基于MADDPG的多园区综合能源系统协同调度方法  

Collaborative Scheduling Method for Multi-park Integrated Energy System Based on MADDPG

作　　者：刘生虎 陈皓勇[1] Liu Shenghu;Chen Haoyong(South China University of Technology,Guangzhou Guangdong 510641,China)

机构地区：[1]华南理工大学,广东广州510641

出　　处：《电气自动化》2025年第1期109-112,共4页Electrical Automation

基　　金：国家重点研发计划项目(2022YFB2403500)。

摘　　要：多园区综合能源系统存在内部复杂的多能耦合关系。为解决系统内部能量互补协同优化困难的问题,建立一种基于数据驱动的多智能体深度确定性策略梯度算法,并以最小运行成本和减碳为目标构建电热气多园区综合能源系统协同调度架构。所提方法通过提取综合能源系统内历史状态信息对神经网络进行训练,网络数据固定后即可根据当前状态进行调度决策。仿真结果表明,所提方法对比非多智能体算法更能有效降低多园区综合能源系统运行成本,并对不同碳价格有良好适用性。There is a complex multi-energy coupling relationship within the multi-park integrated energy system.A data-driven multi-agent deep deterministic policy gradient algorithm was established to solve the problem of complementary and collaborative optimization of energy within the system.The collaborative scheduling architecture of the electric,heat and gas multi-park integrated energy system was constructed with the goal of minimizing operating costs and reducing carbon emissions.The proposed method trained the neural network by extracting historical state information from the integrated energy system.Once the network data was fixed,scheduling decisions can be made based on the current state.The simulation results show that the proposed method is more effective in reducing the operating costs of multi-park integrated energy systems compared to non-multi-agent algorithms,and has good applicability to different carbon prices.

关 键 词：综合能源系统 电热气联供 协同优化 深度强化学习 低碳经济运行 

分 类 号：TK01[动力工程及工程热物理]