Discourses on solar radiation modeling

All solar energy applications require readily available, site-oriented and long-term solar radiation data. A typical database comprises of global, direct and diffuse solar irradiance, duration of sunshine and complementary data like cloud cover, atmospheric turbidity, humidity, temperature, etc. However, most of these stations do not provide complete if any information on solar data, chiefly due to the capital and maintenance costs that measuring instruments incur. For instance, global radiation is the most frequently measured parameter, its two components, i.e. diffuse and direct irradiance are often not measured. Improvements have been made to the meteorological radiation model MRM which, had been developed by Muneer et al. as a simple broadband irradiance estimation model based on synoptic information, by incorporating the sunshine information in the model's regressions. The result of the improvement of the model is a considerable reduction in biases and scatter in the comparison between estimated and measured data. The improved meteorological radiation model, IMRM is more accurate, by up to 70% in some cases, than its predecessor in estimating, global and diffuse horizontal irradiance. When sunshine, atmospheric pressure and temperature are not measured by a nearby station, yet cloud information is recorded, radiation estimation models based on cloud cover, CRM, can be used. Three CRMs have been compared to newly proposed models. It was found that models with locally fitted coefficients gave a more accurate estimation of the solar radiation than CRMs with generalized coefficients. The newly proposed model performed better that the older generation models. The third section of the article deals with estimation of diffuse radiation and possible improvements in its modeling. In this section, apart from clearness index (kt), influence of the synoptic parameters of sunshine fraction (SF), cloud cover (CC) and air mass (m) on diffuse fraction of global radiation (k) is studied both qualitatively and quantitatively. It is found that, SF shows a strong bearing on the k-kt relationship followed by CC and then m. As a next step, a series of models are developed for k as a polynomial function of kt, SF, CC and m. After an extensive evaluation procedure, a regression model is selected such that the diffuse radiation can be estimated with reasonable accuracy without making the model overtly complex. It was found that among all the models, the composite model involving all parameters provides the most accurate estimation of diffuse radiation. The site-specific models are further investigated for any appreciable correlations between different locations and their possible attributions. It was also found that a single model could more than adequately estimate the diffuse radiation for the locations within a given region. Three optimum models are also recommended for each region, in view of the fact that information on all parameters is not necessarily available for all sites. This study reveals a significant improvement from the conventional k-kt regression models to the presently proposed models, therefore, leading to more accurate estimation of diffuse radiation by approximately 50%.