Artificial neural network based approach compared with stochastic modelling for electrical load forecasting

Accurate load forecasting is very important for electric utilities in planning for new plants. Also it is very significant for the routine of maintaining, scheduling daily, electrical generation, and loads. The main mission for a forecaster is to study the behaviour of the collected historical data (which is called data mining), and determining the different patterns of the time series. In this study, emphasis was considered on short-term load forecasting which is important for real time operation and control of power systems. Artificial intelligence such as ANN, ANFIS and stochastic forecasting models were used in this paper. The performance of these models is dependent on the characteristics of electric loads and is based on the assumption that electric load patterns are basically invariant with time. Three different models were considered and a new stochastic model (called REGARIMA) was introduced and compared the ANN and ANFIS models. A comparison study is performed between the results of the three methods. The simulations show that they are very promising methods in short term forecasting techniques, which could be applied as well on wind speed forecasting.