Machine learning predicts properties of hydrochar derived from digestate

Abstract

Background

Hydrothermal carbonization (HTC) is a promising solution for digestate valorization, and machine learning (ML) is a helpful tool for modeling hydrochar properties.

Methods

This study utilized two ensemble tree-based ML algorithms, the random forest (RF) and the eXtreme Gradient Boosting (XGB), for predicting digestate-derived hydrochar yield, properties (Cc, Hc Nc, Oc, Sc, Ashc, HHVc), and HTC process index including energy yield (EY), energy densification (ED), and carbon recovery (CR).

Significant Findings

In most cases, XGB showed better predictive performance, including yield, Cc, Hc, Nc, Ashc, HHVc, EY, and ED prediction, while RF revealed better performance in Oc, Sc, and CR prediction. XGB and RF showed satisfactory performance in predicting Cc, Hc, Oc, Sc, Ashc, and HHVc, with test R² of 0.856–0.942 and 0.864–0.947, respectively. The multi-task model for predicting yield and hydrochar properties (Cc, Hc, Nc, Oc, Sc, Ashc, HHVc) was also developed. XGB reveals better performance than RF, with the average test R² of XGB could achieve 0.895, which is comparable to the current published work. The SHapley Additive exPlanations (SHAP) analysis reveals that digestate ash content, C content, and HTC temperature (T) dominate multi-task predictions. The chain regressor technique enhanced the model performance toward multi-task prediction, including EY, ED, and CR: in RF, the test R² of ED and CR were increased by 38 % and 26 %, respectively, while in XGB, the test R² of ED was improved by 48 %. The developed ML model in this work could satisfactorily predict hydrochar properties, forming a basis for optimizing HTC process parameters and determining suitable applications for digestate valorization. ML effectively maps the correlation between input features and output responses, making ML a time-efficient and practicable tool for prediction tasks and identifying essential features, especially for multi-output prediction with high-dimension.