Solar radiation Prediction using fuzzy logic and neural networks

Abstract

Solar energy is the main renewable clean energy in the world which is deployed as the source of all other types of energy on earth, since it can directly or indirectly be transformed into different forms of energies. Iran having approximately 300 clear sunny days in a year is categorized among the countries with high solar energy potentials. Taking the geographical location of Iran into consideration, using solar energy seems to be vitally important. Photovoltaic and solar thermal are the main sources of electricity generation from solar energy in comparison to other models of energy transmission. Large-scale PV power plants have been built using photovoltaic (PV) all over the world. However, the weather conditions play key role in performance of PV. In other word, due to the high variability of ambient temperature, the power output of PV plants is stochastic. This paper attempt to predict the solar radiation using a method based on the fuzzy neural networks of the third kind. It should be noted that prediction is very useful in generating solar data. All data is collected from the Urmia Meteorological datum plane (NDP). This paper proposes a new solar radiation forecast method using neural networks and fuzzy logic. The proposed method aims to achieve efficient accuracy at different conditions.
In our proposed method, Fuzzy logic and neural networks are combined. Due to the fact that the third type of the fuzzy neural networks owns a higher accuracy for estimation of linear models, the current study has used this very type of fuzzy neural network. The gained results shows that with the rise of the hidden neurons and learning steps, the final output of the network becomes less different than the actual value with this in mind that a large number of neurons were not manipulated.