Day Ahead Hybrid Forecasting of Global Horizontal Irradiance using Machine Learning (Random Forest Algorithm) and Time-Series Model (SARIMAX)

Abstract

This paper aims to propose and analyze a method to combine the Machine learning model with the Time-series model for hybrid forecasting of Global Horizontal Irradiance (GHI). This hybrid model exploits the performance of the Time-series model and Machine learning model, which perform differently on a different set of weather conditions, to give a more accurate result. For this research, Random Forest has been used as a machine learning model, and for the Time-series model, Seasonal Autoregressive Integrated Moving Average with exogenous regressors (SARIMAX) model has been used. The machine learning model considers weather conditions such as humidity, cloud cover temp, etc., to predict GHI. The time series model only depends on past data values, which makes it independent of weather conditions. A hybrid forecast tends to exploit the advantages of both models and overcome limitations. The final estimates from the Hybrid model contain the weight of each model, which is calculated during the validation period using a regression algorithm.