Research on Working Condition Diagnosis of Submersible Screw Pump

. As the development of each oil field enters the middle and late stage, sand production, wax deposition and the continuous increase of well depth, electric submersible screw pump wells frequently break down, and hidden troubles of equipment can not be found in time, which makes oil wells produce with faults. It increases the wear and tear of equipment, seriously affects the production of oil wells, and causes unnecessary economic losses. Therefore, improving the accuracy of the working condition diagnosis of the electric submersible screw pump production system plays a vital role in improving the overall management level of the electric submersible screw pump well. This can not only enhance the managers' correct understanding of this kind of oil well system, but also comprehensively monitor the working state of the well. It can find the fault as soon as possible, analyze the cause of the fault, and avoid the accident by adjusting the production parameters, so that the downhole screw pump can work safely, reliably and effectively, and maximize the maintenance free period and economic benefits of the electric submersible screw pump well.


Introduction
In addition to normal working conditions, screw pump wells also have many other abnormal working conditions, which are regarded as fault conditions.Although there are many kinds of faults in screw pump wells, generally speaking, faults can be divided into surface faults and downhole faults.For the submersible screw pump production system, since the submersible motor and reducer are placed underground, their faults are mainly concentrated in the underground.At the same time, the submersible motor is connected to the screw pump rotor through the motor protector and flexible shaft, so there is no rod breaking and rod tube eccentric wear fault.There are five typical working conditions of submersible screw pump, including flexible shaft breaking, wax deposition, pump jamming, pump leakage and normal operation.
Cui Jinbang and others from North China Bureau, based on understanding the testing method of screw pump well working conditions, researched and designed the working condition test technology of screw pump well mainly consists of working condition test and replay diagnosis [1].Wang Haiwen of the University of petroleum and others proposed a working condition diagnosis technology of the bare rod stress method by analyzing the stress on the bare rod of the screw pump oil production system, developed the bare rod stress test system, and analyzed some commonly used working condition diagnosis models of screw pump wells by studying the relationship between the working conditions of screw pump wells and relevant data, and compiled the corresponding working condition diagnosis software [2].Liu Wenbo of Southwest Petroleum Institute has developed a test method that can measure the stress on the polished rod of screw pump well System, and the corresponding auxiliary process is given [3].Gong Junfeng of Shengli Oilfield realized that there are few working condition diagnosis methods for screw pump wells, and developed a working condition test model for screw pump, which includes two modules -working condition diagnosis and optimization design.In working condition diagnosis, various methods such as current method and liquid production method are used to diagnose oil wells [4].After analyzing the production status of oil wells, Wu Xiaodong and others from the University of petroleum selected five characterization parameters of the production status of oil wells, such as production, dynamic liquid level and working current.Combined with the two judgment theories of threshold diagnosis and logic diagnosis, they formulated the rules for deriving the working status of oil wells from the state variables and studied a new diagnostic method for working conditions of screw pump wells [5].Jose F. Correa et al.Proposed a new working condition analysis method of screw pump well, which combines rough set theory with expert system.This method can not only judge the working condition of screw pump well, but also give the optimization scheme of fault well, which is convenient for managers to adjust the fault well [6].

State diagnosis model based on artificial neural network
By analyzing the fault diagnosis of screw pump well, combining the advantages of wavelet packet decomposition theory in extracting signal features and artificial neural network in pattern recognition, a diagnosis method based on wavelet packet extraction of fault features and artificial neural network for online fault recognition is proposed.

Wavelet packet algorithm
The decomposition algorithm and reconstruction algorithm of wavelet packet are shown in the following formulas: Wavelet packet decomposition algorithm: It can be seen from the above formula that the wavelet packet transform can decompose both the low-frequency part and the high-frequency part of the image.It has a stronger adaptability and a wider application range, so it is more suitable for various image processing.

Feature extraction of fault signal based on
wavelet packet Wavelet packet decomposition can extract fault features in a more refined frequency band.The following takes the decomposition of three-layer wavelet packet as an example to illustrate the decomposition process of wavelet packet.

Figure 1. Three layer wavelet packet decomposition tree
The node (I, J) represents the jth node (I = 0, 1, 2,...) of the i-layer, and each node represents certain signal characteristics.Where node (0, 0) represents the original signal s, node (1, 0) represents the low-frequency coefficients of the first layer of wavelet packet decomposition, node (1, 1) represents the high-frequency coefficients of the first layer of wavelet packet decomposition, node (2, 0) represents the low-frequency coefficients of the second layer of wavelet packet decomposition, and so on.

