A review of validation methodologies and statistical performance indicators for modeled solar radiation data: Towards a better bankability of solar projects

In the context of the current rapid development of large-scale solar power projects, the accuracy of the modeled radiation datasets regularly used by many different interest groups is of the utmost importance. This process requires careful validation, normally against high-quality measurements. Some guidelines for a successful validation are reviewed here, not just from the standpoint of solar scientists but also of non-experts with limited knowledge of radiometry or solar radiation modeling. Hence, validation results and performance metrics are reported as comprehensively as possible. The relationship between a desirable lower uncertainty in solar radiation data, lower financial risks, and ultimately better bankability of large-scale solar projects is discussed. A description and discussion of the performance indicators that can or should be used in the radiation model validation studies are developed here. Whereas most indicators are summary statistics that attempt to synthesize the overall performance of a model with only one number, the practical interest of more elaborate metrics, particularly those derived from the Kolmogorov-Smirnov test, is discussed. Moreover, the important potential of visual indicators is also demonstrated. An example of application provides a complete performance analysis of the predictions of clear-sky direct normal irradiance obtained with six models of the literature at Tamanrasset, Algeria, where high-turbidity conditions are frequent. (C) 2014 Elsevier Ltd. All rights reserved.