A review and comparison of fuzzy regression models for energy consumption estimation

The objective of this study is to examine the most well known fuzzy regression approaches with respect to energy consumption estimation. Furthermore there is no clear cut as to which approach is superior for energy consumption estimation. This is quite important in developing countries such China and Iran severe fluctuation for energy consumption. Where classic regression approaches do not provide a suitable prediction. In the present study, monthly data for electricity consumption in Iran are studied from 1992 to 2004. For suitable anticipation of electricity demand fluctuations, sixteen fuzzy regression models are considered in this research. Each fuzzy regression model has different approach and advantages. Auto correlation function was applied for defining input data of each of these models. By using this technique a few combinations are considered for selecting the input of each model. After calculating each model, their outputs will be an estimated function of the rate of electricity consumption in Iran. For determining the rate of error of fuzzy regression models estimations, the rate of output of each model is compared with the actual rate of monthly electricity consumption in test data. Five types of errors are considered for each model. Also an analysis of variance and Duncanpsilas multiple range tests are performed to formally select the best fuzzy regression model. The results show that Peterpsilas model is out performs the other by considerable margin.