Predicting fuel properties of biomass using thermogravimetry and multivariate data analysis

Simple and reliable characterization methods for determining fuel properties of biomass are needed for several different applications. This paper describes and demonstrates such a method combining thermogravimetric analysis with multivariate data analysis, based on the thermal decomposition behavior of the fuel. Materials used for the tests were milled samples of wood chips thermally pretreated under different conditions in a torrefaction pilot plant. The predictions using the multivariate model were compared to those from a conventional curve deconvolution approach. The multivariate approach showed better and more flexible performance, with error of prediction of 2.7% for Mass Yield prediction, compared to the reference method that resulted in 29.4% error. This multivariate method could handle samples pretreated under more severe conditions compared to the curve deconvolution methods. Elemental composition, heating value and volatile content were also predicted with even higher accuracies. The results highlight the usefulness of the method and also the importance of using calibration data of good quality.