This paper was motivated by a practical optimization problem that appeared in electricity market of Mongolia. We consider the total average cost minimization problem of power companies of the Ulaanbaatar city. By solving an identification problem, we developed a fractional model that quite adequately represents the real data. The obtained problem turned out to be a fractional minimization problem over a box constraint, and to solve it, we propose a method that employs the global search theory for d.c. minimization.