Computational methods for optimizing operational strategies of energy storage systems based on generative artificial intelligence

INTRODUCTION: The energy storage systems (ESSs) require the coordination of renewable generation, load demand, electricity prices and safety constraints but operational strategy optimization is much dependent on manual modeling and is hard to change based on users preferences. THE OBJECTIVES ARE: The study will come up with an automated and constraint-based optimization structure of ESS scheduling. It is proposed that a constraint-improved generative artificial intelligence (AI)-based model to optimize energy storage operation strategies, called CGAI-ESO, be developed. METHODOLOGY: The model is based on natural language processing to understand the requests of energy management system (EMS) developers or end-users, retrieval-augmented generation (RAG) to fetch domain and optimization data, and a generative AI agent to develop objective functions, constraints, and executable optimization codes. The combination of an optimization solver with a digital twin feedback module checks the validity of economic performance, safety compliance, battery degradation and constraint satisfaction. The results of experiments with public energy data and microgrid simulations demonstrate that CGAI-ESO minimizes operational cost by 18.72 percent, maximizes peak shaving rate to 28.5 percent, enhances the use of renewable energy to 96.0 percent and minimizes the level of constraint violation to 0.02 percent. It is also better than large language model (LLM)-only and LLM-RAG baselines in terms of model generation, constraint completeness and code executability. CONCLUSION: CGAI-ESO provides a feasible route for intelligent and reliable ESS operation strategy optimization.