ADVANCES IN ALTERNATIVE FUEL CONTENT MODELING USING NEAR-INFRARED SPECTROSCOPY

摘要

Partial least squares (PLS) regression models can be constructed from near-infrared (NIR) spectroscopic data to predict critical fuel specification properties. Implementing such models into the data handling algorithms of small, field-portable NIR equipment would allow for fast in-situ identification of unknown fuels and assessment of their suitability for a given application. However, the presence of alternative fuels in fuel blends can affect the most fundamental compositional aspects of NIR spectroscopic data and, consequently, limit the applicability of PLS models constructed from petroleum fuels to predict the fuel properties of these blends. To compensate for this, specialized PLS models have been developed to identify and predict the amounts of alternative fuels blended with a petroleum fuel. The results from these alternative fuel models can be used then to derive correction factors to correct the predictions made by the corresponding PLS fuel property models. Several alternative fuel types have been incorporated into a generalized content and property modeling strategy that will be discussed during the course of this presentation. The developed strategy provides the means to allow petroleum fuel based NIR PLS models to predict and quantify alternative fuel contents in blends in a robust fashion that, due to at least some fuel-independent predictive capabilities, better accommodates a future of unknown and unknowable fuel types.