Rolling-window Multi-stage Stochastic Programming for a Virtual Power Plant in the Real-time Market

Considering the uncertainty of real-time market prices, a rolling-window multi-stage stochastic programming scheme is proposed to solve the dispatch problem of a virtual power plant (VPP), aiming to minimize its operating cost. On this basis, the decision-making process of the VPP is modeled as a continuous sequence procedure and the influence of uncertainty information that would be revealed in the future can be taken into account at the current time interval. Meanwhile, to better reflect the reality, the nonanticipativity constraints are introduced to guarantee that the decisions at the current time slot should be made only relies on the information already observed without the hindsight of future realizations of uncertainties. Experimental results show the proposed scheme could simulate the influence of future uncertain information on scheduling policy by determining operation strategies under different scenarios, and also specify the unique dispatch strategy of the VPP in a rolling continuously way based on nonanticipativity constraints.