A simulated-based genetic algorithm for the forecasting of monthly peak electricity demand

This paper is focused on the forecast of the monthly peak electricity demand using stochastic search processes that are the basis of genetic algorithms. Three different models, namely, quadratic, logarithmic and exponential, are developed based on the economic indicator GDP and the weather factors such as temperature, hours of sunshine and humidity. The models are applied to Mauritius which is chosen as an application country. The genetic algorithm is then implemented and the best coefficients with minimum root mean square error are obtained. Monthly data from January 2005 to December 2008 are considered for estimating the values of the coefficients. Each independent variable is estimated for the year 2009 using regression analysis and the models are tested for that year. Our results show that the quadratic equation yields a slightly better estimate of the future monthly peak electricity demand. We also forecast the monthly peak electricity demand up to the year 2014. Finally, we apply multiple regression to find the values of coefficients. We see that the values obtained by GA provides a better estimate for monthly peak electricity demand as compared to that found by multiple regression.