虚拟电厂自趋优负荷跟踪控制策略

Essentially, the virtual power plant (VPP) is an aggregation unit of distributed energy resources (DERs) based on internet of things, which promotes the coordination and optimization of electric power, grid, load and storage by the exploiting the flexibility of distributed power, energy storage and flexible load. This paper proposed a novel load tracking control strategy for VPP. Based on self-approaching optimization theory, the VPP could realize automatic tracking control of the given target load curve and achieve effective utilization of massive flexible resources. The content mainly included three parts: Optimality, approaching optimization and self-approaching optimization. To assess the "optimality" of VPP, two indicators of load tracking error and incentive cost were established. In terms of how to "approaching optimization", an event-driven stimulus-feedback control scheme was proposed. The event was triggered by the deviation of load tracking error, initiating a process to guides the independent decision-making of DERs, and driving the overall optimization of VPP. To ensure the process of approaching optimization was automatic and autonomous, the response rules that match the stimulus-feedback scheme was built, and an end-to-end automatic response model of DERs was established to achieve the consistency between individual decisions and overall goals. The deep reinforcement learning algorithms was introduced to optimize the decision-making process of DERs. The numerical results show the proposed strategy can quickly assess the potential of VPP in complex environment, and accurately track the target load curve within its capability range. Moreover, it can effectively drive VPP approach to the theoretical optimal operation point, while the individual approach to the optimality as well. The paper provides a feasible scheme of how VPPs participate in the economic dispatch of power system.