Machine Learning Approach for the Prediction of Biomass Waste Pyrolysis Kinetics from Preliminary Analysis.

Area of Science:

• Biomass energy conversion
• Chemical engineering
• Machine learning applications

Background:

• Biomass pyrolysis is a key thermochemical conversion process.
• Accurate prediction of pyrolysis kinetics is crucial for process optimization.
• Machine learning offers potential for modeling complex kinetic behaviors.

Purpose of the Study:

• Develop artificial neural network (ANN) models to predict biomass pyrolysis kinetics.
• Evaluate different input parameter sets for ANN model accuracy.
• Enhance prediction accuracy using optimization algorithms.

Main Methods:

• Collected thermogravimetric analysis and feedstock characterization data for 32 biomass types.
• Developed three ANN models using ultimate analysis, proximate analysis, and component-based (cellulose, hemicellulose, lignin) inputs.
• Utilized particle swarm optimization to refine the Backpropagation ANN (BP-ANN) model.

Main Results:

• ANN models achieved prediction deviations of 20.80%, 14.06%, and 12.85% for different input sets.
• Component-based input (cellulose, hemicellulose, lignin) yielded the highest accuracy.
• Optimized BP-ANN model reduced the maximum deviation for activation energy prediction from 12.85% to 6.72%.

Conclusions:

• ANN models are effective tools for predicting biomass pyrolysis kinetics.
• Biomass composition, particularly cellulose, hemicellulose, and lignin content, is vital for accurate kinetic modeling.
• Particle swarm optimization significantly enhances the predictive performance of ANN models for biomass pyrolysis.