A Multiscale Framework for Predicting Biomass Gasification Tar Yields

Thesis Abstract

Biomass gasification tar formation poses a major barrier to efficient operation and scale-up of gasification systems, impacting tar cleanup costs, catalyst fouling, and overall energy efficiency. This study addresses the fragmented understanding of tar yields by developing a multiscale framework that integrates reaction kinetics, reactor-scale transport, and feedstock variability to predict tar yields with quantifiable uncertainty. The aim is to deliver a predictive framework capable of guiding process design and operation to minimize tar production across diverse biomass feedstocks and operating conditions. Specific objectives are to (1) synthesize a multiscale tar-yield model that links devolatilization and tar-forming reactions at the molecular level to macroscopic reactor performance; (2) calibrate and validate the framework against experimental data from pilot-scale gasification tests; (3) quantify the influence of feedstock composition, moisture, and equivalence ratio on tar yields; (4) assess the sensitivity and uncertainty of tar predictions using probabilistic methods; and (5) develop actionable guidelines for tar mitigation strategies, including process parameter optimization and catalyst integration. The methodology combines experimental and modeling components. A factorial pilot-scale gasifier (50 kWth) will be operated with wood-derived and agricultural residue feeds (n=6 feedstock types), each tested at three equivalence ratios and three bed temperatures, yielding 54 operational runs. Tar yields will be quantified using standardized elutriation-gas sampling followed by GC-MS and GC-FID analyses to identify and quantify primary tar compounds, complemented by off-line high-temperature molecular sieve adsorption to capture heavier tar fractions. Kinetic data for primary devolatilization and secondary tar-formation reactions will be obtained from thermogravimetric analysis (TGA) coupled with differential scanning calorimetry (DSC) and micro-reactor experiments, providing rate expressions for key tar precursors. The modeling framework integrates three scales (i) molecular-scale reaction kinetics for tar formation derived from density functional theory (DFT)–informed pathways, (ii) meso-scale kinetic Monte Carlo (kMC) simulations to capture tar evolution within devolatilization zones, and (iii) macro-scale computational fluid dynamics (CFD) with reduced-order models to simulate gasification reactor performance and heat/mass transfer effects. Parameter estimation will employ Bayesian inference and Markov chain Monte Carlo (MCMC) to calibrate rate constants against the experimental tar data, yielding posterior distributions and predictive intervals. Data analysis will proceed in stages. Descriptive statistics will summarize tar yield distributions across feedstocks and operating conditions. Multivariate regression and partial least squares (PLS) will relate measured tar yields to feedstock chemistry metrics (volatile matter, lignin, cellulose, extractives) and process variables. ANOVA will test the significance of feedstock type, temperature, and equivalence ratio on tar yields. The multiscale framework will be validated using k-fold cross-validation and compared to conventional single-scale models via root mean square error (RMSE) and R-squared metrics. Uncertainty quantification will be conducted through propagation of posterior parameter uncertainties to tar yield predictions, enabling probabilistic tar forecasts under new operating scenarios. A sensitivity analysis (Sobol indices) will identify dominant factors driving tar production. Expected findings include (i) a robust, quantitative linkage between feedstock chemistry and tar yields across scales, (ii) identification of critical process windows that minimize tar formation, and (iii) a validated, uncertainty-aware predictive tool suitable for pilot and industrial application. The study contributes to knowledge by providing a transparent, physics-informed multiscale model that bridges molecular-level tar chemistry with reactor-scale performance, addressing a persistent gap in predictive capability for biomass gasification tar. Practical implications include improved design guidelines for tar mitigation, optimized operating envelopes, and enhanced reliability of gas-cleanup systems. The main conclusion anticipates that integrative multiscale modeling, combined with targeted experimental validation, can reduce tar yields by a quantifiable margin across diverse biomass feeds, thereby lowering downstream processing costs and enabling more efficient gasification-based energy systems. Recommendations emphasize expanding the feedstock library, incorporating real-time sensor data for online calibration, and extending the framework to catalyst-assisted tar cracking to further improve gasifier performance.