Designing an optimized model to forecast short-term electricity demand based on ARIMA and Wavelet decomposition neural network: Composition of linear and non-linear model (A case study in Iran)

In this paper we designed a new model that it integrate the linear and no linear models also we surveyed the trend electricity daily demand of country and the effective factors on daily demand of this energy carrier. Next, we have forecasted the electricity daily demand for next 10 days in forms of “step to step” by models of ARIMA, Feed forward Artificial Neural Network, and Neural Network-Wavelet Transform and Proposal model. Finally, the quantities have been forecasted by each one of models tested by criterions of forecast accuracy. Results present that proposal model has fewer forecasting error and high accuracy in forecasting of electricity daily demand of country. Its behind, the Neural Network-Wavelet Transform, Feed forward Artificial Neural Network and ARIMA lie respectively in further preferences.