1School of Electrical Engineering, Shandong University, Jinan 250061, China;2Academy of Intelligent Innovation of Shandong University, Jinan 250100, China;3Dispatching Center of State Grid Shandong Electric Power Company, Jinan 250012, China

Power flow adjustment in power systems requires dynamic correction of control variables in response to problems such as limit violations. However, result judgment and data adjustment rely heavily on manual work, and expert experience is difficult to quantify and generalize. With the rapid iteration and advancement of large language model (LLM) technology, LLM-driven agent frameworks have demonstrated strong adaptability, flexibility, and robustness in handling complex tasks and realizing automated decision-making. This paper proposes an adaptive calculation and adjustment method for power flow in power systems based on LLMs. The reasoning capability of LLM is used to enhance the adaptability of strategy adjustment, and the power flow calculation module is invoked to verify the feasibility of strategies in real time, forming a closed-loop workflow with dynamic perception, autonomous decision-making, and continuous optimization. Retrieval augmented generation is introduced during the calculation and adjustment process to provide safe operation constraints and auxiliary decision-making support, realizing the full-process automation of power flow calculation and adjustment under complex constraints. Case studies on the IEEE 30-bus and IEEE 118-bus systems verify the application potential of LLM in the automated calculation and analysis of power systems.