Llm helps deep reinforcement learning improve real time power flow safety

LLM-Enhanced Deep Reinforcement Learning with Human-in-the-Loop Calibration for Real-Time Optimal Power Flow

Artificial Intelligence

Summary

Managing electricity flow in real time is tricky because the system changes a lot and has many safety rules that are hard to explain with numbers alone. The authors propose using a large language model (LLM) to add helpful feedback in plain language to the learning process, combined with expert human checks to catch mistakes. This makes the AI better at making safe and reliable decisions about electricity flow. Tests on standard power networks show the method works better and more stably than traditional AI training.

What this means in practice

  • For power grid operators: Enhance decision-making for real-time power distribution while ensuring safety through language-guided training and expert oversight.
  • For energy management teams: Improve training of automated control policies that integrate operational preferences expressed in language and calibrated by human experts.

Authors

Jiayang Tu, Bozhen Jiang, Bojun Du, Yunxiao Ling, Ershun Du, Qin Wang

Abstract

Real-time optimal power flow (RT-OPF) requires fast and reliable decision-making under time-varying operating conditions. However, many practical operating preferences and safety related requirements are difficult to encode using conventional numerical rewards alone, which limits the effectiveness and controllability of deep reinforcement learning (DRL) for RT-OPF. To address this issue, this paper proposes an LLM-enhanced DRL framework with human-in-the-loop (HITL) for RT-OPF. The proposed method augments the physics-based reward with sparse language evaluative feedback during offline training, so that semantic operational requirements can be incorporated into policy learning. To improve reliability in RT-OPF, a selective HITL calibration mechanism is further introduced to inspect and correct questionable LLM judgments before they continuously shape the learning process. Case studies on the IEEE 30-bus and IEEE 118-bus systems show that LLM feedback improves policy safety over vanilla TD3, while further HITL calibration achieves the best overall performance and more stable training behavior. These results demonstrate the effectiveness of combining language guidance and expert calibration for RT-OPF.