Comparison of fuzzy time series, ANN and wavelet techniques for short term load forecasting

Abstract

The present article presents the load forecasting for a power system (substation) load demands using techniques based on fuzzy time series (FTS), artificial neural network (ANN), and wavelet transform (WT). The mean absolute percentage error (MAPE), integral absolute error (IAE), integral of time multiplied error (ITAE), integral square error (ISE) along with integral time multiplied square error (ITSE) criteria are used for determining the performance indices and minimizing the error. From the investigations of the results obtained in the study, it is inferred that forecasting of electric load based on WT and ANN offers less error as compared to FTS. The suggested integrated model captures the useful properties of artificial neural networks and wavelet transforms in time series and is found to be accurate for real-time data.