Ahmad Al-Subhi and Chokri Belhadj Ahmad
This paper presents Short Term Load Forecasting (STLF) model using mode-regression based technique for a residential area in Yanbu Industrial City (YIC) in the Kingdom of Saudi Arabia (KSA). Hourly load, temperature and humidity data are collected for three consecutive years from 2009 to 2011. In this technique, all the days with similar characteristics are classified into groups called modes. The daily load behavior is classified based on identified behavioral modes representing religious, social, and official occasions, in addition to environmental conditions. The daily load signal is decomposed into time-varying and non-time varying components. Each component is forecasted individually. The first model uses harmonics regression analysis to forecast the time-varying component which is only the daily load curve with zero mean value. To forecast the non-time varying component, regression analysis using Eureqa software is used to forecast the average load consumption for the next day. In the end, the two models are added algebraically to constitute the next day load forecasting model. The obtained model formulation testing has shown satisfactory forecasting results. A comparative study is conducted to prove the effectiveness of the model proposed .The results obtained in this work are compared with other published work that uses different method applied to the same data.