BAYESIAN MODEL COMPARISON OF SOLAR RADIATION MODELS
In this paper, we propose a new statistical method: the Bayesian Model Comparison (BMC) method for selecting an adequate hourly diffuse fraction correlation. Six models are investigated and compared according to the BMC method. The selection of the best model is based on a Bayesian criterion called the Deviance Information Criterion (DIC). In this article, we demonstrate the usefulness of the DIC criterion in the model selection process and we issue a caution regarding the selection of a model with standard statistical methods. The aim of this paper is also to introduce the DIC to the solar radiation modeling community.
Philippe Lauret Carine Rivière
Laboratoire de Physique du Batiment et des Systèmes 15, Avenue René Cassin BP 97715 Saint-Denis Cedex
国际会议
2007世界太阳能大会(Proceedings of ISES Solar World Congress 2007)
北京
英文
2007-09-18（万方平台首次上网日期，不代表论文的发表时间）