A Review on Machine Learning Models for Solar Irradiation Prediction

In recent years, one of the most developed renewable energy sources, is the solar energy, along with wind power. In the last years more investment has been made in the development of these two technologies, due to the amount of places with high irradiation and wind possibilities. However, the intermittent nature of solar irradiation increases the need for more flexible and reliable energy generation. Hence, in order to fulfill the requirements of the grid, energy system operators use conventional technologies. Therefore, if a higher penetration of renewable generators is desired, a decrease of the unreliability factors must be achieved by the design of accurate forecasters. Forecasting renewable energies’ output power is not a new assignment. Several methods can be found throughout literature. Forecasting is based on predicting future figures within historical databases One application that has been grabbing attention is solar irradiation prediction using data driven models. Machine learning models can find it difficult to follow the pattern of solar irradiation owing to the fact that solar irradiation varies significantly and some may even become zero during nights. This discontinuity causes even more problems. Hence a two-fold approach has been used for solar irradiation prediction using the wavelet transform as a data processing tool for all the relevant parameters and subsequently utilizing the processed data to train a neural network. This paper presents a comprehensive review of the existing techniques in the domain.