基于格兰杰因果检验的过程波动路径及根源识别分析

为解决流程工业中大范围过程变量波动时波动传播路径的确定和波动源所在回路的定位问题,提出了一种基于格兰杰因果检验的分析方法。该方法结合已有的频域分析方法,筛选出含相近频率成分的受扰波动变量,对其进行格兰杰因果检验,检验得到的回路变量间的因果影响关系表征了波动干扰的传播路径,以因果关系图的形式直观表达。利用先验过程知识化简因果关系图,并通过切除阈值搜索方式过滤掉次要因果关系分支,得到波动干扰的主要传播路径,由此定位波动源。仿真及工业数据的验证结果表明该方法是可行的。

Propagation and Root-Cause Identification of Process Oscillations via Granger Causality Test

To identify the propagation paths of multi-loop process oscillations in process industry and locate the origin of the oscillation disturbance, a novel method based on Granger causality test is proposed.Pre-processed by existed spectral analysis, process variables of oscillatory control loops with similar frequency components are selected for Granger causality test, which can reveal the complex causality among the selected process variables.The causality is visualized in the form of causality diagram, which represent the propagation paths of the oscillations and can be further simplified by the understanding of process knowledge.Less significant causality pathways are filtered out by manual cutting-off threshold searching.The resulting diagram represents the main oscillation propagation paths, which is traced to identify the source of the oscillation disturbance.Both simulated and real plant data tests are presented that the proposed method is feasible.