SOH estimation of lithium battery based on improved quantum particle swarm optimization hybrid neural network

Accurate prediction of the State of Health (SOH) of batteries is crucial for ensuring their long-term safe and effective operation. Aiming at the challenges faced in the SOH prediction research of lithium batteries, such as the limited processing of health features, the stochastic nature of the attention allocation, and the scientific nature of model hyperparameter settings, this paper proposes an improved quantum particle swarm optimization hybrid neural network for SOH estimation of lithium batteries. Firstly, health features are extracted from multiple dimensions such as voltage, current, temperature, and incremental capacity. Secondly, the importance indicators of these features are calculated and optimally ranked by fusing the random forest algorithm and the mutual information approach, and the obtained low-importance features are downscaled to obtain the indirect health features, which are inputted into the squeeze-excitation attention enhanced convolutional neural network along with the high-importance features to sufficiently extract features. Subsequently, the bidirectional long short-term memory neural network is used to fully extract long-term dependencies. Finally, the improved quantum particle swarm is used for hyperparameter optimization to achieve global optimization. The proposed method has been verified to have superior predictive performance using the NASA and Oxford battery datasets. Experimental results show that the mean absolute error, mean absolute percentage error, and root mean square error of the proposed method are within 1.5% and 0.8% in the two datasets, respectively, far lower than other methods, and have a high accuracy of SOH estimation. Therefore, the proposed method is expected to be an effective information guide for battery health management.