Solar energy, which has always been at the forefront, has discovered numerous
uses in a variety of fields. One of the key targets of scientists and producers
in the twenty-first decade is sustainable solar energy collecting. The
maximization of solar energy is totally dependent on the radiation absorbed by
the photovoltaic panels. Radiation is observed using numerous equipment and
calculated using diverse methods. If the device is to be totally reliant on
solar energy, it must be calculated far ahead. It is difficult to work because
solar radiation is affected by various factors, including region as well as
seasonality. In forecast scenarios, Artificial Neural Networks (ANN) is a
popular approach among scientists. Therefore, this research provides a technique
for estimating solar radiation that makes use of back-propagation algorithms.
The data of 17 stations in Tamil Nadu, India, were acquired for analysis and
split into three clusters: training, validation, and testing. This research is
focused on nine input variables and one outcome variable. Solar radiation is
estimated via feed-forward back-propagation in this case. The presented approach
is ascertained for training techniques such as Levenberg Marquardt (LM),
Bayesian-Regularization (BR), as well as Scaled Conjugate Gradient (SCG). In all
three scenarios, the resulting statistical error values and regression values
prove the adequacy of the presented approach. The Root Mean Square Error (RMSE)
for the BR approach is found to be the lowest, with a value of 0.0013. Also,
from all statistical error and regression values, the training approach BR
produces the best value than LM and SCG training approaches.