Advanced dual-artificial neural network system for biomass combustion analysis and emission minimization

Area of Science:

• Agricultural Engineering
• Environmental Science
• Biomass Energy

Background:

• Agricultural waste management is crucial for plantations.
• Composting is not always efficient for all waste types.
• Thermal treatment of waste requires optimization to minimize greenhouse gas emissions.

Purpose of the Study:

• To investigate the feasibility of using artificial neural networks (ANNs) to predict feedstock and emission parameters from vineyard biomass combustion.
• To develop a novel dual-ANN system for accurate prediction and optimization.
• To provide recommendations for feedstock selection for bioenergy production and emission reduction.

Main Methods:

• Construction of a novel dual-ANN system comprising two cascade-forward ANNs.
• Training ANNs on independent datasets with three hidden layers each.
• Benchmarking the dual-ANN system's accuracy and relative error.

Main Results:

• The dual-ANN system achieved a maximum relative error of 2.09%.
• ANN models can predict feedstock parameters for optimal bioenergy production.
• Specific feedstock characteristics (leaf amount, distribution, mass, area) influence calorific values.
• Different feedstock conditions are required for emissivity reduction.

Conclusions:

• It is not possible to provide universal recommendations for both energy and carbon benefits.
• Feedstock selection should be tailored based on specific goals (energy production vs. emission reduction).
• The study outlines a clear direction for seeking consensus in optimizing biomass combustion.