Forecastingmethods are one of the most efficient available approaches to make managerialdecisions in various fields of science. Forecasting is a powerful approach inthe planning process, policy choices and economic performance. The accuracy offorecasting is an important factor affects the quality of decisions thatgenerally has a direct non-strict relationship with the decisions quality. Thisis the most important reason that why the endeavor for enhancement theforecasting accuracy has never been stopped in the literature. Electricity loadforecasting is one of the most challenging areas forecasting andimportant factors in the management of energy systems and economic performance.Determining the level of the electricity load is essential for precise planningand implementation of the necessary policies. For this reason electricity loadforecasting is important for financial and operational managers of electricitydistribution. The unique feature of the electricity which makes it moredifficult for forecasting in comparison with other commodities is theimpossibility of storing it in order to use in the future. In other words, theproduction and consumption of electricity should be taken simultaneously. Ithas caused to be created a high level of complexity and ambiguity inelectricity markets. Computational intelligence and soft computing approachesare among the most precise and useful approaches for modeling the complexityand uncertainty in data, respectively. In the literature, several hybrid modelshave been developed in order to simultaneously use unique advantages of thesemodels. However, iterative suboptimal meta-heuristic based models are alwaysused for combining in these models. In this thesis, a direct optimum parallelhybrid (DOPH) model is proposed based on multilayer perceptrons (MLP) neuralnetwork, Adaptive Network-based Fuzzy Inference System (ANFIS), and SeasonalAutoregressive Integrated Moving Average (SARIMA) in order to electricity loadforecasting. The main idea of the proposed model is to simultaneously useadvantages of these models in modeling complex and ambiguous systems in adirect and optimal structure. It can be theoretically demonstrated that theproposed model due to use the direct optimal structure, can achieve non-lessaccuracy than iterative suboptimal hybrid models, while its computational costsare significantly lower than those hybrid models. Empirical results indicatethat the proposed model can achieve more accurate results rather than itscomponent and some other seasonal hybrid models