Random forest-based modeling for insights on phosphorus content in hydrochar produced from hydrothermal carbonization of sewage sludge

The hydrochar produced from hydrothermal carbonization(HTC) of sewage sludge (SS) usually has a high phosphorous (P) content, and that would result in fouling and energy efficiency reduction. Therefore, it is important to monitor the P content during the hydrochar production process. This work suggests a data -driven Random Forest-based model to predict the total P content in the hydrochar (TP-hc) from the HTC of SS. Various configurations of inputs features were examined, including the data of proximate analysis, ultimate analysis, ultimate and proximate analyses, and for each configuration, either if the total P in the SS (TP-ss) was known or not. Overall, the models including TP-ss as input have accurately predicted the TP-hc with an R-2 located in [92-95%]. Features' importance approach and partial dependence analysis pointed out that the TP-ss, ash content, reaction temperature (T), reaction time (t), and initial pH of feedwater exhibit positive effect on the TP-hc. In contrast, contribution of the volatile matter (VM) of SS was mostly negative. Dry matter loading exhibits no obvious monotonicity with TP-hc. This work could guide the production of SS-hydrochar with the desired P content, and thus avoid time and resources consuming for many trials. (C)& nbsp;2022 Elsevier Ltd. All rights reserved.

Automated summary

This study proposes a data-driven model to predict the total phosphorous content in sewage sludge hydrochar. The results show that the total phosphorous content in the sludge, ash content, reaction temperature, reaction time, and pH of the feedwater have a positive effect on the total phosphorous content in the hydrochar, while the contribution of volatile matter in the sludge is mostly negative. There is no clear monotonic relationship between dry matter loading and the total phosphorous content in the hydrochar.