Volume 07: Proceedings of Applied Energy Symposium: CUE2019, China, 2019

Multi-Task Prediction of Fuel Properties of Hydrochar Derived From Wet Municipal Wastes With Random Forest Jie Li, Xinzhe Zhu, Yinan Li, Yen Wah Tong, Xiaonan Wang*

Abstract

Waste to energy is a promising way to ease the urban burden of waste treatment and hydrothermal carbonization (HC) can dewater the municipal wastes with high moisture efficiently with hydrochar left. The hydrochar with outstanding fuel characteristics can be used as fuel for incineration to generate power. To predict the fuel characteristics of hydrochar including the yield, higher heating value (HHV) and carbon content (C_char) based on the information of the wet municipal waste, machine learning methods have been explored in this work. Results show that the optimized Random Forest (80 trees with 10 maximum depths) has good multi-task prediction capability of fuel characteristics. The R 2 for the predictions of the yield, HHV and C_char are 0.80, 0.91 and 0.95, respectively. Moreover, according to the feature importance analysis, the yield of hydrochar is mainly determined by the temperature and water content of HC, while the HHV and C_char are dominated by the carbon and ash content of the feedstock, respectively.

Keywords machine learning, multi-task prediction, waste to energy, hydrothermal carbonization, hydrochar

Copyright ©
Energy Proceedings