(37c) Kinetic Studies on Pyrolysis of Invasive Reed Canary with a Combined Scheme of Parallel-Reaction Kinetic Model and Multi-Layer Artificial Neural Network Model

In this work, a new combined scheme is proposed to investigate pyrolysis kinetics on biomass samples of Invasive Reed Canary to evaluate the potentials as a feedstock to the generation of biofuels. In the combined scheme, a kinetic model is coupled with a multi-layer artificial neural network (ML-ANN) model to predict detailed profiles of biomass pyrolysis at 10, 20, 30, and 40 K/min. A 15-step parallel-reactions mechanism is developed and utilized in the kinetic model. The optimal values of pre-exponential factors (A) and activation energies (Ea) at each reaction step are achieved by an optimization model. The iso-conversional method is applied to calculate initial values of kinetic parameters in the optimization model, and multi-objective functions are defined to minimize the difference between model predictions and experimental data. The kinetic model is validated with a selection of data points at 4 heating rates and then is applied to predict the major profile of biomass pyrolysis. A ML-ANN model is developed to provide the details on biomass pyrolysis. The Bayesian optimization is applied to optimize the hyperparameters of the ML-ANN model. Finally, the kinetic model combined with the deep neural network model is validated with 9559 data points and is applied to predict full profiles of biomass pyrolysis at 4 heating rates.