Implementation of a cloud-based solar radiation model using Regression Analysis for estimating photovoltaic power generation

The accurate estimation of irradiance and PV electricity output necessitate costly equipment that need regular maintenance, thereby inflating the capital cost of investment of solar PV system. In this paper, a cloud-based regression model is implemented to estimate irradiance and electricity parameters using cloud cover data. Testing the model in a site having sky regimes characterized as clear sky with scattered clouds revealed the accurate performance of the model. Daily estimations of irradiance were performed to acceptable levels of accuracies, with root mean square error of 90.2 W/m
²
and correlation coefficient value of 0.77. Even higher accuracies were reported for monthly mean estimations of irradiance, with root mean square error of 60.9 W/m
²
and correlation coefficient value of 0.94. Using the monthly mean irradiance estimates, the PV electricity generation output was estimated and compared to simulated performance on NREL PV Watts for a 20 kW DC system at Universite des Mascareignes. Relatively high levels of accuracies have been reported for the Black model with root mean square error of about 149 Wh and correlation of 0.92.