基于识别规则的注水站离心泵故障预测方法

Fault prediction method of centrifugal pump in water injection station based on identification rules

针对油田注水站通过传统的人工“望闻问切”等手段,分析判断离心泵机组工况现状的不足,提出了一种基于识别规则的智能故障诊断方法,实现了离心泵实时故障诊断。首先,通过采集注水站离心泵正常状态以及异常状态时的参数,接着分析两种状态下这些参数的变化趋势,将其进行对比分析,研究出离心泵在正常状态和异常状态下参数变化的特征。再次,通过这些离心泵机组参数的变化率及其阈值,计算出离心泵机组故障异常的边界条件。最后,研制出一种离心泵故障识别的规则,实现对离心泵故障的智能预测。通过油田离心泵机组的现场验证,该方法预测故障的准确率达到了99.5%,确保了离心泵机组正常生产运行。

In view of the inadequacy of the traditional manual diagnostic methods such as“looking,hearing,inquiring and touching”used in oilfield water injection stations for analyzing the operational status of centrifugal pump units,an intelligent fault diagnosis method based on identification rules is proposed in this paper,which realize real-time fault diagnosis of centrifugal pumps.Firstly,parameters from the normal operation and fault states of the centrifugal pumps at the water injection station are collected.Then,by analyzing the trend of these parameter changes under both conditions and comparing them,the characteristic differences in parameter variations between the normal and abnormal states of the centrifugal pumps are identified.Subsequently,by examining the change rate and threshold value of these parameters,the boundary conditions for detecting anomalies in the centrifugal pump units were established.Finally,a set of rules for identifying centrifugal pump faults is developed,enabling intelligent prediction of centrifugal pump faults.Field validation of this method at oilfield centrifugal pump units shows that the accuracy rate of predicting faults reaches 99.5%,ensuring the normal production and operation of the centrifugal pump units.