Prediction of instantaneous yield of bio-oil in fluidized biomass pyrolysis using long short-term memory network based on computational fluid dynamics data

Bio-oil (or tar) production by the fast pyrolysis of biomass is one of the important technological paths for efficient utilization and conversion of biomass. However, there are many types of biomasses with different compositions. There is an urgent need to select the biomass type that is more suitable for fast pyrolysis to produce bio-oil. In this work, the fast pyrolysis characteristics of three main components of biomass (cellulose, hemicellulose, and lignin) in a fluidized bed reactor were studied by establishing a comprehensive mathematical model of multiphase flow-heat transfer-pyrolysis reaction in the fast pyrolysis process of biomass in a fluidized bed reactor using a numerical simulation method. The fast pyrolysis characteristics of two typical biomasses such as hardwood and softwood with different contents of three main components were further analyzed. The results show that the yield of bio-oil or tar is 3% higher and the yield of biochar is 6% lower for the fast pyrolysis of hardwood compared to softwood at 773 K. Simulation results demonstrated that hardwood with high hemicellulose content and low lignin content is an ideal feed for bio-oil production by fast pyrolysis.

Section: Resultsmentioning

confidence: 99%

“…and reactor design. Zhong et al 21 proposed a method by coupling CFD and machine learning (ML) in predicting biomass pyrolysis, and their simulation method is favorable in handling large amounts of CFD results.…”

Section: Introductionmentioning

confidence: 99%

Numerical Simulation on Effects of Biomass Type on Its Fast Pyrolysis in Fluidized Bed Reactor

Shi

Liu

Zhang

et al. 2023

“…Machine learning (ML) has recently emerged as a critical technique to investigate multiphase flows and reactors owing to its fast advancement of computational theory and capability, − and shows great potential in speed up CFD modeling of biomass fast pyrolysis in fluidized-bed reactors. − Zhong et al designed a reduced-order model based on a back-propagation network, which trained CFD data from multifluid model (MFM) simulations. With significantly less calculation effort but still high accuracy, time-averaged species distributions in the reactor at various temperatures were forecasted.…”

Section: Introductionmentioning

confidence: 99%

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

Zhong

Wei

et al. 2023

Self Cite

Computational fluid dynamics (CFD) has evolved into a vital tool for 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/industrial units is still extremely time-consuming. Because deep learning (DL) is a promising method to attain both precision and speed, two new DL models, which added an attention mechanism or a convolutional neuron network (CNN) layer in the basic long short-term memory (LSTM) model, were established to predict instantaneous mass flow rates of major species for biomass fast pyrolysis in a bubbling fluidized bed. Historical mass flow rates from a multifluid model (MFM) simulation were considered as the time series of data for the model training process. Influencing factors, including sequence length, learning rate, convolutional kernel and stride sizes in the CNN layer, and number of neurons and layers in LSTM module, were examined to improve forecasting ability. The results demonstrated that the hybrid model including both CNN and LSTM outperforms other models in predicting instantaneous mass flow rates of biomass fast pyrolysis in bubbling fluidized beds.

“…18−28 Hanbin Zhong et al employed CFD to simulate the instantaneous yield of bio-oil in fluidized biomass pyrolysis, and further optimized their simulation by using long short-term memory network methods based on CFD data, thereby significantly reducing computational time. 29 This powerful technique has also proven effective for TBR research, providing detailed information about the processes within the reactor. 30−33 Simulation approaches have been used to investigate the effects of gas and liquid feeds, particle arrangement, reactor temperature, and pressure on the liquid flow process.…”

Section: Introductionmentioning

confidence: 99%

“…Previous studies on poor liquid distribution in the TBR have primarily relied on macroscopic indicators such as pressure drop, porosity, liquid holdup, and outlet liquid distribution. − Advances in detection technologies, such as gamma-ray tomography, electrical capacitance tomography, and magnetic resonance imaging, can make it possible to obtain detailed information on the pore structure and liquid distribution in TBR. − However, such techniques often prove costly, requiring expensive equipment and intricate experimental procedures. In recent years, computational fluid dynamics (CFD) and simulation approaches have emerged as cost-effective and convenient alternatives, gaining attraction in various fields. − Hanbin Zhong et al employed CFD to simulate the instantaneous yield of bio-oil in fluidized biomass pyrolysis, and further optimized their simulation by using long short-term memory network methods based on CFD data, thereby significantly reducing computational time . This powerful technique has also proven effective for TBR research, providing detailed information about the processes within the reactor. − Simulation approaches have been used to investigate the effects of gas and liquid feeds, particle arrangement, reactor temperature, and pressure on the liquid flow process. ,− Nonetheless, the mechanisms underlying liquid flow require further elucidation.…”

Section: Introductionmentioning

confidence: 99%

Numerical Simulation of Liquid Flow Characteristics on a Single Spherical Particle for Trickle Bed Reactor

Wei,

Shi,

Guo

et al. 2023

Liquid distribution plays a crucial role in the performance of the trickle bed reactor (TBR). Therefore, understanding the flow behavior of a droplet impacting a spherical catalyst particle in the TBR is essential for the design and operation of the reactor. In this study, the volume of fluid (VOF) model was used to investigate this phenomenon. The simulation results revealed two distinctive flow patterns: a uniform liquid film flow pattern and a droplet flow pattern. To quantitatively evaluate the two flow patterns, a postprocessing method was proposed based on the Fibonacci sequence. The results indicate that the uniform liquid film flow pattern can maintain larger total gas–liquid interfacial area and specific interfacial area (thinner liquid film) for a longer residence time compared to the droplet flow pattern, demonstrating that the uniform liquid film flow pattern is a promising way to optimize the TBR performance. To facilitate the formation of a uniform liquid film on the particle, using liquids with smaller contact angles, lower surface tension, and adjusting the droplet/particle size ratio to a suitable range can be suggested.