基于PDC的厂级振荡根源定位研究
发布时间：2018-11-03 20:13
【摘要】：过程工业既是能源、各种原材料的生产者,同时也是能源的主要消耗者,节能降耗对提高企业的经济效益和促进国家低碳环保战略的施行有重大的意义。过程工业中由于控制回路参数调校不当,阀门粘滞,过程非线性,控制系统设计低劣,外部输入干扰等原因,往往导致了整个过程系统的厂级范围大规模的振荡。及时评估多回路控制系统性能,检测系统中的过程干扰及异常工况,从而准确定位并诊断引发振荡的根源,并对故障源进行维护修理,确保企业产品的质量和生产效率,是目前过程工业中关注的主要问题之一 本课题以偏有向相干分析PDC(Partial directed coherence)为基础,针对过程工业中采集和存储的大量过程历史数据,提出了新的基于PDC的厂级振荡根源定位方法。该方法通过传统功率谱图和频谱ICA(Independent Component Analysis)算法来筛选数据,然后对筛选数据进行偏有向相干分析,对过程变量间波动传播方向和强度大小进行辨识,同时构建变量间的PDC因果关系图,最后应用DFS/BFS的权值阈值搜索算法和过程先验知识对因果关系图进行化简,删除次要因果关系分支,直观辨识波动传播路径。 本课题中,利用该方法对基于SIMULINK的仿真案例数据,以及EASTMAN化工厂、上海某氯碱厂工业历史数据进行分析,进行厂级振荡根源定位,结果表明,这一套根源定位方案是切实可行的。
[Abstract]:Process industry is not only the producer of energy and raw materials, but also the main consumer of energy. Energy saving and reducing consumption are of great significance to improve the economic efficiency of enterprises and promote the implementation of national low-carbon environmental protection strategy. Due to improper adjustment of control circuit parameters, valve stickiness, process nonlinearity, poor design of control system, external input interference and so on, large scale oscillation of the whole process system is often caused. Evaluate the performance of the multi-loop control system in time, detect the process interference and abnormal working condition in the system, and accurately locate and diagnose the source of the oscillation, and carry on the maintenance and repair to the fault source to ensure the quality and production efficiency of the enterprise product. It is one of the most important problems in process industry at present. Based on the partial directed coherence analysis (PDC (Partial directed coherence), a large number of process history data are collected and stored in the process industry. A new method based on PDC is proposed to locate the source of plant oscillation. In this method, the traditional power spectrum and spectrum ICA (Independent Component Analysis) algorithm are used to screen the data, and then the biased coherent analysis is carried out to identify the direction and intensity of the wave propagation between the process variables. At the same time, the PDC causality graph between variables is constructed. Finally, the weight threshold search algorithm of DFS/BFS and the process prior knowledge are used to simplify the causality graph, to delete the secondary causality branch, and to identify the wave propagation path intuitively. In this paper, the simulation case data based on SIMULINK and the industrial historical data of EASTMAN chemical plant and a chlor-alkali plant in Shanghai are analyzed, and the source of the oscillation is located. The results show that, This solution is feasible.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
本文编号：2308898
[Abstract]:Process industry is not only the producer of energy and raw materials, but also the main consumer of energy. Energy saving and reducing consumption are of great significance to improve the economic efficiency of enterprises and promote the implementation of national low-carbon environmental protection strategy. Due to improper adjustment of control circuit parameters, valve stickiness, process nonlinearity, poor design of control system, external input interference and so on, large scale oscillation of the whole process system is often caused. Evaluate the performance of the multi-loop control system in time, detect the process interference and abnormal working condition in the system, and accurately locate and diagnose the source of the oscillation, and carry on the maintenance and repair to the fault source to ensure the quality and production efficiency of the enterprise product. It is one of the most important problems in process industry at present. Based on the partial directed coherence analysis (PDC (Partial directed coherence), a large number of process history data are collected and stored in the process industry. A new method based on PDC is proposed to locate the source of plant oscillation. In this method, the traditional power spectrum and spectrum ICA (Independent Component Analysis) algorithm are used to screen the data, and then the biased coherent analysis is carried out to identify the direction and intensity of the wave propagation between the process variables. At the same time, the PDC causality graph between variables is constructed. Finally, the weight threshold search algorithm of DFS/BFS and the process prior knowledge are used to simplify the causality graph, to delete the secondary causality branch, and to identify the wave propagation path intuitively. In this paper, the simulation case data based on SIMULINK and the industrial historical data of EASTMAN chemical plant and a chlor-alkali plant in Shanghai are analyzed, and the source of the oscillation is located. The results show that, This solution is feasible.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
【参考文献】
相关期刊论文 前5条
1 孙金明,左信,季德伟,陈育昆;PID控制器性能评价[J];化工自动化及仪表;2004年05期
2 夏春明;吴钦;郑建荣;;基于格兰杰因果检验的过程波动路径及根源识别分析[J];华东理工大学学报(自然科学版);2010年01期
3 陈耀,罗晓,孙优贤;过程控制工程中的若干问题讨论[J];基础自动化;2002年02期
4 王印松;张彩;;基于设定值响应的PI控制系统性能评估[J];控制工程;2007年01期
5 张旭昶,胡立生;模型预测控制的性能监视[J];机电一体化;2004年04期
，本文编号：2308898
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2308898.html
