A smart predict-and-optimize framework for microgrid's bidding strategy in a day-ahead electricity market

This paper deals with the microgrid's bidding strategy (MGBS) problem in a day-ahead (DA) electricity market. To this end, the DA electricity prices, demand, and renewable energy uncertainties are modeled by different scenarios via generating several representative daily curves. However, the real-time (RT) electricity prices are modeled using robust optimization (RO) via predicting appropriate intervals. The existing paradigm to predict the RT intervals is to minimize the statistical errors, while this accuracy-oriented method may not necessarily yield the best MGBS plan against the actual realization of RT prices. The novelty of this research is to present a smart predict-and-optimize (SPO) methodology in which a cost-oriented prediction model replaces the existing accuracy-oriented methods. Trilevel mathematical programming is established to construct a cost-oriented model where the first level maximizes/minimizes the actual profit/cost by adjusting the predicted interval bounds of the RT prices. Knowing these trained interval bounds, the second level runs a scenario-based DA-MGBS. Then, a rescheduling problem is solved at the third level based on the actual realization of uncertainties. To solve this trilevel problem, a reformulation-and-decomposition (R&D) algorithm is utilized. The numerical experiment illustrates the advantages of the SPO method in terms of the microgrid (MG) cost.