Electricity load forecasting using hybrid models based on Multi-Layer Perceptrons Neural Network and Seasonal Auto-Regressive Integrated Moving Average models

Nowadays, saving time and economy of each country requires proper planning, decision making, and rational forecasts in different areas. One of the most well-known areas that has received a lot of attention is electricity forecasting. The features of the electricity which makes it distinguished from other commodities are the impossibility of storing it and the existence of seasonality and nonlinear and ambiguity pattern in electricity data set. These features of the electricity makes it more difficult to forecast using traditional methods. Therefore, in this paper, a parallel optimal hybrid model using seasonal linear and nonlinear methods is proposed to forecast the electricity load forecasting. The main idea of this model is the use of the advantages of individual models in the modeling of complex systems in a structure, simultaneously. Experimental results indicate that in this method due to the use of a direct weighting method, the computational cost of modeling it is significantly lower than other parallel hybrid methods.