基于多尺度建模的炼油化工过程报警根源分析

为了及时找出故障原因以抑制过程故障传播，保证生产过程安全及产品质量，有必要对炼油化工过程报警根源进行分析研究。而对于过程风险表征指标的选择，现有方法多是根据主观经验直接挑选而得，对于一些复杂的大型系统，由于过程复杂，设备众多，常涉及大量监控变量，主观挑选表征指标常常导致报警根源分析过程中遗漏某些重要原因变量。因此提出多尺度建模的方法，通过对炼油化工生产系统的空间尺度进行划分，针对某一风险过程，分析其不同尺度下的过程风险表征指标，基于互相关函数建立多尺度下的炼油化工过程报警根源分析模型。案例分析将其应用于某石油化工企业分馏塔冲塔风险过程。结果表明，对于过程风险报警根源分析而言，需对整个风险单元进行分析，仅考虑与风险直接相关的设备可能遗漏某些重要原因变量，以致无法推绎出引发报警的扰动根源。通过对不同尺度下的模型进行应用对比分析，从而选择合适的尺度进行建模，有助于保证炼油化工过程报警根源分析的准确性。

Alarm Root Cause Analysis in Refinery Process Based on Multi Scale Modeling

In order to find the fault cause timely to suppress fault propagation effectively and ensure production safety and product quality, it is necessary to conduct the process alarm root cause analysis. For the choice of process risk indicators, most of the existing methods were based on subjective experience. Due to the complexity of processes and large amount of equipments in some complex large scale systems with a large number of monitoring variables, subjective selection of risk indicators often led to the omission of some important variables in the alarm root cause analysis. Therefore, a multi scale modeling method was used by dividing the spatial scales of refinery production system. The process risk indicators under different scales were analyzed to establish the risk process alarm root cause analysis model at different scales based on a cross correlation function. The method was demonstrated and validated by the flooding process of a fractionator in a petrochemical enterprise. The results indicate that it is necessary to analyze the whole risk unit for alarm root cause analysis. Some important process variables can be missed if only considering devices that are directly related to the risk, so that the initial disturbance which trigger the alarm will not be derived. The comparison and analysis of models in different scales are made so as to choose the appropriate scale for modeling and ensure the accuracy of process alarm root cause analysis.