dc.description.abstract | We consider a mandatory pool, based to the one established in the Greek electricity market, in which the unit commitment and the scheduling of energy and reserves are the solution of Day-Ahead Scheduling (DAS), an optimization problem that is solved daily and aims to minimize the system cost for the next day. The single-day horizon of DAS may be rather short for capturing the effects of the long start-up times and large commitment costs of slow-start lignite units; hence, the DAS solution may be myopic, resulting in higher total costs in the long-run. To tackle this problem, the Greek market uses a heuristic approach, in which the units' shut-down costs are replaced by their start-up costs and the start-up costs are suppressed; this facilitates the start-up and discourages the shutdown of slow-start units. To address and evaluate the myopic solution issue of DAS more rigorously, we extend the unit commitment problem to a longer horizon of several days, and keep only the solution for the next day as binding (rolling horizon). We call the resulting approach Medium-Term Unit Commitment (MTUC). We compare the long-run average performance of the MTUC output for different horizon lengths (2, 4 and 7 days) to that of the heuristic DAS approach used in the Greek market. The results show that MTUC brings in a small reduction in the total system cost. © 2011 IEEE. | en |