Solar Irradiance Prediction by a New Forecast Engine Composed Wavelet Packet Transform and Adaptive Neuro-Fuzzy Inference System.

A novel approach, combining wavelet packet transform and adaptive neuro-fuzzy inference system is proposed in this study for solar irradiance short term prediction. The use of wavelet techniques is to overcome the discontinuities and a non periodicity in the change on the data and to increase the accuracy of time series prediction. However, the original time series data are decomposed into number of wavelet coefficient signals then used as an input vectors to ANFIS. The outputs from the ANFIS are recombined using the same wavelet technique to predict solar irradiance. Solar informationas a time series from a realworld case study of Nevada desert is used for model development. The results obtained with the proposed model, showed that the modified mean absolute error, the root mean square error and the daily peak error in solar irradiance short term prediction with values0.425, 1.823and 0.0034 by comparison is less than ANFIS and Wavelet-ANFIS model.