Investigation On Data Analysis And Fault Diagnosis Of Electric Submersible Pump In Offshore Oil,Gas And Injection Wells

As the core equipment of offshore oil industry production,electric submersible pumps need to work under high temperature and high pressure production conditions for a long time,and are a production node with frequent failures.Once the failure of the electric submersible pump,which leads to stop working and stop production,will cause serious economic losses.Efficient fault diagnosis of electric submersible pumps can help production managers to quickly and accurately determine faults and reduce equipment repair time.It can also help producers to find out early faults of electric submersible pumps in time,improve repair effects,and delay equipment scrap time,prevent a series of chain reactions and improve the safety and durability of the entire system.Therefore,it is of great practical significance to conduct fault diagnosis research on electric submersible pumps.The traditional fault diagnosis of electric submersible pumps in offshore oil,gas and injection wells mainly relies on manual calculation and expert experience.With the complexity of electric submersible pump oil well equipment and the precision of information acquisition technology,the data information of electric submersible pumps presents the characteristics of higher dimensionality and nonlinearity.Traditional methods cannot dig out the deep features of the fault information of electric submersible pumps,and hence the accuracy of the fault prediction of the electric submersible pump is not high enough and the fault diagnosis ability becomes dissatisfactory.Therefore,how to fully excavate the deep features contained in the intricate fault information has become a major problem in the fault diagnosis of electric submersible pumps.This article uses the data of electric submersible pumps in offshore oil,gas and injection wells,to carry out the research on data preprocessing,data analysis and fault diagnosis on electric submersible pumps.The main tasks of this article are as follows:1.Data preprocessing.The collected data of electric submersible pumps in offshore oil and gas wells are all dirty data,which has many abnormalities and missing problems.So the data preprocessing is required.Firstly delete the abnormal,missing,duplicate and irrelevant data;then propose the use of linear interpolation,quadratic spline interpolation,cubic spline difference and nearest neighbor algorithm to interpolate the data,and calculate the data and variables of various interpolation algorithms Euclidean distance to select the optimal interpolation algorithm;finally the data is transformed.This will provide data support for subsequent experiments.2.Data analysis.Electric submersible pump data only has the information of the point of failure,and the target variables that affect the failure of the electric submersible pump are not known in advance.Therefore,a large amount of data is used to explore and analyze the curve change trend of the data characteristics of the electric submersible pump to initially select the main characteristics implying the failure of the electric submersible pump.Then,these features are combined with the maximum information coefficient,feature elimination and random forest algorithm to select main features to establish relevant data sets.3.Fault diagnosis.Kmeans clustering,the algorithm based on the principal component analysis and Mahalanobis distance,and deep autoencoder are used to conduct intelligent fault diagnosis research on the established data set to inform the fault occurrence time.These three diagnostic algorithms have their own advantages,but the deep autoencoder algorithm is more effective in diagnosing the time of the electric submersible pump failure in advance.