A neuro-fuzzy algorithm for improved gas consumption forecasting with economic, environmental and IT/IS indicators

In energy sector, accurate prediction of long term gas consumption is very important for decision making and policy process. In addition, conventional approaches may not provide precise results. In this paper, an integrated forecasting algorithm based On Adaptive Neuro Fuzzy Inference System and Computer Simulation (ANFIS-CS) for long term gas consumption has been proposed. Standard input variables include different economic, environmental and IT/IS (number of internet users divided by population in each year) indicators, and the output variable is gas consumption. The concepts of post-processing and pre-processing are considered in the proposed method. At first the best distribution function is identified for each year and then CS is used to create random variables for each year to predict the effects of probabilistic distribution on annual gas consumption finally, data is fed into ANTIS model to find the network with the lowest mean absolute percentage error (MAPE). To show the quantitative benefits of the ANFIS-CS, 12 different structures of a well-known class of adaptive neural networks (ANNs), namely Multi-Layer Perceptron (MLP) as well as 10 different types of regression models are developed and the MAPE values of ANN-MLP models and regression models are compared with the MAPE of proposed model. The results of this comparison show the applicability and superiority of the proposed method. This is the first study that presents an integrated intelligent forecasting approach for accurate gas consumption considering the economic, environmental and IT/IS indicators. (C) 2015 Elsevier By. All rights reserved,