Artificial intelligence in state of charge estimation: Pioneering approaches across energy storage systems

This review investigates the role of artificial intelligence in predicting the state of charge for thermal energy storage devices. Traditional estimation methods often struggled with complex dynamics and large-scale data, showing accuracy limitations of 5–10 % under dynamic conditions. Artificial intelligence significantly improved accuracy, efficiency, and scalability, achieving 98 % prediction accuracy in electrical storage, a 30 % efficiency gain in thermal energy storage, a 77 % reduction in power fluctuations for mechanical storage, and a 40 % efficiency boost in chemical storage. The review analysed various artificial intelligence methodologies applied to thermal energy storage, including neural networks, support vector machines, reinforcement learning, and hybrid models, which reduced computational time by up to 60 %. Integrating artificial intelligence with the internet of things and big data enabled real-time analysis of thermal energy storage systems, reducing monitoring latency by 70 %. However, challenges persisted regarding data integrity, integration costs, and ethical concerns. The study also revealed implementation gaps within thermal storage technologies, with artificial intelligence adoption at 15 % in latent thermal energy storage compared to other energy systems like electrical storage where adoption reaches 85 %. Future research should focus on explainable artificial intelligence models, robust data quality frameworks, and standardized integration protocols, specifically tailored for the unique challenges of thermal energy storage including sensible, latent, and thermochemical systems. This review highlights the transformative impact of artificial intelligence on state of charge estimation in thermal energy storage systems, paving the way for more efficient and reliable energy management strategies.