Skip to main content
ARS Home » Northeast Area » Wyndmoor, Pennsylvania » Eastern Regional Research Center » Sustainable Biofuels and Co-products Research » Research » Publications at this Location » Publication #320235

Research Project: Farm-Scale Pyrolysis Biorefining

Location: Sustainable Biofuels and Co-products Research

Title: Prediction of properties and elemental composition of biomass pyrolysis oils by NMR and partial least squares analysis

item Strahan, Gary
item Mullen, Charles
item Boateng, Akwasi

Submitted to: Energy and Fuels
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/4/2015
Publication Date: 1/1/2016
Citation: Strahan, G.D., Mullen, C.A., Boateng, A.A. 2016. Prediction of properties and elemental composition of biomass pyrolysis oils by NMR and partial least squares analysis. Energy and Fuels. 30:423-433.

Interpretive Summary: Biomass or non-food plant material (e.g. crop residues, grasses, forest residues) is the most abundant renewable source available for the production of fuels and chemicals. It can be converted to bio-oil through a thermochemical conversion process known as fast pyrolysis, the heating of organic matter in the absence of oxygen, to a liquid called bio-oil. Bio-oil can be refined into gasoline and diesel fuels or used as a feedstock to produce commodity or specialty chemicals. However, bio-oils are highly complex mixtures which vary with biomass source and the particular process used to produce them making their complete analysis a challenge for analytical chemists. Currently bio-oils are usually characterized using numerous different techniques and instruments. In this work we have performed these analyses and also nuclear magnetic resonance (NMR) spectroscopy on more than 70 bio-oil samples and related materials. Chemists normally can use NMR analysis to determine structures of molecules, but with mixtures as complex as bio-oils only basic information can be gained by manual interpretation. However, much more detailed information is contained within the spectra. Here we report a model that uses a statistical method, partial least squares (PLS) analysis, to predict several properties of the bio-oils that normally require separate analyses, including its mass fraction of carbon, hydrogen, nitrogen, and oxygen, the samples heating (calorific value), their total acid number (TAN) and the concentration of phenols they contain. The accuracy of the model varies among the properties, but the model can continue to be improved via addition of more sample data. This can allow for more rapid nondestructive analysis of bio-oils, and can also be used to test the performance of an ongoing pyrolysis operation. This information will be valuable to those producing or utilizing pyrolysis bio-oils.

Technical Abstract: Several partial least squares (PLS) models were created correlating various properties and chemical composition measurements with the 1H and 13C NMR spectra of 73 different of pyrolysis bio-oil samples from various biomass sources (crude and intermediate products), finished oils and small molecule standards. Two models based exclusively on 13C-NMR data demonstrated the best over-all ability to predict these same characteristics for an unknown sample. The bio-oils were generated from feedstocks such as energy crops, woods, animal wastes and oil seed presscakes using various treatment protocols. The models also include a variety of standard small molecule samples, biodiesel, and fossil fuels (gasoline and diesel). The intensities of the NMR spectra from these samples were binned and subjected to partial least squares analysis to correlate them to their fractional mass content of H, C, O, and N, as well as to their values for higher heating value (HHV), total acid number (TAN), and total concentration of phenol and cresols. The first two latent variables of the PLS models based on 13C spectra are qualitatively similar to the equivalent first two principle components reported previously (Strahan et. al. 2011), but their chemical properties and compositions are now explicitly included. The R2 and RMSE of the predicted values are discussed in detail, and are acceptable for most biofuel-related research, and may be useful for wider applications as well.