State of Charge Estimation for Hybrid Supercapacitors Under Multiple Temperature Conditions Based on A Hybrid Neural Network Model

The dual advantage of power and energy density renders hybrid supercapacitors(HUCs) a strategic focus in next-generation energy storage research. Achieving reliable and precise estimation of the state of charge (SOC) in HUCs is vital for enhancing system management efficiency and promoting their industrial-scale deployment. In practical applications, energy storage systems often face challenges due to significant temperature fluctuations. In response to this challenge, we propose a deep learning architecture that integrates a convolutional neural network, an attention mechanism, and a Bi-directional long short-term memory (CNN-AM-Bi-LSTM) network. This model first uses an attention mechanism to weight the output features of convolutional layers, enhancing the model’s reliance on important information, It then model captures temporal dependencies across multiple time scales through the use of a Bi-LSTM network, further improving the model's accuracy and generalization ability. The model is validated under different temperature conditions of DST operation, and experimental results show that the CNN-AM-Bi-LSTM network can adapt to the characteristics of HUCs at different temperatures. Typically, the estimation error remains below 2%, and the model maintains good performance even at lower temperatures.