Comparative analysis of hourly load forecast for a small load area

Accurate load forecasting plays a key role in economical use of energy and real time security analysis of system. In this paper a practical case of small load area of a town getting supplied by nineteen distribution feeders is considered. Four months exhibiting different daily load-curve variation pattern are selected. Graphical analysis of the daily load curves for a week in each month is performed. Also statistical data analysis of hourly load data for each month is conducted. Artificial Neural Networks (ANN) is used for hourly forecasting. Input vector is designed which includes the historical load data, minimum and maximum temperature data as vector elements. Artificial Neural Network models are trained for each month using Back-Propagation algorithm with Momentum learning rule. For the selected months the network performances are evaluated using the mean absolute percentage error (MAPE) criterion. The variation in forecasting ability of ANN for different months is also discussed.