Survey on Machine Learning Approaches for Solar Irradiation Prediction

Renewable energy technologies are clean sources of energy that have a much lower environmental impact than conventional energy generation methods. Researches focusing on different energy generation techniques are gaining much importance worldwide, to manage exponential increase in the energy requirements. Solar energy is used in various applications like solar charged sensor nodes, solar charged vehicles, agriculture, electricity production etc. This solar energy can be harnessed using a range of technologies such as solar heating, solar photovoltaic cells, solar thermal electricity, solar architecture and artificial photosynthesis. The need for solar energy requires the estimation of solar energy production at various atmospherical conditions. This estimation involves the prediction of solar irradiation. Machine learning techniques based on Support Vector Machine (SVM), Neural Networks, Multilayer Perception(MLP), etc as well as Gaussian Process Regression method are normally applied for learning and predicting solar parameters. These models make use of parameters like air temperature, wind direction, relative humidity, and total rainfall as input to predict the temperature for a particular day. This paper highlights on the features of these different approaches for prediction and various metrics that are normally used for measuring the accuracy of the prediction process