基于深度强化学习的园区综合能源系统低碳经济调度  

Deep Reinforcement Learning-based Low-carbon Economic Dispatch of Park Integrated Energy System

作　　者：杨挺[1] 刘豪 王静 党兆帅 耿毅男 盆海波[1] YANG Ting;LIU Hao;WANG Jing;DANG Zhaoshuai;GENG Yinan;PEN Haibo(Key Laboratory of Smart Grid of Ministry of Education(Tianjin University),Nankai District,Tianjin 300072,China;Shenzhen Power Supply Co.,Ltd.,Shenzhen 518000,Guangdong Province,China)

机构地区：[1]智能电网教育部重点实验室(天津大学),天津市南开区300072 [2]深圳供电局有限公司,广东省深圳市518000

出　　处：《电网技术》2024年第9期3604-3613,I0022,I0023,共12页Power System Technology

基　　金：中国南方电网公司总部科技项目(090000KK52220020)。

摘　　要：为降低园区综合能源系统的运行成本和碳排放量,同时应对系统不确定性带来的随机波动,提出一种考虑阶梯式碳交易的园区综合能源系统低碳经济调度模型,并采用深度强化学习方法求解。首先构建园区阶梯式碳交易模型,将碳交易成本考虑在内对园区综合能源系统低碳经济调度问题进行数学描述;其次将该调度问题表述为马尔可夫决策过程框架,定义系统的观测状态、调度动作和奖励函数;继而采用近端策略优化算法进行低碳经济调度决策。所提方法无需进行负荷预测或不确定性建模,能够对源和荷的随机波动做出实时响应。最后基于多场景多算法进行算例仿真,结果表明所提方法提高系统运行经济性的同时降低了系统碳排放量。To reduce the park-integrated energy system's operating costs and carbon emissions while solving the random fluctuations caused by the system uncertainty,a low-carbon economic dispatch model of the park-integrated energy system considering ladder-type carbon trading is proposed and solved by the deep reinforcement learning method.Firstly,the ladder-type carbon trading model is proposed,and the low carbon economic dispatch problem of the park-integrated energy system is mathematically described by taking carbon trading costs into account;secondly,the dispatch problem is formulated as a Markov decision process framework,defining the observation state,dispatch action and reward function of the system;then the proximal policy optimization algorithm is used to make low carbon economic dispatch decisions.The proposed method does not need to predict load or model the uncertainty,and the network is trained to respond to the system state in real-time.Finally,the simulation is based on multiple scenarios and algorithms,and the results show that the proposed method improves the system operation economy while reducing the system's carbon emissions.

关 键 词：园区综合能源系统 阶梯式碳交易 深度强化学习 近端策略优化算法 低碳经济调度 

分 类 号：TM721[电气工程—电力系统及自动化]