Abstract
Use of Artificial intelligence in power systems is the mainstream theme in the scientific and engineering communities. The integration of renewable energy sources and the increasing demand for electricity have led to a significant increase in the complexity of modern power systems. To address this challenge, there is a need for advanced control and analysing techniques that can ensure the stable and efficient operation of the power grids. This paper explores the application of large language models (LLMs) for control of power transmission and distribution networks, specifically focusing on the use of an LLM-based agent in the control and analysis power system networks. The paper begins by reviewing the current state of LLMs and their potential applications in power systems. It then presents a detailed description of the proposed Llama3-based agent, which is designed to interpret and execute control commands using natural language processing (NLP) techniques. The agent can interface seamlessly with the Pandapower model, allowing it to monitor the network in real-time and make informed decisions to maintain the system’s stability and efficiency. To evaluate the effectiveness of the proposed approach, the paper presents case studies using simulated power transmission and distribution network models. The results demonstrate that the LLM agent can quickly and accurately respond to changes in the system, maintaining its stability and efficiency even under challenging conditions. Second example is contingency analysis with llama3 based agent. AI agent gives recommendations for mitigation of contingencies in the predefined system. The paper also discusses the potential, limitations and challenges of using LLMs for the control and analysis of power systems and suggests directions for future research. Uncertainties of the LLMs are also interesting topic and opens the new point of view on power system problems. Overall, this paper provides insights into the potential of LLMs for control and analysis of power transmission and distribution networks and demonstrates the feasibility of using LLM agents for this purpose. It highlights the benefits of this approach, including improved system stability and efficiency, and provides a foundation for further research in this rapidly evolving field.