Collaborative optimization method for scalable prosumers’ participation in frequency regulation ancillary services

As the demand for frequency regulation resources in power systems increases, collaborative optimization of flexible resources with rapid frequency regulation response capabilities, particularly by enabling scalable prosumers in local areas to participate in the frequency regulation ancillary service market, can effectively enhance safety, stability, and frequency regulation ability of power system. Therefore, this paper first establishes a collaborative optimization framework for scalable prosumers in frequency regulation and describes the operation model of prosumers. Considering the uncertainties that can impact prosumers' power decisions during frequency regulation, a scenario-augmented dataset generation method based on a denoising diffusion probabilistic model is proposed to improve decision adaptability under extreme scenarios with insufficient regulation capabilities. Additionally, to enhance the scalability and applicability of the training method in scalable prosumer collaborative optimization scenarios, a multi-agent attention proximal policy optimization algorithm combined with a global attention mechanism is introduced. The effectiveness of the proposed method in improving decision timeliness, operation benefits, scalability, and policy adaptability during scalable prosumers’ participation in frequency regulation ancillary services is validated using the IEEE standard node test system under various scales and scenarios.