Development of Statistical Models for Predicting Solar Radiation: A Case Study in Guwahati City, India

Environmental degradation and gradual depletion of fossil fuels have dramatically enhanced necessity to shift towards an alternative source of energy. In response to this, utilisation solar energy has emerged as a promising solution with various advantages over conventional sources of energy which are diminishing and polluting. Historically, correlations have been made and used to assess the solar potential of a location as measurement of solar radiation at every location is not yet viable due to the high cost of instruments and geographical constraints. Hence, statistical models for predicting solar radiations become very useful. In this work, Angstrom-Prescott model and its modified forms for the Guwahati city, India were developed to estimate the monthly global solar radiation. Error estimation tools like mean bias error ( MBE) , root mean square ( RMSE) , mean absolute percentage error ( MAPE) and R 2 were used to indicate the model performance. All the error indicators showed very low error values for the newly developed models compared to few of the already existing models. A R 2 value of 0.96 was achieved for the linear Angstrom-Prescott equation. Statistical analysis showed that the newly developed model is suitable to predict monthly global solar radiation for the Guwahati city.