Random Forest (RF) with Daubechies Wavelet and Multiple Time Lags (MTL) for Solar Irradiance Forecasting

Accurate forecasting of solar irradiance is important as the dependency of this clean energy towards weather condition may affect the efficiency of solar power plant. Various types of models has been used to forecast solar irradiance and maintain the efficiency of solar power grid but not many can provide a high accuracy forecasting. In this paper, raw information of solar irradiance, current, temperature and power are collected as input for the random forest (RF) techique. These data went through a noise elimination process before it can use as input data for training and testing procedures. Daubechies wavelet based decomposition concept has been used in filtering the data from any unwanted noise as it may increase the percentage of error. The feature extraction of multiple time lags (MTL) is used to further improve the contents of input data. The proposed method used for solar irradiance forecasting is very much needed to reduce the forecasting error and useful for maintaining the stability of generated energy and energy consumption.