Load forecasting for remote area power supply systems

An artificial neural net approach is applied to short-term load forecasting in three remote area power supply systems (RAPS) in Western Australia. Such systems are usually in the kW range and are characterised by very irregular load profiles which make prediction difficult. The data used was collect...

Full description

Saved in:
Bibliographic Details
Published inProceedings the 11th Conference on Artificial Intelligence for Applications pp. 231 - 237
Main Authors Cheok, K., Kottathra, K., Pryor, T.L., Cole, G.R.
Format Conference Proceeding
LanguageEnglish
Published IEEE Comput. Soc. Press 1995
Subjects
Artificial neural networks
Australia
Backpropagation
Digital filters
Error analysis
Frequency
Load forecasting
Passband
Power supplies
White noise
Online AccessGet full text

Cover

More Information
Summary:An artificial neural net approach is applied to short-term load forecasting in three remote area power supply systems (RAPS) in Western Australia. Such systems are usually in the kW range and are characterised by very irregular load profiles which make prediction difficult. The data used was collected over a year in W.A. and evaluated for each season. The feedforward backpropagation network outperformed the statistical techniques and mean absolute percentage errors between 3.9% and 13.5% were obtained. Digital filters were used to decompose the load into low and high frequency passbands in a hypothetical case where known data is used to determine the learning abilities of the ANN. An upper limit on the accuracies of between 3.2% and 9.8% was achieved in this case. However an error analysis of the residuals shows that these have not yet been reduced to white noise indicating that further improvements are still possible.< >
ISBN:0818670703
9780818670701
DOI:10.1109/CAIA.1995.378818