Estimation of Lithium-ion Batteries State of SOC Based on Borges Difference Structural Adaptive Gated Recurrent Unit Networks

A gated recurrent unit (GRU) network based on Borges difference is proposed, which is called B-GRU network. The B-GRU network aims to solve the limitation that the integer-order GRU network lacks sufficient flexibility in adjusting the proportion of updated information and forgotten information. The B-GRU network is specifically applied to the state of charge (SOC) estimation of lithium-ion batteries (LIBs). By introducing Borges difference into the GRU network structure and combining with strict order interval analysis, the convergence of the improved B-GRU network is guaranteed. The optimization of the network is achieved by adjusting the order rules in the B-GRU network. The experimental results show that the proposed B-GRU network effectively improves the SOC estimation accuracy of LIBs.