A heuristic for the long-term electricity generation planning problem using the Bloom and Gallant formulation

Long-term power planning is a stochastic problem often confronted by electrical utilities in liberalized markets. One can model it for profit maximization—using market-price estimation functions for each interval—by posing it as a quadratic programming problem with some linear equalities and an exponential number of load-matching linear inequality constraints. In order to avoid handling all the inequalities when one is attempting to solve the problem, column generation methods have been employed herein. In this paper, we describe the foundations and implementation of a heuristic that tries to iteratively guess the active set of constraints at the optimizer, alongside a normal quadratic programming solution used at each iteration. The two methods are compared and the heuristic procedure is shown to be more efficient.