A comparative study for optimum short-term forecasting of electricity price with uncertainty

This paper presents an integrated algorithm consisting of artificial neural network (ANN), fuzzy linear regression (FLR) and conventional linear regression (CLR) for optimum short-term electricity price forecasting. Eight ANN models and seven well-known FLR models as well as CLR model are considered simultaneously to provide a robust framework for electricity price forecasting. Thus, it can be easily applied to uncertain and complex environments due to its flexibility. The input data are divided into train and test parts to test the results for further times. Data sets are compared with ANN, fuzzy regression (FR) and conventional regression (CR) techniques in a comprehensive empirical process, with respect to mean absolute percent error measure. It is shown that ANN leads to the most efficient outputs, although, sometimes FR models could yield suitable results. Then, the procedure is repeated for normalised data, and recent results are obtained again, which puts a validation on what was concluded before. At last, the research has been practiced in Iran and proved to be efficient, as the circumstances dominating the society do not follow a steady trend. This is the first study that investigates intelligent models for short-term electricity price forecasting with uncertainty.