Mehrdad Beykverdi - Department of Electrical Engineering Pardis Branch, Islamic Azad University Pardis, Tehran, IRAN
Ahmad Ashouri - Department of Electrical Engineering Khodabandeh Branch, Islamic Azad University Khodabandeh, Zanjan, IRAN

This investigation presents a new method for solving and formulating the unit commitment problem that results in a considerable reduction in the number of decision variables. The scheduling variables are coded as integers representing the operation periods of generating units. The unit commitment problem is solved using a new hybrid method with the name of Artificial Neural Networks (ANN) and Imperialistic Competitive Algorithm or Nero-ICA hybrid procedure. This method provides solutions to the major demerits of Nero-ICA such as type of Neural Network, parameter tuning, selection of objective function size and problem dependent penalty functions. The constrained optimization problem is solved using adaptive penalty function approach. This paper describes the proposed method with the new variable formulation and presents test results on 10 units test systems. The results demonstrate the robustness of the new method in solving the unit commitment problem.

Artificial Neural Networks, Nero-ICA, Imperialistic Competitive Algorithm, Penalty function, Unit commitment, Operation cost