Research On Well Fault Diagnosis Method Based On Electrical Parameter Data Augmentation

Sucker rod pumping system is the most widely used oil production method in beam pumping Wells at home and abroad.It usually works in the position of thousands of meters deep.Because the working environment is very bad,after a long time of operation,all kinds of faults occur frequently,which seriously affects oil production and even causes oil well accidents.Therefore,an effective oil well fault diagnosis method is urgently needed to solve the above problems.Traditional oil well fault diagnosis methods are mainly based on indicator diagram technology,which is collected by dynamometer or load sensor installed at suspension point.However,this method requires shutdown to affect production,and installation is difficult,the sensor is easy to damage,the use of maintenance costs greatly increased.In addition,the distribution of oil Wells is relatively dispersed and there are a large number of well heads.It is time-consuming and laborious to collect data one by one and patrol well in a large range by human,which makes it difficult to diagnose the oil Wells.Therefore,in order to efficiently diagnose oil well faults in real time,with low cost,convenient data acquisition,reduced labor degree and high fault diagnosis accuracy,this paper adopts an oil well fault diagnosis method based on electrical parameters.With the development of computer network,oil field data acquisition and fault diagnosis tend to be intelligent,and the intelligent oil field is gradually established.In industrial oil production sites,due to the different types of basic data and combination devices of pumping units and motors,different reservoir properties and pump depth,there are slight differences in the electrical parameters collected at different periods of the same well,and the electrical parameters of different Wells under different working conditions may be similar,making it difficult to read the data.At present,the field of oil production industry has not established a scientific method of electrical parameter diagnosis and standard electrical parameter data template,insufficient samples,especially few fault samples,which greatly limits the development of intelligent oil well diagnosis technology.In order to solve the above problems,an oil well fault diagnosis method based on electrical parameter data is proposed.Firstly,based on the analysis of normal and typical fault working conditions of rod pump,the whole process simulation model of mechanical production system is established,and the electrical parameter data used in well diagnosis is generated.Secondly,in view of the limited types of data generated by the mechanism model,the TSC-DCGAN method is used to extend the electrical parameter data.Finally,multiple deep learning network models are used for fault diagnosis of the extended electrical parameter data.Experimental results show the effectiveness of the proposed method.