An Interval Optimization and Generalized Nash Bargaining Approach for Power Trading Among Multiple Virtual Power Plants

To address the impact of distributed source–load uncertainty on the optimization and power trading of virtual power plants (VPPs), this paper integrates interval optimization (IO) and generalized Nash bargaining (GNB) to develop a power trading model for multiple virtual power plants (MVPPs). The source–load uncertainty is represented by day‐ahead prediction intervals of photovoltaic (PV) generation and flexible loads. The bargaining power of each VPP is assessed to allocate cooperative profits. A case study demonstrates the scheduling, trading, and cost intervals of each VPP, showing that IO effectively captures power flow uncertainty between VPPs and the grid under source–load variability, while GNB ensures fair allocation of cooperative profits. By applying the Hurwicz criterion to compare IO and robust optimization (RO), it is observed that when the decision‐maker’s optimism level exceeds a critical threshold, IO yields less conservative scheduling solutions than RO. In addition, a comparison with stochastic optimization (SO) shows that while SO can achieve lower expected costs under accurate probabilistic information, IO provides a distribution‐free and computationally efficient alternative when such information is unavailable. Furthermore, a comparison of GNB, Nash bargaining (NB), and the Shapley value method for profit allocation among MVPP shows that GNB, unlike NB, distributes cooperative profits in proportion to each VPP’s contribution to the coalition. In addition, compared with the Shapley value method, GNB achieves significantly lower computational costs when a large number of VPPs participate in cooperative operation, thereby enhancing its engineering feasibility. The proposed model also exhibits superadditivity, indicating that larger alliances generate higher profits. These findings demonstrate that the integration of IO and GNB offers a computationally tractable and engineering‐feasible approach for MVPP coordination under source–load uncertainty.