Hybrid optimization for multi-condition kinetic modeling of biomass pyrolysis with improved numerical stability

Accurate kinetic modeling of biomass pyrolysis is essential for understanding thermochemical conversion processes and supporting energy-system design. However, parameter estimation remains challenging because of nonlinear reaction kinetics, competing pathways, and variability in experimental conditions. This study presents a hybrid optimization framework for multi-condition kinetic modeling of biomass pyrolysis, combining global search using genetic algorithms with local gradient-based refinement. A parallel–competitive Arrhenius-based kinetic model represents lignocellulosic biomass decomposition into gas, tar, and char. The model was calibrated using experimental datasets for three biomass feedstocks across multiple temperatures and holding times, allowing the estimated parameters to represent the investigated operating ranges. Reliable model evaluation during optimization was supported by a stiff ODE solver, adaptive tolerance settings, non-negativity constraints, and zero-time protection. The hybrid approach consistently outperformed the standalone genetic algorithm, reducing RMSE by 38.17–74.77% and SSE by 61.76–93.63% across the three feedstocks. Multi-condition calibration constrains parameter estimation using information from several experimental conditions rather than individual operating points. Although the model captures trends in product distribution, limitations remain in representing complex behavior such as tar stabilization at high temperatures. Overall, the framework provides an approach to biomass pyrolysis kinetic modeling through global and local optimization with numerical implementation.