Estimating state of charge of battery in renewable energy systems: a data-driven approach with artificial neural networks

Abstract Accurately determining the state of charge is crucial for efficient battery management and reliable operation in renewable energy systems. This study presents a novel approach utilizing an artificial neural network to estimate the state of charge of a battery based on key variables such as battery voltage, charging current, load current, and temperature. A photovoltaic solar energy system was installed at the American University of Ras Al Khaimah, where a comprehensive experiment was conducted over 146 hours, resulting in the collection of 3930 high-resolution data points at 2-minute intervals. The artificial neural network model was trained on 70% of the data, while the remaining 30% was used for validation and testing. The trained model achieved a mean square error of 0.000896 and an error rate of less than 2% in the state of charge estimation. The model demonstrated excellent predictive accuracy, with a determination coefficient (R²) of 99.9%, indicating a strong correlation between the predicted and actual state of charge values. These results validate the robustness of the developed artificial neural network model and its potential for real-time state of charge estimation in renewable energy systems, providing a reliable and computationally efficient alternative to traditional state of charge estimation methods.