Medium term electricity demand forecasting of Bursa

In this study, medium term electricity demand forecasting of Bursa region is done with linear, quadratic and exponential curve fitting methods for the months of 2007 by using monthly electricity consumption data between 2002-2006 years. When we examine the results, we see that the best model for the past data is quadratic fitting. By the way, the best forecasting results are achieved with linear regression for the year of 2007. The mean absolute percent error rates are lower than 7 % for the all three methods.The additional study is done about the artificial neural networks. Again, the yearly total electricity consumptions are analyzed with the above three methods. And the time series neural network approach, that provides flexible forecast by establishing nonlinear relationship between inputs and outputs, is used to predict the annual total electricity consumption of the years 2007 and 2008. The absolute percent error rate in the prediction is 3.36 % for the year of 2007 which is very promising.