The University of Hong Kong Shenzhen Institute of Research and Innovation, Shenzhen 518063, China

The virtual power plant (VPP) aggregates massive distributed resources to form a large-scale market entity interacting with the power grid. However, the energy management of large-scale distributed resources faces a contradiction between global optimization and local demands. Centralized optimization and regulation struggle to balance electricity costs and comfort of the users, while local user energy management lacks coordinated control over the overall operational state of the VPP. Therefore, a lightweight user energy management strategy for cloud-edge collaborative VPPs is proposed. Firstly, a hierarchical decision-making framework for energy management of VPPs is constructed to shift unidirectional control to bidirectional interaction. Secondly, considering the limited computational resources of low-cost edge devices, a local decision-making algorithm based on sparse reinforcement learning is proposed, which uses adaptive topology evolution to reduce the number of trainable parameters in neural networks. Finally, case studies conducted on real-world datasets validate the effectiveness and efficiency of the proposed lightweight strategy.