Application of Large Language Models in Decision-Making Systems of Power Engineering Tasks

Managing large power systems involves complex tasks traditionally handled by human operators. The emergence of Smart Grids, integrating small distributed power generation, has further complicated grid management due to increased structural complexity and the involvement of unqualified stakeholders. In response to these challenges, there is growing interest in utilizing artificial intelligence to assist or replace human operators. However, existing artificial intelligence systems are often scenario-specific and lack the adaptability needed for dynamic grid configurations. This article introduces an innovative AI-based control system that leverages natural language processing capabilities of large language models. By analyzing grid states and user instructions provided in natural language, the AI generates adaptive control actions to maintain grid stability. Extensive experiments demonstrate the system's effectiveness in handling various load scenarios, including both gradual and abrupt changes. The artificial intelligence-based control system successfully performs actions akin to those of a human operator, such as activating additional generation or disconnecting loads to stabilize the grid. The system's flexibility allows it to incorporate user commands and adapt to specific conditions, ensuring precise and dynamic control actions.