Research On Diagnosis Of Indicator Card Of Pumping Well Based On Elman Neural Network

Now in the process of pump well’s oil extraction, it is a urgent issue to be resolvedin oil industry to keep abreast of and mastery of the pumping unit working conditionsunderground and to achieve automatic monitoring of the entire production systems.Pumping equipment often suffer undue damage and seriously affect the oil yield andeconomic benefits due to the complex environment under the pumping well.Therefore, these thesis study the intelligent diagnosis technology of pumping wellcondition is studied on the basis traditional condition diagnosis.Traditionally, the pumping well staff through the measured ground indicatordiagram to determine the working conditions of the pumping well, but the variants ofthe pumping unit poles, the viscous resistance of the rod, the vibration and inertia,theground indicator diagram difficult to make the correct judgment based on the workingcondition of the pumping well. So, this thesis propose a new method that through usethe pumping well’s current variation to react the working situation of the pumping well,these method called current method. then apply this principle to establish of a pumpingwell current changes over time indicator diagram diagnosis model. Then these thesiselaborated the dynamometer graphics features and formation process, pretreatment andmake the indicator diagram size normalized, build the indicator diagram typical faultsample as the subject investigated. Through study the indicator diagram shape features,in accordance with the principles of the image processing, make the indicator diagramgrayscale matrix, then extracted six features based on the indicator diagram grayscalematrix parameters, these six parameters constitute a classification feature vector. Thenthese thesis study the basic structure of the BP neural network and the Elman neuralnetwork, include the network integration and network configuration parameters. Thenwe compare the BP neural network and Elman network’s advantages and disadvantages.finally,we use Elman neural network to build the pumping well intelligent diagnosisclassification model, the classification feature vector as the input of the network, thetype of the failure mode as the output, Elman neural network learning through a largenumber of samples, it makes the weights and threshold stored in the form of a network. Finally, we write indicator diagram graphical extraction module and Elman neuralnetwork classifier module through using matlab, the trained network is applied to thesimulation test and the simulation results are given. The test results show that thismethod is feasible and has some theoretical significance and practical value.