As a new type of regional energy management system, the virtual power plant (VPP) can efficiently participate in the secondary frequency regulation auxiliary services of the power grid through the coordinated optimal scheduling of "source-load-storage". This paper introduces the internal structure of the VPP, and models and analyzes the characteristics of new energy units and controllable loads. A two-stage scheduling model for the VPP participating in secondary frequency regulation is established, which can balance the net profit and frequency regulation effect of secondary frequency regulation. An improved quantum particle swarm optimization (QPSO) algorithm with adaptive weights is studied. By introducing an adaptive weighting mechanism, the weight parameters are dynamically adjusted during the quantum particle update process to improve the search ability and convergence speed of the algorithm. The improved algorithm is applied to the two-stage optimization process, enabling the VPP to achieve higher net profits from secondary frequency regulation and better frequency regulation effects. Simulation results demonstrate that the proposed improved algorithm has a faster convergence speed and stronger global optimization ability.