基于SVR-PCA的DMTO过程监测研究  

Research on process monitoring of DMTO based on SVR-PCA

作　　者：赵泽盟 李洋 李超 史元腾 Zhao Zemeng;Li Yang;Li Chao;Shi Yuanteng(China Coal Energy Research Institute Co.,Ltd.,Xi′an Shaanxi 710054,China;China Coal Xi′an Design Engineering Co.,Ltd.,Xi′an Shaanxi 710054,China;Wuxi Research Institute of Applied Technologies,Tsinghua University,Wuxi Jiangsu 214072,China)

机构地区：[1]中煤能源研究院有限责任公司,陕西西安710054 [2]中煤西安设计工程有限责任公司,陕西西安710054 [3]清华大学无锡应用技术研究院,江苏无锡214072

出　　处：《煤化工》2024年第4期109-114,共6页Coal Chemical Industry

基　　金：中国中煤能源集团重点科技项目(20211CY005)。

摘　　要：由于DMTO反应器的反应温度和建模变量呈非线性特征,常规的线性模型无法准确捕捉其特征,为此建立了基于SVR-PCA的过程监测模型,能够避免非线性特征对模型的影响。结果表明:SVR-PCA过程监测模型能够比PCA过程监测模型提前63 min、比DCS系统预警提前163 min对DMTO反应温度长时间大幅波动进行报警,能够减少反应温度短时间小幅波动和反应温度长时间小幅波动的误报率。Due to the nonlinear nature of DMTO reaction temperature and modeling data,conventional linear model could not accurately capture its characteristics.Therefore,the process monitoring model based on SVR-PCA was constructed to mitigate the impact of nonlinear features on the model.The results showed that the SVR-PCA process monitoring model could provide an alarm for significant and long-term fluctuations in DMTO reaction temperature 63 minutes earlier than the PCA process monitoring model,and 163 minutes earlier than the DCS system.It could also reduce the false alarm rate for short-term small fluctuations and long-term small fluctuations in reaction temperature.

关 键 词：DMTO 反应温度 过程监测 支持向量回归 主元分析模型 

分 类 号：TQ53[化学工程—煤化学工程]