A working condition diagnosis method combining analytical solution and exact solution

Abstract

The present invention provides a working condition diagnosis method combining analytical solutions and exact solutions, which relates to the technical field of equipment big data analysis and diagnosis. The working condition diagnosis method combining analytical solutions and exact solutions includes the following diagnosis steps: S1. First, establish a training sample library based on system dynamic simulation, and comprehensively record and store oil well working condition data; S2. Establish a numerical simulation mathematical model of the oil well working condition according to the data in the sample database. By taking the measured electric power curve, torque factor, crank unbalance weight, walking beam unbalance weight, balance device balance weight, etc. as the inputs of a radial basis artificial neural network, a dynamometer card inversion model with a torque factor less than the threshold is established. By taking the measured electric power curve, torque factor, crank unbalance weight, walking beam unbalance weight, balance device balance weight, etc. as the inputs of a radial basis artificial neural network, a dynamometer card inversion model with a torque factor less than the threshold is established.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment big data analysis and diagnosis, and specifically provides a working condition diagnosis method that combines analytical solutions with exact solutions. Background Technique

[0002] At present, the dynamometer cards of pumping wells are obtained through load sensors, which have problems such as high cost, low popularity rate, and easy data drift and distortion, restricting the development of the oilfield Internet of Things construction and digital management. Electrical parameters are the most basic operating parameters of pumping wells, with advantages such as high popularity rate, low acquisition cost, and stable data. Using electrical parameters for the working condition diagnosis of pumping wells can avoid using load sensors and achieve low-cost and high-efficiency digital management of oil wells. There are problems with the dynamometer card test diagnosis being untimely and the test being cumbersome. The working condition diagnosis of pumping units basically starts from the basic definition of the dynamometer card of the pumping unit. Currently, by measuring the data of the polished rod load and displacement of the pumping unit, the analysis and drawing of the dynamometer card are carried out. This method has a long acquisition cycle, with a dynamometer card being collected once a month, and it is impossible to grasp the operating conditions of the oil well in real time. Moreover, the acquisition method is relatively cumbersome. There are problems with the oil well energy consumption test having a large workload, poor accuracy and timeliness of the energy consumption test, and a large error in the total power consumption of the mechanical production system. The liquid production unit consumption index recorded through a long-cycle test cannot reflect the true energy consumption index of the current mechanical production system. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a working condition diagnosis method that combines analytical solutions with exact solutions, solving the problem in the prior art that the data acquisition cycle of pumping units is long, with a dynamometer card being collected once a month, and it is impossible to grasp the operating conditions of the oil well in real time.

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: A working condition diagnosis method that combines analytical solutions with exact solutions, which is a combined diagnosis method mainly for the working conditions of oil wells, including the following diagnosis steps:

[0005] S1. First, establish a training sample library based on system dynamic simulation and comprehensively record and store the working condition data of the oil well;

[0006] S2. Establish a numerical simulation mathematical model of the oil well working conditions according to the data in the sample database;

[0007] S3. Analyze the measured motor input power curve according to the simulation mathematical model, establish simulation models for the instantaneous output power of the motor, the torque of the input shaft, and the torque of the crank load, and perform normalization of the load torque TF and the crank load torque factor M z and optimization and noise reduction of the crank load torque M z ;

[0008] S4. For the crank load torque M zAfter noise reduction optimization, a mathematical model for the analytical solution inverse dynamometer card can be established;

[0009] S5. According to the simulation model of the ground node force and energy parameter relationship and the polished rod load and crank load torque, a mathematical model for the accurate solution of the dynamometer card inversion (TF > a) can be determined and obtained;

[0010] S6. When the torque factor is greater than a, the accurate solution model is used to invert the polished rod dynamometer card. When the torque factor is less than or equal to a, the analytical solution model of the radial basis neural network is used to invert the polished rod dynamometer card;

[0011] S7. Use Matlab to compile the mathematical model of the accurate solution of the dynamometer card inversion (TF > a) into an application software;

[0012] S8. According to the established application software, perform the inversion process, calculate the final polished rod load and polished rod displacement, and draw the polished rod dynamometer card.

[0013] Preferably, in the step S2, the oil well working condition numerical values include the oil well pump fullness working condition, the traveling valve leakage working condition, the fixed valve leakage working condition, the plunger leakage working condition, the plunger out of the pump barrel working condition, the upper bump pump working condition, the lower bump pump working condition, the sucker rod string breakage and disconnection working condition, the oil well wax deposition working condition, the gas influence working condition, and the insufficient liquid supply working condition numerical values.

