Yan Mingyue, Bi Huiyang, Wang HuanXu, Xu Caicai, Chen Lihao, Zhang Lei, Chen Shuangwei, Xu Xuming, Li Zhongjian, Hou Yang, Lei Lecheng, Yang Bin
Key Laboratory of Biomass Chemical Engineering of Ministry of Education, College of Chemical and Biological Engineering, Zhejiang University, Hangzhou 310027, China.
Institute of Zhejiang University-Quzhou, No. 99 Zheda Road, Quzhou 324000, China.
Chem Bio Eng. 2024 Nov 26;2(2):110-122. doi: 10.1021/cbe.4c00146. eCollection 2025 Feb 27.
Accurate prediction of syngas compositions in multicomponent organic waste gasification is challenging because of its intricate composition and abundant volatile matter, which contrasts with traditional coal gasification influenced mainly by oxygen-coal ratio. Through process analysis, we identified the furnace temperature as a crucial factor directly impacting gasification reactions. Herein, we developed a hybrid backpropagation neural network (BPNN) model integrating furnace temperature data obtained from a temperature soft-sensing model and utilizing principal component analysis (PCA) for dimensionality reduction. The resulting T-PCA-BPNN model demonstrated outstanding predictive performance, achieving values of 0.95, 0.97, and 0.94 for CO, CO, and H, respectively. Compared to the base BPNN model, the total mean square error (MSE) and mean absolute error (MAE) decreased by 49.4% and 13.3%, respectively. Furthermore, the percentage of predictive errors within 1% (QR) surpassed 90%, underscoring the model's practical applicability. Leveraging PCA and SHapley Additive exPlanations (SHAP) analysis, we established a syngas regulation strategy that controls critical parameters to identify postdimensionality reduction through practical operational adjustments. This data-driven model enhances syngas prediction, thereby facilitating improved process control and optimization in complex organic waste gasification.
多组分有机废弃物气化过程中合成气成分的准确预测具有挑战性，因为其成分复杂且挥发性物质丰富，这与主要受氧煤比影响的传统煤炭气化不同。通过过程分析，我们确定炉温是直接影响气化反应的关键因素。在此，我们开发了一种混合反向传播神经网络（BPNN）模型，该模型整合了从温度软测量模型获得的炉温数据，并利用主成分分析（PCA）进行降维。所得的T-PCA-BPNN模型表现出出色的预测性能，CO、CO₂和H₂的R²值分别达到0.95、0.97和0.94。与基础BPNN模型相比，总均方误差（MSE）和平均绝对误差（MAE）分别降低了49.4%和13.3%。此外，预测误差在1%以内（QR）的百分比超过90%，突出了该模型的实际适用性。利用PCA和SHapley加性解释（SHAP）分析，我们建立了一种合成气调节策略，通过实际操作调整来控制关键参数以识别降维后的情况。这种数据驱动的模型增强了合成气预测能力，从而有助于在复杂的有机废弃物气化过程中改进过程控制和优化。
