Risk-Averse Optimal Scheduling of Virtual Power Plants Considering PV-CCS Fluctuation Coupling

Abstract

This article proposes a two-stage risk-averse scheduling model for an industrial-park virtual power plant in which carbon capture and storage (CCS) is coordinated with photovoltaic (PV) generation under restricted tie-line correction capability. Unlike formulations that treat CCS only as an emission-reduction device, the proposed model separates flue-gas capture, solvent regeneration, and rich-solvent inventory, thereby representing CCS parasitic power as a controllable load without confusing solvent-side energy shifting with geological carbon storage. Day-ahead unit schedules and tie-line nominations are common to all scenarios, whereas real-time grid deviations, PV curtailment, and CCS operation are scenario-dependent recourse decisions. Five time-correlated PV scenarios are obtained by multi-start clustering of 212 daily forecast-error trajectories derived from measured rooftop PV data; all 92 held-out days are used for out-of-sample assessment. The objective combines expected and worst-retained-scenario operating costs, with an explicit day-ahead and real-time settlement model. Compared with stochastic expectation minimization, the proposed setting reduces the in-sample worst-retained-scenario cost from 85,155 to 77,734 CNY at a 0.68% increase in expected cost. Flexible CCS with solvent storage reduces expected cost and PV curtailment by 9.99% and 34.38%, respectively, relative to fixed CCS, while eliminating 125.99 kWh of expected unserved energy. Held-out mean costs of the proposed and CVaR schedules are statistically indistinguishable; the remaining comparisons demonstrate the economic and low-carbon value of PV-CCS fluctuation coupling without claiming universal tail protection.