Operation and expansion strategies for virtual power plants participating in electricity maarkets

This doctoral thesis tackles the optimal operation and planning strategies of virtual power plants (VPPs) that participate in electricity markets. An accurate model is provided for the components of the VPPs, which comprise conventional and renew- able generating units, storage facilities, and demands. Moreover, the uncertainty related to the components of the VPP, i.e., renewable generation levels, production costs for the conventional generating units, and demand consumption levels, as well as the uncertainty associated with electricity market prices are both modeled. Two models are developed for the optimal operation of VPPs that participate in different electricity markets. One of them addresses the offering strategy in the day-ahead (DA) energy market of a VPP that also participates in the real-time energy market to compensate for the power deviations. In this problem, the uncertainty related to renewable power production levels and market prices is modeled using confidence bounds and scenarios, respectively. The other operational model deals with the DA self-scheduling problem of a VPP trading in both energy and reserve electricity markets. As a major novelty, uncertain requests for reserve deployment are modeled using intervals. Moreover, the uncertainty in renewable power generation levels and market prices is modeled using confidence bounds and scenarios, respectively. Both models are based on a stochastic adjustable robust optimization (ARO) approach. This dissertation also addresses the optimal expansion planning of VPPs that participate in the electricity market via both a stochastic model and an ARO model. Uncertain parameters considered in these expansion planning models are future production costs of conventional generating units, future consumption levels of the flexible demands, and future market prices. In the stochastic approach, uncertainty sources are modeled using a set of scenarios, whereas in the ARO approach, uncertain parameters are characterized by confidence bounds