芳烃异构化反应系统动态建模及性能评估研究

项目名称： 芳烃异构化反应系统动态建模及性能评估研究

项目编号： No.61304125

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 傅永峰

作者单位： 浙江工业大学

项目金额： 26万元

中文摘要： 本项目针对芳烃异构化系统反应机理复杂、非线性强、时变性强而导致的建模困难，实时监控与评估缺失，先进控制技术无法实施等问题，拟开发基于静动态异类多模型动态融合的自适应混合统计模型评估系统的反应性能，分析操作条件等工艺变量对反应性能参数的影响；为提高监控模型的实时计算性能，控制样本规模，开发ILS-SVM算法作为上述动态模型；提出一种新的基于模糊曲线加权的MMMD算法剔除建模样本中的粗差并用基于小波变换的算法处理样本中的随机误差，以确保建模数据的准确性，提高模型的计算性能；最后，提出一种新的动态建模数据处理算法，并基于该算法建立动态模型以描述过程的动态特性,提高模型的实用性。本项目成果将用于芳烃异构化系统反应性能的实时监控与评估，以保障反应系统安全、稳定、高效地运行，为芳烃异构化反应系统的先进控制提供理论和技术支撑，具有非常重要的科学研究意义与实际应用价值。

中文关键词： 芳烃异构化系统；过失误差侦测；多模型动态融合；双重校正策略；软测量

英文摘要： Because of the complex reaction mechnism,strong nonliearity and time variability, it is very difficult to modeling aromatics isomerization systems. So, real-time monitoring, evaluation and advanced control technologies can not be implemented in these systems.To solve these problems, an adaptive hybrid statistical model based on static and dynamic heterogeneous multi-model dynamic fusion was proposed in this project to evaluate the performance of the reaction system. Based on this model,the impact of process variables such as operating conditions to performance parameters of the reaction can be analyzed.To improve the real-time computing performance and cotrol the number of data samples of the model, an ILS-SVM ( Incremental Least Squares Support Vector Machines) algorithm was proposed and was applied in the model mentioned above as the dynamic model.Furthermore, in order to ensure the accuracy of the modeling samples and improve the comuting performance,a new MMMD (Modified Median Minimum Distance) algorithm weighted by fuzzy curve was proposed to exclude the gross errors in the modeling samples. Finally, a new dynamic modeling data processing algorithm was proposed. The model built based on this algorithm can describe the dynamic characteristics of the process.This makes the usefulness of the model better. The

英文关键词： Aromatics isomerization system；Gross error detection；Multi-model dynamic fusion；Dual updating strategy；Soft sensor