A compact and optimized neural network approach for battery state-of-charge estimation of energy storage system. (15th March 2021)

• Record Type: Journal Article
• Authors: Guo, Yuanjun; Yang, Zhile; Liu, Kailong; Zhang, Yanhui; Feng, Wei
• Is Part Of: Energy. Volume 219(2021)
• Publication Date: 2021-03-15
• Subjects: State-of-charge estimation -- Energy storage system -- Neural network -- JAYA optimization; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2020.119529
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Accurate estimations of battery state-of-charge (SOC) for energy storage systems are popular research topics in recent years. Numerous challenges remain in several aspects, especially in dealing with the conflict of high model accuracy and complex model structure with heavy computational cost. This paper proposes a compact and optimized SOC estimation model, integrating a fast input selection algorithm to choose important terms as input variables, followed by a simple and efficient JAYA optimization scheme to tune the key parameters of neural network functions. From the real-system experiment results, it can be seen that the estimation model errors are greatly reduced by applying optimization method, and the model performance is validated through statistical error values including root mean square error, mean absolute error, mean absolute percentage error and SOC error. The experimental results demonstrate that the SOC estimations can be greatly improved after optimization of neural network parameters under different charging and discharging process. Highlights: A novel compact and optimized radial basis function neural network is proposed. The method is adopted for the SOC estimation of energy storage systems. Simulation results validate the excellent performance of the proposed method.