基于大维数据驱动的油气管网泄漏监控模糊决策方法

A Large Dimensional Data-driven Fuzzy Detection Method for Oil-gas Pipeline Network Leakage

输油管网状态量多及工艺复杂,难以建立精确的管网数学模型,为了能够实时监控管网的安全运行情况,本文提出一种基于大维数据驱动的管网泄漏监控模糊决策方法.首先利用管网现有的数据信息,在不对数据进行降维处理的情况下,从信息物理系统的角度出发,将油气管网的拓扑结构、阀门开度等管道物理数据以及压力、流量等运行信息数据结合起来对复杂管网系统建立数据驱动模型.然后基于大维随机矩阵谱理论,将得到的信息物理数据协方差矩阵谱分布及圆环率作为模糊决策的条件对管网运行情况进行判断.当管网拓扑发生动态变化时,提出的方法可以有效地解决误报率高的问题.最后通过仿真及实例的分析,可以证明所提出方法的有效性.

It is difficult to build an exact math model for the oil-gas pipeline network because there are multiple state variables and complex processing technology. A leakage detection method using fuzzy detection based on large dimensional data for oil pipeline network is proposed, which can monitor oil pipeline network operational status in real time. Firstly, from the view-point of cyber-physical system （CPS）, existing oil pipeline network data containing network topology, physical data like valve opening, and operating data like pressure and flow are used to build a data-driven model without dimensionality reduction. Then, as for fuzzy decision conditions, the eigenvalue spectrum distribution and ring law based on spectral analysis of large dimensional random matrices are used to judge the network operating status. When the topological structure of oil pipeline network is dynamically altered, the fuzzy method can solve the problem of high rate false alarm. Finally, the effectiveness of the proposed method is demonstrated through simulation and case study.