Abstract:

With the rapid development of renewable energy technologies, energy-storage batteries have gained widespread application in power systems. Accurately estimating the state of charge (SOC) of batteries is critical for ensuring their performance and safe operation and extending their lifespan. To improve the accuracy of SOC estimation for grid energy-storage batteries under varying power demands, we propose a method based on dynamic capacity correction. First, the error-generation mechanism of traditional SOC estimation methods under complex operating conditions was analyzed, and a general improvement strategy was proposed. Second, the capacity variation characteristics of batteries at different discharge rates were analyzed, and a quantitative model that characterizes the relationship between discharge rate and capacity was established, providing a theoretical foundation for accurate SOC estimation. Next, a hybrid estimation algorithm, CLA-EKF, was developed by integrating deep neural networks with the extended Kalman filter (EKF). This approach leverages the advantages of both methods in handling complex nonlinear relationships and resisting disturbances. Furthermore, an SOC estimation method with adaptive capacity correction based on discharge rate was developed. The experimental results demonstrate that the proposed capacity-correction-based CLA-EKF method significantly improves the accuracy of SOC estimation under various fluctuating power conditions, outperforming conventional methods. This study provides an effective solution for SOC estimation in grid energy-storage systems with high practical application value.

Key words: energy storage battery, SOC estimation, discharge rate, Kalman filter method, neural network