Data-driven worst conditional value at risk energy management model of energy station

The extensive integration of photovoltaics (PVs) leads to great uncertainties in the energy management of energy station (ES). In this context, this paper proposes a novel data-driven worst conditional value at risk (DWCVaR) energy management model for evaluating the operational risk brought by the uncertainty of PV outputs. Firstly, the wait-and-see energy management model is formulated to minimize the daily operational cost of ES. Then, the objective function is reformulated by the worst conditional value at risk (WCVaR) with the discrete probability distribution (PD). Furthermore, unlike the traditional WCVaR method with a box-like uncertainty set, the PD uncertainty set of PV outputs is described by the integrated norm (e.g., 1-norm and ∞-norm) constraints to cover possible PDs of PV outputs. Specifically, the integrated norm uncertainty set enables the explicit expression of the impact of historical data amount on the uncertainty set. Meanwhile, confidence levels of the uncertainty set can be adjusted for different conservative requirements. After the linearization of norm constraints, a bi-level primal-only iterative algorithm is proposed to solve the “min-max” optimization model. Finally, the whole DWCVaR energy management model is solved by the CPLEX solver. Case studies and comparative analysis are performed based on a representative ES. Numerical results show that the proposed DWCVaR energy management model is capable of achieving a reasonable and acceptable solution, and it significantly outperforms other methods.