| Summary: | Short term load forecasting (STLF) and very short term load forecasting (VSTLF) play an important role in economy running of power system. It is a key part of Supervisory Control and Data Acquisition (SCADA). So improving the forecast accuracy of short term load forecasting is very crucial in day-to-day operation. Many methods of load forecasting have been tried. Artificial neural network (ANN) is one of them. Because ANN has advantages in nonlinear prediction, more and more ANN research work applied to electric load forecasting were carried out in recent years. In this report, two back-propagation(BP) networks were designed to forecast integrated load. One BP network was designed to forecast load every 15 minutes. All models were trained using the historical load data supplied by the Energy Market Company (EMC) of Singapore. Holiday load was also considered in this report. Through limited testing conducted in the lab, the average absolute error (MAPE) for a 24-hour ahead forecast using the actual load is shown to be 3.22% for Mondays through Sundays and 5.00% for holiday load forecasting. The average MAPE for a 15-minute ahead forecast using the actual load is shown to be 0.194% for Mondays through Sundays and 0.226% for holiday load forecasting. It is evident from the findings that the BP-network is suited for STLF and VSTLF applications.
|
|---|