A Logic-Based Mixed-Integer Nonlinear Programming Model to Solve Non-Convex and Non-Smooth Economic Dispatch Problems: An Accuracy Analysis (1810.02608v2)

Abstract: This paper presents a solver-friendly logic-based mixed-integer nonlinear programming model (LB-MINLP) to solve economic dispatch (ED) problems considering disjoint operating zones and valve-point effects. A simultaneous consideration of transmission losses and logical constraints in ED problems causes difficulties either in the linearization procedure, or in handling via heuristic-based approaches, and this may result in outcome violation. The non-smooth terms can make the situation even worse. On the other hand, non-convex nonlinear models with logical constraints are not solvable using the existing nonlinear commercial solvers. In order to explain and remedy these shortcomings, we proposed a novel recasting strategy to overcome the hurdle of solving such complicated problems with the aid of the existing nonlinear solvers. The proposed model can facilitate the pre-solving and probing techniques of the commercial solvers by recasting the logical constraints into the mixed-integer terms of the objective function. It consequently results in a higher accuracy of the model and better computational efficiency. The acquired results demonstrated that the LB-MINLP model, compared to the existing (heuristic-based and solver-based) models in the literature, can easily handle the non-smooth and nonlinear terms and achieve an optimal solution much faster and without any outcome violation.