AbstractLoad forecasting plays an important role in load management, optimisation and control of a power station. An Artificial Neural Network was used to provide short term load forecasting at a small remote hybrid power station supplying Djilkminggan Aboriginal community near Elsey Station, Northern Territory. Commercial software on Neural Networks was purchased and used to produce a short term load forecast.
Data was collected and carefully prepared. A pre-processing procedure in selection of training data is presented.
An Artificial Neural Network was developed using the Back Propagation method. Networks which forecast the next half hour and the next hour load were formed and examined. Predicted load was compared with the actual load to determine the accuracy of the forecast. An attempt to minimise the error of forecast data was made by examining several network architectures. The selected network successfully produced a short term load forecast with minimum error of 8.34%. An analysis of several aspects of forming a short term load forecasting using different artificial neural network architectures is included. The forecast will partly solve the problem of efficient operation of a small remote hybrid power station.
|Date of Award||Jun 1996|
|Supervisor||Dean Patterson (Supervisor)|
Load forecasting at Djilkminggan Power Station using an artificial neural network
Soeparwo, S. (Author). Jun 1996
Student thesis: Masters by Research - CDU