What type of study is this?

September 5, 2025

Estimation of lithium-ion power battery state of charge based on SRC-GRU modeling

Key Points

The SRC-GRU model significantly improves state of charge estimation accuracy for lithium iron phosphate batteries.
The model achieved reduced errors compared to the standard GRU model, enhancing battery management system effectiveness.
Time-series data including voltage, current, and temperature were utilized to train the SRC-GRU model under varying conditions.
Evaluation metrics indicated that the SRC-GRU outperformed the original GRU in SOC prediction accuracy.

Abstract

Battery Management System (BMS), as a bridge between the power battery and the vehicle, has an important role in monitoring and protecting the power battery. Accurate estimation of the state of charge (SOC) of power batteries is critical for energy management and safe operation of electric vehicles. In this paper, an enhanced gated recurrent unit (Swish-activated, Regularized, and Clipped-output Gated Recurrent Unit, SRC-GRU) neural network approach is proposed based on Recurrent Neural Network (RNN); The A123 lithium iron phosphate battery data from the University of Maryland's battery test open-source dataset was selected to model the time-series characteristics of battery voltage, current and temperature under constant-current charging and discharging conditions at different ambient temperatures, to achieve the prediction of the SOC; and the root-mean-square error and the average absolute error were used to measure the estimation accuracy. Finally, by comparing the estimation results with the original GRU (Gated Recurrent Unit, GRU), the estimation accuracy of SRC-GRU network is significantly better than that of the original GRU network.

AI से पूछें

Bookmark