An evolutionary programming solution to the unit commitment problem

The work was conducted with the aim of finding a general method for solving the unit commitment (UC) problem. The proposed algorithm employs the evolutionary programming (EP) technique in which populations of contending solutions are evolved through random changes, competition, and selection. In the subject algorithm an overall UC schedule is coded as a string of symbols and viewed as a candidate for reproduction. Initial populations of such candidates are randomly produced to form the basis of subsequent generations. The practical implementation of this procedure yielded satisfactory results when the EP-based algorithm was tested on a reported UC problem previously addressed by some existing techniques such as Lagrange relaxation (LR), dynamic programming (DP), and genetic algorithms (GAs). Numerical results for systems of up to 100 units are given and commented on.