Long-Range Generation Planning Using Generalized Benders' Decomposition: Implementation and Experience

This paper describes experience in implementing and using a generalized Benders' decomposition model for planning electricity generating capacity expansion. The model divides the problem into a master linear program, which generates trial expansion plans, and a set of nonlinear subproblems, which compute production cost and system reliability for the trial plan. Modifications of the original model described in the paper include a more efficient method for the subproblem computations based on the Gram-Charlier representation of probability distributions, representation of multiple unit plants, computation of upper and lower bounds on the optimal cost, and inclusion of nonthermal generating technologies. The paper also describes computational experience with the model and comparison with a dynamic programming model of the same problem. The Appendix discusses the convexity properties of the model with the modifications introduced in the paper.