Multi-output fusion SOC and SOE estimation algorithm based on deep network migration

Abstract

The burgeoning fields of electric vehicles and renewable energy systems necessitate precise battery state estimation to optimize battery management and enhance both performance and reliability. This study presents a novel MS-SENet-GRU-UKF-TL model which incorporates a multi-scale squeeze-and-excitation networks (MS-SENet), gated recurrent units (GRU), unscented Kalman filter (UKF), and transfer learning (TL) to refine the accuracy of estimating the state of charge (SOC) and state of energy (SOE). SENet networks, featuring varying convolutional kernels, are deployed to extract features effectively across multiple scales. The effective deployment of the multiscale attention mechanism enhances the precision in capturing essential data information. Additionally, the application of the UKF algorithm improves the precision and smoothness of the model's outputs. The framework is verified under four data sets at different temperatures. Experimental results show that compared with other models, the average accuracy of the proposed SOC estimation method is enhanced by at most 37.486 %, and the average accuracy of SOE is enhanced by 35.432 %. Finally, the novel algorithm perform transfer learning on different battery data to validate the adaptive performance of the proposed method under different materials and capacity.