Research On Multi-faults Diagnosis Method Based On Data In Oilfield Producing Process

The thesis takes oilfield production process as study background, and ensureslong-term, safe, reliable and efficient operation of various testing instruments andequipment in production process as the main purpose, mainly focus on multi-faultsdiagnosis of oil field sensor equipment. By using the PCA method in multivariate statisticsto solve a few questions that sensor fault diagnosis is facing with. It provides a new ideaand new method for oil field sensor equipment fault diagnosis. The thesis main workincludes:According to the problems that the traditional PCA fault diagnosis method isimmobilized and the high rate of false alarm in dealing with dynamic data, the iterativemulti-model method is proposed. The data are sorted in accordance with differentoperating conditions by the shortest distance. Building multiple PCA models by using theclassified data, and updating models by iterative algorithm to realize process monitoringand improve the accuracy of fault diagnosis.Allowing for data characteristic of oilfield produce process, the thesis uses weightedsquare prediction error (SWE) to realize fault monitoring. The method builds differentresidual space according to different fault types. Compared with the SPE method in theability of fault diagnosis, the SWE can better use the fault information in residual space,and improve the fault diagnosis accuracy. For the problem that SPE statistic’s faultrecognition ability limited in multi-faults diagnosis process, fault reconstruction methodbased on SWE makes the different fault types of combination and realizes faultreconstruction in different residual spaces which correspond to different fault types ofcombination. It can realize fault recognize effectively.According to data characteristics of sensor equipment in oil field production process,the thesis proposes a kind of method that based on multivariate statistics to solve thetraditional PCA’s several problems that exist in sensor fault diagnosis. Using the real dataof the oil field production is carried out offline fault diagnosis experiments. Experimentsvalidate the effectiveness of the method. Applying the method in oil field sensor equipment for fault diagnosis has the vital significance in ensuring safe and effective production,reducing production cost and improving block production.