Improving Kitchen Waste Composting Maturity by Optimizing the Processing Parameters Based on Machine Learning Model

Shang Ding,Wuji Huang,Wei‐Jian Xu,Yiqu Wu,Yuxiang Zhao,Ping Fang,Baolan Hu,Liping Lou
DOI: https://doi.org/10.2139/ssrn.4128995
2022-01-01
Abstract:As a new method of big data analysis, machine learning model can explore the relationship between different parameters and draw universal conclusions, which was used to predict composting maturity and identify key parameters in this study. The results showed that the Stacking model exhibited excellent prediction accuracy (Testing set: R 2 =0.8746, RMSE=6.4539, MAE=3.7936). SHAP analysis and Partial Dependence Analysis (PDA) were performed to determine the importance of each feature as well as the optimal interval of processing parameters. Optimal starting conditions should be maintained during the mesophilic state (temperature: 30-45℃, moisture content: 55-65%, pH: 6.3-8.0), and nutrients (TN>2.3%, TOC>35%) should be adjusted during the thermophilic state. Experiments revealed that model-based optimization strategies could improve composting maturity because they could optimize compost microbial flora and perform complex carbon cycle functions in the cooling and maturing stage. In conclusion, this study contributes to enhancing composting processes and improving the composting quality.
What problem does this paper attempt to address?