多智能体深度强化学习的分布式园区综合能源系统经济调度策略

Economic Scheduling Strategy for Integrated Energy System in Distributed Parks Based on Multi-agent Deep Reinforcement Learning

为解决综合能源系统内不同种类能源间复杂的耦合关系,以及综合能源系统内的多能互补与经济运行存在困难这一问题,本文提出一种以最优经济运行为目标的分布式园区综合能源系统优化调度架构,并采用基于数据驱动的多智能体深度Q网络对其进行求解。所提方法通过深度强化学习中的神经网络读取综合能源系统中的信息,不需要对模型内存在的复杂耦合关系进行建模,训练后的各智能体仅依赖本地园区数据即可实现对用户负荷需求的实时响应。仿真结果表明,所提方法在促进各园区内光伏消纳的同时,提高了综合能源系统运行的经济性。

To solve the problem that the coupling relationship between different types of energy in an integrated energy system(IES)is complex and there exist multi-energy complementarity and difficulty in the economic operation of the IES,an optimal scheduling architecture for the IES in distributed parks is proposed in this paper,which aims at the optimal economic operation and is further solved by using a data-driven multi-agent deep Q network(DQN).The proposed method reads the information in the IES through the neural network in deep reinforcement learning,and there is no need to model the complex coupling relationship within the model.The trained agents can only rely on the data of the local park to realize the real-time response to users’load demand.Finally,simulation results verify that the proposed method can not only promote the photovoltaic absorption in each park,but also improve the operation economy of the IES.