Using artificial neural network model with time series monthly average electricity demand in Kurdistan region

Abstract EN

Forecasting in the time series is one of the important subjects in statistical science.

It helps the government to make a good plan in order to get an accurate decision in future.

This research deals with methods for forecasting which is artificial neural networks models; specifically MultiLayer Networks as adopted back propagation algorithm feedback and Box-Jenkins model (seasonal multiplicative model).

It has been applied on the series of time for the electricity demand in Kurdistan Region for the period between (2011– 2015).

We compared both results in models based on their prediction accuracy.

It was found that SARIMA (1, 1, 0) (0, 1, 1) 12 gives the better result than the neural network model based on statistical criteria (RMSE, MAE, MAPE), used for comparison.

Finally, using the best model to forecast the electricity monthly average demand in the Kurdistan Region for a period (12 months in 2016).