INTELLIGENT MODELLING FOR SOLAR ENERGY FORECASTING AND APPLICATIONS

Abstract

Due to the growing population of the world, there is a surge in the demand for energy – specifically element energy. The production of electricity contributes the largest share in the emission of greenhouse gases that emanate from the burning of fossil fuel. So, a need has arisen for a clean form of energy i.e. renewable energy that can contribute to the increasing demand for energy worldwide. For the accurate design and development of solar energy based systems, solar radiation resource data plays a significant role. Unfortunately, the measured data is rarely available for research purpose for these stations where measurement is already being done. Therefore, it is essential to develop modelling techniques that can forecast global solar energy for such locations where measurements have not been done with reasonable accuracy. The number of mathematical models has been developed for assessment of solar energy under cloudless skies. In fact, it is rather difficult to forecast accurately the behaviour of solar radiation by using these stochastic models, as they need the bases of the precise definition of problem domains and these stochastic models were found with relatively large errors and sometimes difficult to be adopted widely. These models developed, however, are not appropriate to forecast solar energy during cloudy sky. So, fuzzy logic based model have been developed in situations when deterministic or probabilistic models do not provide a realistic description of the phenomenon under study. A lot of research has been carried out and it is observed that for modelling complex systems involving large data sets, it is difficult to maintain accuracy with so much of data sets by employing fuzzy logic approach. So, an Artificial Neural Network (ANN) models were introduced, that employ artificial intelligence techniques and are data-driven which can subsequently simulate the structure. For complex function estimation, the prediction of a number of hidden layers and hidden neurons accurately using ANN is difficult as they are large in number. Also, the training time for the conventional neural network is too large, which results in a slower response of the system. The existing neural model performs only vii

the operation of summation of its weighted inputs; it does not perform the operation of products on its weighted inputs. Therefore, hybrid intelligent models have been introduced for forecasting solar energy which integrates the features of fuzzy logic approach and ANN. Further, the obtained results have been implemented for short-term power forecasting in solar photovoltaic (PV) system under composite climate zone.