ELECTRICITY FORECASTING FOR SMALL SCALE POWER SYSTEM USING ARTIFICIAL NEURAL NETWORK

This project presents a practical short term load forecasting (STLF) for
small scale power system using artificial neural network (ANN) method. The
project applies a generic three-layered feedforward network. The network is
trained in a supervised manner and used backpropagation as a learning teclmique.
In addition, a configuration consisting of a hidden layer that uses a hyperbolic
tangent sigmoid transfer function and the output layer with a pure linear transfer
function is applied. Gas District Cooling (GDC) is chosen as a case study for
small scale power system since this plant was designed to produce electrical
power supply and chilled water for Universiti Teknologi PETRONAS (UTP)
campus and in-plant use. As a sole customer of GDC power plant, the load data
from 2006 till 20 I 0 are gathered and utilized for model developments. There are
two models developed based on UTP normal operating semester (Semester On)
and during break (Semester Off). The developed models can forecast electricity
load for the one week ahead. The computation experimental of the proposed
network applies MATLAB software and its toolbox. The mean absolute
percentage error (MAPE) is used as the measurement for the forecasting
performance. At the end of this project, the proposed method using ANN manages
to get average MAPE of 6.72 % for Model I (Semester Off) and 3.92 % for
Model 2 (Semester On) which is considered relatively good result.