New Regression Model to Estimate Global Solar Radiation Using Artificial Neural Network

The main objective of the present study was to develop a new model for the solar radiation estimation in hilly areas of North India for the determination of constants ‘a’ and ‘b’ by taking only latitude and altitude of the place into consideration. In this study, new model was developed based on Angstrom-Prescott Model to estimate the monthly average daily global solar radiation only using sunshine duration data. The monthly average global solar radiation data of four different locations in North India was analyzed with the neural fitting tool (nftool) of neural network of MATLAB Version 7.11.0.584 (R2010b) with 32-bit (win 32). The neural network model was used with 10 hidden neurons. Eight months data was used to train the neural network. Two months data was used for the validation purpose and the remaining two months for the testing purpose. The new developed model estimated the values of ‘a’ which range from 0.209 to 0.222 and values of ‘b’ ranging from 0.253 to 0.407. The values of maximum percentage error (MPE) and mean bias error (MBE) were in good agreement with the actual values. Artificial neural network application showed that data was best fitted for the regression coefficient of 0.99558 with best validation performance of 0.85906 for Solan. This will help to advance the state of knowledge of global solar radiation to the point where it has applications in the estimation of monthly average daily global solar radiation.