction is achieved while retaining key degradation information, effectively solving the curse of dimensionality problem caused by high-dimensional features. By introducing the POA to adaptively optimize the penalty factor and kernel parameters, a prediction model with dynamic parameter adjustment capabilities is established, which significantly improves the degradation trajectory fitting ability. Based on the NASA dataset, experiments are carried out by comparison with other optimization algorithms, which show that the value of the mean absolute percentage error (MAPE) of the POA-SVR remains below 0.7 %, and the value of root mean square error (RMSE) remains below 0.02, achieving higher prediction accuracy and demonstrating the high robustness and reliability even with small samples. It can be concluded that the proposed method offers an interpretable and low-data-dependent lifespan prediction tool for health assessment of energy storage experimental platforms, extending their effective operational cycles.