Journal of Applied Mathematics & Bioinformatics

Time-series modeling and prediction of weather-driven system-level electrical load – Case of Abu Dhabi

• [ Download ]
• Times downloaded: 10900

• Abstract

   

  Load forecasting has long been used in operations and planning of the electric power system. In this study, weather variables were used for modeling and prediction of the system-level electrical load of the city of Abu Dhabi, UAE. A Transfer Function (TF) model was developed and its accuracy was compared to that of an Autoregressive Integrated Moving Average (ARIMA) model. We also tested an Artificial Neural Network (ANN) model based on the same weather variables that were used in the TF model. Assuming perfect knowledge of the weather variables over the forecasting horizon, the TF model was more accurate for forecast horizons of up to 48 hours. The ANN model, on the other hand, was more accurate for one-week ahead forecasts. Assuming imperfect knowledge of the weather variables (i.e., they are not known over the forecasting horizon and have to be forecasted first), the TF model was more accurate than the ANN model in all cases. Average accuracy of the best TF method does not exceed 1.5% for 24-hour horizon, 2.5% for 48-hour horizon and 4% for 168-hour horizon. With the added uncertainty of forecasted weather drivers, the accuracy of the proposed method degrades only slightly, while the ANN model is much less robust and becomes unusable beyond a two-day horizon.