Mohammad Hossein Pourkazeni - School of Economics and Political Sciences, University of Shahid Beheshti, Tehran, Iran.
Roya Aghaeifar - School of Economics and Political Sciences, University of Shahid Beheshti, Tehran, Iran.

Electricity cannot be stored and needs huge amount of capital so producers and consumers pay special attention to predict electricity consumption. Besides, time-series data of the electricity market are chaotic and complicated. Nonlinear methods such as Neural Networks have shown better performance for predicting such kind of data. We also need to analyze other variables affecting electricity consumption so as to estimate their quantitative effects. This paper presents a new approach for forecasting: a combined method of Neural Networks (ANN) and econometrics methods which can also explain the effect of rising electricity prices on consumption after the Subsidies Reform Plan. Data is from ۱۹۸۸-۲۰۰۸, and the method is compared with Neural Network and ARIMA based on the RMSE performance function that shows the advantage of the combined approach. The provident prediction is done for ۲۰۰۹- ۲۰۱۴ and indicated that after decreasing subsidy, electricity consumption would increase slightly until ۲۰۱۴.

Keywords: forecasting, electricity consumption, neural network, ARDL model, ARIMA method