Short-term load forecasting using time series and data mining analysis : a comparison between FFANN and sARIMA

الملخص EN

In a restructured power market, the accurate forecasting of load demand has drawn attention of power researchers.

Electricity demand is a complex signal due to its non-linear, non-stationary and time variant behavior.

In this paper, two models are proposed, namely, “Feed Forward Artificial Neural Network (FFANN)” and “Seasonal Autoregressive Integrated Moving Average (sARIMA)” to forecast load demand.

Here the data of electricity market of Victoria, Australia, in year 2010 are used into two time series case studies.

To achieve an accurate forecasting, additional to load data, temperature, humidity and wind speed data, which have important impacts on load level have been considered.

The results shows that FFANN model can predict electricity demand more precisely than sARIMA model.