Power system cloud edge multi-scale cooperative scheduling method and system based on multi-agent reinforcement learning

Abstract

The invention discloses a multi-agent reinforcement learning-based power system cloud edge multi-scale cooperative scheduling method and system. The method comprises the steps of obtaining a real-time multi-agent reinforcement learning model comprising a day-ahead agent model, an intra-day agent model and a real-time agent model in response to a scheduling demand of an edge side, and sending the real-time multi-agent reinforcement learning model to the edge side; acquiring first real-time state data of the power system on the edge side to determine a day-ahead predicted output value corresponding to each adjustable power supply unit in combination with the day-ahead agent model; correcting the day-ahead predicted output value according to the second real-time state data of the power system on the edge side and the intra-day intelligent agent model to obtain an intra-day corrected output value; and acquiring state data of the power system at the edge side, adjusting the intra-day correction output value by combining with the real-time agent model to obtain a real-time output value corresponding to the adjustable power supply unit, and performing scheduling control on each adjustable power supply unit according to the real-time output value to improve the flexibility and accuracy of scheduling.

Images