Electricity Load Forecasting by Combining Adaptive Neuro-fuzzy Inference System and Seasonal Auto-Regressive Integrated Moving Average

Nowadays, electricity load forecasting, as one of the most important areas, plays a crucial role in the economic process. What separates electricity from other commodities is the impossibility of storing it on a large scale and cost-effective construction of new power generation and distribution plants. Also, the existence of seasonality, nonlinear complexity, and ambiguity pattern in electricity data set makes it more difficult to forecast by using the traditional methods. Therefore, new models, computational intelligence and soft computing tools and combining models are the most accurate and widely used methods for modeling the complexity and uncertainty in the data set. In this paper, a parallel optimal hybrid model using computational intelligence tools and soft computations is proposed to forecast the electricity load forecasting. The main idea of this model is the use of the advantages of the individual models in the modeling of complex systems in a structure and elimination of  the limitations of them, simultaneously. The experimental results indicate that the proposed hybrid model has a higher performance accuracy in comparison to iterative suboptimal hybrid models and its computational cost is lower than the other hybrid models; also, the proposed model can achieve more accurate results, as compared with its component and some other seasonal hybrid models.