文献类型：期刊文献

英文题名：Prediction and optimization of coal ash flow temperature using machine learning approaches

作者：Liu, Xia[1];Wang, Xiaoyu[1];Gao, Yunfei[1];Zhang, Yuqing[1];Chen, Xueli[1];Dai, Zhenghua[1];Yu, Guangsuo[1];Wang, Fuchen[1]

机构：[1]East China Univ Sci & Technol, Inst Clean Coal Technol, Shanghai 200237, Peoples R China

年份：2026

卷号：404

外文期刊名：FUEL

收录：;EI(收录号：20252718728259);WOS:【SCI-EXPANDED(收录号:WOS:001528844700001)】；

基金：This work was supported by National Key R & D Program of China (2023YFB4103800) , National Natural Science Foundation of China (22208104) and Shanghai Yangfan Program (22YF1410300) .

语种：英文

外文关键词：Coal gasification; Flow temperature; Coal ash composition; Machine learning; Optimization

摘要：Coal gasification is one of the important ways to achieve clean and efficient utilization of coal, and the fusion characteristics of coal ash play a crucial role in guiding industrial gasifier operations. This paper establishes prediction models for the fusion characteristic temperature (FT) of coal ash based on machine learning (ML) algorithms using a comprehensive dataset that encompasses a vast majority of coal types in China. It was found that the XGBoost model optimized by Bayesian optimization algorithm exhibited the best predictive performance, achieving a test set R2 of 0.886 and a test set RMSE of 29.166. Feature importance analysis revealed that Al2O3 and A/B have higher importance in predicting FT. The constructed ML model is further applied to guide the optimization of FT of high silica-aluminum content coal through the addition of CaO. It was found that the optimization strategies obtained by ML follow the experimental validations. The fundamental difference between the ML model and the FactSage model is also elaborated. Overall, this work presents important progress on coal ash flow temperature prediction and optimization using machine learning approaches and could be applicable to industrial applications.