Electrical Energy Consumption Estimation by Genetic Algorithm and Analysis of Variance

This study presents a genetic algorithm (GA) with variable parameters to forecast electricity demand in agricultural, low energy consuming and energy intensive sectors using stochastic procedures. The economic indicators used in this paper are price, value added, number of customers and consumption in the last periods for agricultural and low energy consuming sectors and price, value added, number of customers, price of the substitute fuel and energy intensity in energy intensive sector. Three kinds of models; linearlogarithmic, exponential and quadratic are used to find which leads us to minimum error for the related sector. The GA applied in this study has been tuned for all its parameters and the best coefficients with minimum error are identified, while all parameter values are tested concurrently. The estimation errors of genetic algorithm models are less than that of estimated by regression method. Finally, analysis of variance (ANOVA) is applied to compare genetic algorithm ( three models), regression and actual data. It is found that at α = 0.05 the five treatments are not equal and therefore Duncan test is applied to see which treatment pair has lead to the rejection of null hypothesis. Furtherer more it is shown that genetic algorithm estimation is closer to actual data with less MAPE (Mean Absolute Percentage Error) error than that of estimated by regression. The data from 1979 to 2003 is used to forecast electricity consumption in the aforementioned sectors as the case study.