Optimum Long-Term Electricity Price Forecasting in Noisy and Complex Environments

Abstract

This article presents an integrated algorithm consisting of artificial neural network (ANN), fuzzy linear regression (FLR), and conventional linear regression (CLR) models for optimum long-term electricity price forecasting. Three ANN models and seven well-known FLR models along with a CLR model are applied all together to provide a robust framework for electricity price forecasting. Analysis of variance for a randomized complete block design and Fisher Least Significant Difference test are performed to compare forecasting results obtained by the ANN, FLR, and CLR models. Results indicate that there is a significant difference between the performance of ANN and FLR models in terms of mean absolute percentage error. Besides, it is shown that the CLR and FLR models considerably outperform the ANN models in this case. The proposed algorithm can be easily used in uncertain and complex environments due to its flexibility and is suitable for the long-term forecasting of electricity price.