Forecasting of Solar Irradiance using Machine Learning Techniques

In modern power system application the forecasting of incident solar irradiance plays very important role for energy management optimally for operation the power grid efficiently. To forecast the output power of solar Photo Voltaic based electricity generation system the prediction of solar irradiance is very helpful. This forecast of solar irradiance is conducted by different techniques under physical models based strategy combined to cloud imagery and algorithm based approach using machine learning (ML) methods to develop prediction models associated with artificial intelligence (AI). This article is giving details of the implementation and execution of a strategy for solar irradiation forecasting developed using ML. Results analysis using proposed method is compared among ML based methodologies on regression model, decision tree, boosting and bagging for performing solar irradiance prediction applications. The performance measurements are challenging due to complexity of data set diversity, forecasting horizon, time step, performance indicators, and setup. The Linear regression gave minimum RMSE, MSE and prediction speed as $0.80359,0.64575$ and 21000 observations per second. Lowest MAE and training time is obtained for decision tree as 0.58321 and 5.3601 sec. The results presented here are highly efficient and helps to improve prediction performance by minimizing error and enhancing speed by the proposed novel approach.