Evaluation of different diffuse radiation models for Indian stations and predicting the best fit model

In the present study, the non-linear solar radiation models for predicting the monthly average daily diffuse radiation are developed using the measured data on global radiation, diffuse radiation and sunshine hours for 12 locations of India. Statistical method is used to derive these correlations. The developed models are employed to estimate the monthly average daily diffuse radiation. The performance of these correlations is compared with existing model. Accuracy of developed relationships is also tested using statistical indicators viz. Percentage error (PE), root mean square error (RMSE), mean percentage error (MPE) and mean bias error (MBE). The study finds that these statistical parameters have very low values for the proposed models. A cubic correlation of diffuse coefficient with percent possible sunshine gives the best fit. The maximum values of RMSE, MPE and MBE for the proposed third order equation are 4.33%, 8.68% and -1.25% respectively while in the case of existing model these values are 13.28%, 13.39% and -3.83% respectively. Hence, it is possible to apply the cubic equation for the prediction of monthly mean daily diffuse radiation. (C) 2011 Elsevier Ltd. All rights reserved.