Solar radiation prediction based on random forest of feature-extraction

Abstract

In order to use solar energy efficiently, the accurate prediction of solar radiation is extremely important. The aim of this paper is to discuss and build the solar radiation prediction model, and an efficient method based on improving random forest is proposed in here. The random forest regression model is created for predicting solar radiation. The nine variables are taken as input of the original model, and the solar radiation as the output. The original data is normalized using Min-Max method and is mapped to the values in the range of zero to one. The large amount of radiation data has been dealt with univariate feature extraction. The classifier of the random forest tree is constructed by characteristic factors that can influence solar radiation, and the optimal parameters of the model are selected by the OOB error analysis. The relationship between M-try and OOB can be obtained by computing of the R platform. Hence, a prediction model of improved random forest solar radiation based on feature extraction has been described. Under the specific area of solar radiation forecast, the studies have been done, and the solar radiation is analyzed and compared with the different prediction methods. The simulation results prove the validity of the improved method; the root mean square error is reduced and the precision of prediction is improved. It is significant for the prediction of the amount of solar radiation under complex environment and the effective use of photovoltaic power generation.