Predictive Model of Solar Irradiance Using Artificial Intelligence: An Indian Subcontinent Case Study

2. Literature Survey

Benghanem (2012) talked about the prediction model's importance to predict the average of long-term solar radiation (global). But this had its own limitations - bearing these parameters was only possible in selective regions as the overall cost equipment is very high. A method to tackle this problem is by using appropriate empirical relations through the measured parameters at selective locations where appropriate equipment is available. Other parameters like - latitude, relative humidity, longitude, sunshine hours, temperature, relative humidity can be used to find the global solar radiation using different empirical models developed by many researchers.

The classification of the empirical models was done by Besharat et. al. (2013) after they reviewed the available models. The categories of models created by them are Sunshine based, Temperature based and Cloud-based. Ennaceri et al. (2018) investigated potential sites for concentrated solar power installations in Morocco based on direct normal irradiation. A sunshine based model (empirical) was proposed by Quansah et al. (2014). Radiation Model for Solar applications was proposed by Xie et al. (2016). Seven empirical models were used for the calculation process by exploring the Angstrom-Prescott model and the Hargreaves-Samani model. Experiments showed that the Angstrom-Prescott model underestimated global solar radiation for few months (April, May, June, October, and November) and overestimated global solar radiation for few months (December, August, and September). All these models were not suitable for the long-term measurement of solar radiation because the input data were site-specific. Because of all these reasons, most researchers made an attempt to make use of meteorological parameters for Global Solar Radiation prediction. The most common prediction approach which is followed recently is the artificial intelligence (AI) for solar radiation prediction. Sfetsos and Coonick (2000) presented an approach for the forecasting of solar radiation by numerous artificial intelligence based techniques.