Adaptive Neuro-Fuzzy Inference System vs. Regression based approaches for annual electricity load forecasting

Electricity demand forecasting is known as one of the most important challenges in managing supply and demand of electricity and has been studied from different views. Electrical load forecast might be performed over different time intervals of short, medium and long term. Various techniques have been proposed for short term, medium term or long term load forecasting. In this study we employ Adaptive Neuro-Fuzzy Inference System (ANFIS) and regression (Linear and Log-Linear) approaches for forecasting Iran's annual electricity load. We use feature selection technique for selecting most influential factors out of twenty socio-economic and energy-economic factors, and present a model that is affected by four economical parameters which are Nonoil Real-GDP, Population, Wholesale Price Index and Energy Intensity. Using Real-GDP instead of nominal-GDP can provide more accuracy because the effects of inflation are considered in the structure of such model and this will cause the results to be more reliable. To improve forecasting accuracy of the model we apply data preprocessing techniques. Forecasting capability of each approach is evaluated by calculating three separate statistical evaluations of the root mean square error (RMSE), the mean absolute error (MAE), and the mean absolute percentage error (MAPE). All evaluations indicate that the accuracy of ANFIS which is trained with preprocessed data is remarkably better than the other two conventional approaches.