Development and Evaluation of Deep Learning Models for Predicting Instantaneous Mass Flow Rates of Biomass Fast Pyrolysis in Bubbling Fluidized Beds

Computational fluid dynamics (CFD) has evolved into avital toolfor advancing bubbling fluidized-bed reactors for biomass fast pyrolysis.However, due to the enormous computational burden of CFD simulations,optimizing working parameters over a broad range or simulating large/industrialunits is still extremely time-consuming. Because deep learning (DL)is a promising method to attain both precision and speed, two newDL models, which added an attention mechanism or a convolutional neuronnetwork (CNN) layer in the basic long short-term memory (LSTM) model,were established to predict instantaneous mass flow rates of majorspecies for biomass fast pyrolysis in a bubbling fluidized bed. Historicalmass flow rates from a multifluid model (MFM) simulation were consideredas the time series of data for the model training process. Influencingfactors, including sequence length, learning rate, convolutional kerneland stride sizes in the CNN layer, and number of neurons and layersin LSTM module, were examined to improve forecasting ability. Theresults demonstrated that the hybrid model including both CNN andLSTM outperforms other models in predicting instantaneous mass flowrates of biomass fast pyrolysis in bubbling fluidized beds.

Automated summary

Computational fluid dynamics (CFD) has become a vital tool for advancing biomass fast pyrolysis in bubbling fluidized-bed reactors. However, the computational burden of CFD simulations makes it time-consuming to optimize working parameters or simulate large-scale units. To overcome this, two new deep learning (DL) models, incorporating an attention mechanism or a convolutional neural network (CNN) layer, were developed to predict mass flow rates in biomass fast pyrolysis.