Predictive modeling and optimization of hydrochar properties from food waste hydrothermal carbonization using machine...

Area of Science:

• Sustainable Chemistry
• Waste-to-Energy Technologies
• Materials Science

Background:

• Hydrothermal carbonization (HTC) converts food waste into valuable hydrochar for energy and soil applications.
• Optimizing HTC is complex due to interacting parameters like temperature, time, and catalyst dosage.
• Machine learning (ML) offers a promising approach to predict and optimize HTC processes.

Purpose of the Study:

• To apply ML techniques for predicting and optimizing hydrochar yield and properties from food waste HTC.
• To evaluate the performance of XGBoost, SVR, and LR models in predicting HTC outcomes.
• To identify key process parameters influencing hydrochar characteristics.

Main Methods:

• Analyzed HTC process parameters: temperature (120-360°C), catalyst dosage (0-2g), residence time (30-270min), and moisture content.
• Developed and compared three ML models: XGBoost (XGB), Support Vector Regression (SVR), and Linear Regression (LR).
• Assessed model predictive accuracy using R², MAE, and RMSE metrics and identified parameter importance.

Main Results:

• XGBoost achieved the highest predictive accuracy (R²=0.87, MAE=0.81, RMSE=1.02), outperforming SVR and LR.
• Temperature (importance=0.91) and catalyst dosage (0.61) were the most significant factors influencing hydrochar yield and properties.
• Optimized conditions produced 48.5g hydrochar/100g dry waste with 68% energy recovery.
• TiO₂ nanoparticles significantly improved carbon retention (56.8% to 77.6%), bulk density (0.37 to 0.89 g/cm³), and heating value (18.4 to 22.7 MJ/kg).

Conclusions:

• ML-driven optimization effectively predicts and enhances hydrochar production from food waste via HTC.
• XGBoost is a suitable model for real-time HTC process control, reducing experimental efforts.
• This approach advances sustainable waste-to-energy conversion, improving resource utilization and environmental remediation potential.