Intraday Optimal Dispatch Method for Power Systems Considering Source‐Load Uncertainty Under Extreme Stagnant and Calm Weather

ABSTRACT

Under high‐penetration renewable energy integration, extreme weather events pose significant risks and challenges to power systems. Extreme static weather, as a special meteorological event, easily creates extreme scenarios of sustained low renewable output, making traditional mathematical optimisation algorithms fail to meet real‐time requirements for intraday optimal dispatch, severely affecting secure and stable operation. To address this risk, this paper proposes an intraday optimal dispatch method based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. First, a Markov Decision Process is employed to simulate the sequential decision‐making of intraday dispatch and construct a complete model framework. Second, a TD3 algorithm integrating actor and critic networks is designed to guide adaptive updates of the dispatch agent. Third, the TD3‐based intra‐day dispatch model is built and trained via multiscenario simulation to enhance generalisation capability. Finally, the effectiveness is validated on the IEEE 24‐bus system. Compared with the Deep Deterministic Policy Gradient (DDPG) and Soft Actor‐Critic (SAC) algorithms, the proposed TD3 method reduces the average load shedding by 11.5%–14.8% and achieves a single‐dispatch computation time of 0.02 s, which is over 500 × faster than the Model Predictive Control (MPC) baseline. The system operating cost is controlled within $247,900–$248,400 across all test scenarios, with zero power flow limit violations.