The economic load dispatch (ELD) problem is known as a hard optimization problem in the field of power system planning and control. Evolutionary algorithms (EAs) have demonstrated high performance in solving ELD problems with the number of power generator up to 40, but they lose their efficiency with high dimensionalities. In this study, we have proposed and have estimated the performance of a selective hyperheuristic for the online synthesis of an optimization algorithm based on large-scale optimization approaches. In the paper, the experimental results for benchmark problems and a real-world ELD problem are presented and discussed.