关键词: 多能互补能源系统, 多智能体强化学习, 动态优化建模, 源网荷储协同, 可再生能源消纳, 协同调度

Abstract: In order to solve the problem of dynamic collaborative optimization of multi-energy complementary energy system under high proportion of renewable energy access,and the limitations of traditional centralized methods in multi-agent interest coordination and real-time response,dynamic optimization modeling research is carried out.A three-layer multi-agent reinforcement learning framework of “ physical layer-decision layer-collaborative layer ” is constructed,which divides energy producers,consumers and system dispatchers into independent agents.Based on the improved proximal strategy optimization algorithm,a dynamic reward function integrating economy,environmental protection and stability is designed,and distributed decision-making and global coordination are realized through centralized training-decentralized execution mechanism.A typical park-level multi-energy complementary system is selected as an example to verify the proposed model.The results show that the proposed model can increase the renewable energy consumption rate to 95.3 % and reduce the cost of electricity by 18.7 %.In the 50 % load mutation scenario,the system recovery and stabilization time is shortened to 90 seconds,which is 90 % less than the traditional mixed integer programming method. In the face of ±20% wind and solar prediction error,the load satisfaction rate remains 98.7 %.The dynamic optimization model can effectively solve the problem of multi-agent coordination and uncertainty adaptation of multi-energy complementary systems,and provide technical support for real-time optimal scheduling of high-permeability renewable energy systems.

Key words: Multi-energy complementary energy system, multi-agent reinforcement learning, dynamic optimization modeling, source network load storage coordination, renewable energy consumption, collaborative scheduling