A Novel Method to Predict Solar Radiation Using Machine Learning Algorithms

In this work, an effort has been made to make the regression algorithms efficient and accurate by using the Ensemble Learning Approach. Data for this work has been collected from the weather station at MNNIT Allahabad, Prayagraj, U.P. [25.4358° N, 81.8463° E], and all the results are compared on the basis of the R2 score performance matrix. In Ensemble Learning Approach, different models are used in layers and with the help of bagging and voting methods; their result is enhanced to 94% R2 score in testing accuracy. The present model (predicted value 173 W/m2K as on 05/02/23 at 9 a.m.) shows good agreement with the real-time measurement of solar radiation (actual value 168 W/m2K as on 05/02/23 at 9 a.m.) with an error within ±3%. The model developed can be used by researchers to get the required data for their research and also plan for experimentation accordingly. The grid operators and power companies in and around Prayagraj can make use of this prediction model to plan for the proper scheduling of the different power generating techniques in order to minimize the instability in the grid.