Optimizing the early-stage of composting process emissions - artificial intelligence primary tests.

Area of Science:

• Environmental Science
• Agricultural Science
• Artificial Intelligence

Background:

• Composting organic waste presents emission challenges, including ammonia (NH3), carbon monoxide (CO), hydrogen sulfide (H2S), and carbon dioxide (CO2).
• Novel analytical methods, particularly those leveraging artificial intelligence (AI), show promise for improving composting processes.

Purpose of the Study:

• To predict and optimize emissions during early-stage composting using machine learning (ML) models.
• To develop AI-driven tools for real-time monitoring and management of composting emissions.

Main Methods:

• Laboratory composting experiments were conducted using varying incubation temperatures (50-70°C) and biochar doses (0-15% dry mass).
• Machine learning models, including Artificial Neural Networks (ANN) and Decision Trees (DT), were selected and trained using emission data (CO, CO2, NH3, H2S).

Main Results:

• ANN and DT models demonstrated satisfactory predictive accuracy for key composting emissions.
• ANN achieved R² accuracies of 0.71 for CO, 0.81 for CO2, 0.95 for NH3, and 0.72 for H2S.
• DT models showed comparable R² accuracies: 0.69 for CO, 0.80 for CO2, 0.93 for NH3, and 0.65 for H2S.

Conclusions:

• ML models successfully predicted CO and H2S emissions during composting for the first time.
• AI-based emission prediction offers a cost-effective and rapid alternative to empirical compost analysis.
• This approach facilitates enhanced control and optimization of composting operations to mitigate harmful emissions.