Fast initialization methods for the nonconvex economic dispatch problem

This paper develops a set of fast initialization methods to generate candidate preliminary points in the search space of the non-convex economic dispatch problem. These initial points are either the global optimal solution or close enough from this solution to clearly facilitate and accelerate the process of solving the problem while increasing the probability of attaining the global optimal solution. The proposed methods can approach the global optimal solution in minimal time irrespective of the size of the system. In addition, a two-stage framework is also proposed to accommodate the proposed initialization methods. In the first stage, initial solutions are generated by the proposed initialization methods and in the second stage, any powerful stochastic solver can be utilized to confirm obtaining the global optimal solution. The proposed framework is flexible with respect to treating the physical constraints and the practical features of the problem such as the valve point effects, prohibited operating zones, and multiple fuel options. To generate initial solutions for specific variants of the problem, three fast initialization methods are proposed, and to generate initial solutions considering several practical features simultaneously, an integrating strategy is developed. The interior-point method implemented in MATLAB is employed to solve the approximated convex economic dispatch problems incorporated within the proposed initialization techniques. Several powerful metaheuristic algorithms and benchmark problems have been simulated to demonstrate the effectiveness of the proposed initialization methods, the generic applicability feature of them, and to evaluate the closeness degree from the global optimal solution. The results demonstrate that the proposed initialization methods are capable of generating high-quality solutions in a highly computational efficient manner. (C) 2020 Elsevier Ltd. All rights reserved.