A major challenge in ensuring the reliability of battery systems is the uncertainty surrounding their service life. An accurate prediction of the remaining useful life (RUL) is essential for effective maintenance and operation. Traditional extended Kalman filter (EKF) algorithms, which rely heavily on historical data, often have limited long-term prediction accuracy. To overcome this problem, a multi-observation fusion approach is proposed to enhance the performance of the EKF for battery life prediction. To improve the practical applicability, the conventional aging capacity test values with operational capacity data obtained under real-world conditions are replaced. This modification refined the trajectory of the battery aging state space, thereby reducing the dependency on the quantity and quality of historical data while simultaneously boosting the long-term prediction accuracy and stability. Furthermore, a semi-empirical aging model is introduced to extract prior knowledge from offline data. This provided valuable insights into the life degradation trends and guided the filtering process. The resulting framework forms the basis for a novel RUL prediction method that utilizes a multi-observation fusion EKF. Validation experiments show that the proposed method enhanced the prediction accuracy by over 60% compared with traditional EKF and particle filter algorithms throughout the lifecycle of lithium-ion batteries. Additionally, the technique exhibited robust stability (with no divergence observed over the full battery life) and demonstrated notable improvements in early stage RUL prediction.

Supercapacitors are widely used in transportation and renewable energy fields due to their high power density, stable cycling performance, and rapid charge–discharge capabilities. To ensure efficient applications of supercapacitors, accurately predicting their degradation trajectories and remaining useful life (RUL) is crucial. For this purpose, a physics-informed neural network (PINN) model is developed using Long Short-Term Memory (LSTM) as the base architecture. Physical equations are embedded into the loss function to ensure consistency with domain knowledge, allowing the loss function to incorporate both physical and data-driven components. The balance between these two loss components is dynamically determined through Bayesian optimization, to enhance the model's accuracy further. Validation results show a root mean square error (RMSE) of 3 ​mF (the rated capacity is 1 F) in the degradation trajectory prediction and a RMSE of 269 cycles (the average cycle life is 5180 cycles) for the RUL. Ablation experiments were conducted to validate the effectiveness of integrating physical information into the LSTM framework. Results demonstrate that the proposed model outperforms both the data-driven LSTM method and the empirical equation-based method that the PINN model can reduce the RMSE by 85% and 87.5% for degradation trajectory prediction, and 86.5% and 94.6% for RUL prediction, respectively. In addition, a comparison with advanced models demonstrates that our model reduces the requirement significantly on training data while maintaining comparable prediction accuracy, which favors scenarios where data is scarce.

This paper proposes a novel power management strategy for plug-in hybrid electric vehicles based on deep reinforcement learning algorithm. Three parallel soft actor-critic (SAC) networks are trained for high speed, medium speed, and low-speed conditions respectively; the reward function is designed as minimizing the cost of energy cost and battery aging. During operation, the driving condition is recognized at each moment for the algorithm invoking based on the learning vector quantization (LVQ) neural network. On top of that, a driving cycle reconstruction algorithm is proposed. The historical speed segments that were recorded during the operation are reconstructed into the three categories of high speed, medium speed, and low speed, based on which the algorithms are online updated. The SAC-based control strategy is evaluated based on the standard driving cycles and Shenyang practical data. The results indicate the presented method can obtain the effect close to dynamic programming and can be further improved by up to 6.38% after the online update for uncertain driving conditions.