Calendar-based short-term forecasting of daily average electricity demand
MetadataShow full item record
CitationKumru, M., & Kumru, P. Y. (2013). Calendar-based short-term forecasting of daily average electricity demand. In 2015 International Conference on Industrial Engineering and Operations Management (pp. 1-5) Piscataway, NJ: IEEE. https://dx.doi.org/10.1109/IEOM.2015.7093940
Short-term electricity demand forecast becomes more and more important due to recent deregulation of electricity market in Turkey. It is affected mainly from several factors that are working days, weekends, feasts, festivals, and temperatures. In the study, contribution of these factors to the consumption is to be analyzed and modeled with nonlinear (quadratic) regression models. First, the variation in Turkey's daily electricity consumption for the years of 2012 and 2013 is determined with respect to actual weather temperatures and calendar events. Then, nonlinear regression models are constructed separately for the demand function of weekends, weekdays, public holidays (feast and festivals), and for the total daily averages. The models are tested on the actual calendar data of the year 2014 for its twelve months period. Mean absolute percentage errors are calculated and compared for each of the regression models. The results indicate that the calendar-based short-term forecasting model slightly outperforms the noncalendar-based forecasting model, and seems more reliable in forecasting the short-term electricity demand.