American V-King Scientific Publishing
In this paper, application of genetic algorithm is discussed in order to improve the response of MLP neural network method for forecasting consumed load peak. In this method genetic algorithm is used which has more power in finding total minimum in comparison with algorithms which are based on Newton. This method could find consumed load peak of west of Iran more precise than other models with MRE% (Mean Relative Error Percent) equal 1.207. Several models of ANN with a hidden layer have been checked and the best of them was chosen as the best structure with 3 neutrons.