Artificial Neural Network Model
Artificial neural network constructs an artificial information processing system by simulating biological neural network processing methods, and connects neurons (processing units) based on specific relationships, thus forming a relatively complex network structure.When a large number of samples are given, the rules in the samples are automatically learned and summarized through the learning of the samples, and the relationship weights connecting each neuron are constantly modified, so that the network tends to be stable.Finally, according to the obtained network, a new input mode is given, and a corresponding output mode can be deduced.
(1) Artificial neuron model (2) Topological structure of neural network By organizing a large number of basic neurons in a certain structure, a neural network can be obtained.According to the combination relationship and action mode of each neuron, neural networks with different characteristics can be obtained.Generally speaking, it can be divided into two types: interconnected type and layered type.The layered type can be further divided into forward type, inter layer interconnection forward type and feedback forward type according to whether there is feedback or not and the type of feedback; Interconnected neural networks can be divided into two types according to the connection relationship of neurons: local interconnection and full interconnection.

A diagnosis model of working conditions based on logical discrimination
According to the analysis of the causes and characteristics of various abnormal working conditions of electric submersible screw pump, the logical relationship between various working conditions and operating parameter States is summarized, and the inference mechanism of various working conditions is designed according to the logical relationship between operating parameter States and operating conditions.
(1) Flexible shaft / tubing disconnection detection If the output is zero and the current is ultra-low, it is judged that the flexible rod is broken; If the output is zero and the current is too high, it is judged that the pump is jammed; If the output is zero and the current is normal, it is judged that the oil pipe has fallen off; In all three cases, it is recommended to shut down the well for workover.
(2) Detection of pump card / stator swelling / wax deposition If the current is too high and the output is zero, it is judged that the pump is stuck, and it is recommended to shut down the well for workover.If the current is too high and the output is not zero, it is judged that the stator is swollen or waxed; The current is on the high side, the pump inlet temperature does not change, and the wellhead temperature increases.It is judged that the stator is swollen or waxed; In the above two cases, it is recommended to run at high speed for 10 minutes first, and then reduce to the original speed.If the front and rear currents do not change much, the working condition is stator swelling, and vice versa; If the working condition is wax deposition, wax removal shall be adopted; If the working condition is stator swelling, reduce the speed and run for a period of time; if there is no improvement, stop the pump for inspection; If the current is only on the high side, it may be motor failure.
(3) Screw pump / oil pipe leakage detection If the output decreases and the working current of the screw pump is low, the leakage of the screw pump should be analyzed.Calculate the volumetric efficiency according to the identification results of screw pump speed and wellhead flow data, analyze the current volumetric efficiency in combination with the pressure difference between the inlet and outlet of the pump and the characteristic curve of the screw pump, and understand the internal leakage of the screw pump.
Analyze the rotational speed identification results of the screw pump to obtain the volumetric efficiency of the screw pump, and classify the leakage according to the ratio between the calculated volumetric efficiency and the volumetric efficiency in the characteristic curve.If the ratio between the calculated volumetric efficiency and the volumetric efficiency in the characteristic curve is greater than or equal to 85%, it is diagnosed as a micro leakage of the screw pump, and it is recommended to continue working at this time; If the ratio between the calculated volumetric efficiency and the volumetric efficiency in the characteristic curve is greater than 60% and less than 85%, it is diagnosed as a slight leakage of the screw pump.At this time, it is recommended to reduce the speed and continue to operate; If the ratio between the calculated volumetric efficiency and the volumetric efficiency in the characteristic curve is greater than or equal to 40% and less than or equal to 60%, it is diagnosed as moderate leakage of the screw pump.At this time, it is recommended to reduce the speed and continue to operate, and observe the effect.If the situation does not improve, it is recommended to stop the pump; If the ratio between the calculated volumetric efficiency and the volumetric efficiency in the characteristic curve is less than 40%, it is diagnosed that the screw pump has serious leakage.At this time, it is recommended to stop the pump immediately and consider replacing the pump.If only the production is reduced, it is judged as tubing leakage.

Figure 2 .
Figure 2. Basic model of neuron