[0014] Preferably, in the step S3, when normalizing the load torque TF and the crank load torque factor M z at the three zero points of the torque factor TF, the crank load torque M z should also be zero. The three zero points of the crank load torque and torque factor curve do not necessarily exactly correspond. In order to avoid affecting the fitting accuracy of the polished rod load, it is necessary to optimize and reduce the noise of the crank load torque.

[0015] Preferably, in the step S4, the mathematical model of the analytical solution inverse dynamometer card includes the following formula:

[0016]

[0017] where a is the threshold of the load torque factor; f is the radial basis neural network model; TF represents the length of the front arm of the walking beam; M Z ' represents Z the sucker rod string of the level; Z represents the level of the oil well to which it belongs; B represents the pumping frequency of the oil pump; Q y represents the mass of a single crank block; g represents the length of the connecting rod; l represents the length of the rear arm of the walking beam; p represents the number of crankshaft balance blocks of the oil well pump; q represents the offset angle of the crankshaft.

[0018] Preferably, in the step S5, the mathematical model of the accurate solution of the dynamometer card inversion (TF > a) includes the following formula:

[0019]

[0020] wherein, i MB represents the horizontal projection of the MB base rod; P out represents the return work of the oil well pump.

[0021] Preferably, in the step S8: the inversion process includes the following steps:

[0022] I. Select the well number on the software interface, then select the data test date corresponding to the well number and the test serial number corresponding to this date, retrieve the structural parameters, mass parameters, oil well parameters and rod string parameters of the beam pumping unit in the database, and read the measured data corresponding to the test serial number at this time;

[0023] II. Substitute the input parameters into the ground node force and energy parameter relationship simulation model and the beam pumping unit motion simulation model to obtain the load torque M Z at the crank and the motion law of the polished rod and the torque factor TF curve;

[0024] III. Use the crank load torque optimization model to optimize and reduce the noise of the load torque at the crank calculated in step II, and the torque factor threshold a;

[0025] IV. Segment the crank load torque curve according to the torque factor threshold a, use the exact model to calculate the polished rod load corresponding to this part. For the part less than the threshold, i.e., |TF|≤a, input the determined sensitive parameters into the trained radial basis neural network to obtain the corresponding polished rod load, and then obtain the polished rod load curve for the entire cycle;

[0026] V. Draw the polished rod dynamometer card according to the calculated polished rod load and polished rod displacement.

[0027] Preferably, in step I of the back-calculation process, when reading the measured data corresponding to the test serial number at this time, it is necessary to check whether the parameters have been changed, and directly input the corrected value in the input box for the parameters that need to be corrected.

[0028] Preferably, in step IV of the back-calculation process, the part greater than the torque factor threshold a is |TF|>a, and the part less than the torque factor threshold a is |TF|≤a.

[0029] The present invention provides a working condition diagnosis method combining analytical solution and exact solution. It has the following beneficial effects:

[0030] 1. By analyzing the measured motor input power curve, the present invention establishes simulation models for the instantaneous output power of the motor, the torque of the input shaft, and the torque of the crank load. Based on this, an inversion model for accurately solving the polished rod dynamometer card is established. Through a radial basis function artificial neural network system, the measured electric power curve, torque factor, crank unbalance weight, walking beam unbalance weight, balance weight of the balance device, etc. are used as the inputs of the radial basis function artificial neural network, and an inversion model of the dynamometer card when the torque factor is less than the threshold is established.

[0031] 2. The present invention compiles relevant simulation software through Matlab, develops a polished rod dynamometer card inversion software based on measured electrical parameters, and applies this software for dynamometer card inversion. By comparing and plotting the measured electric power, the measured polished rod dynamometer card, and the inverted polished rod dynamometer card, the average error of the maximum load at the polished rod is 2.55%, and the maximum error is 9.6%; the average error of the minimum load at the polished rod is 6.1%, and the maximum error is 13.22%; the average error of the dynamometer card area is 9.88%, and the maximum error is 19.91%. The accuracy of this software is verified through on-site analysis, and the data inversion results all meet the technical requirements and satisfy engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a schematic diagram of the torque M of the load at the motor crank of the present invention Z and the change curves of the polished rod motion law and the torque factor TF;

[0033] Figure 2 is a schematic diagram of the software system architecture of the present invention;

[0034] Figure 3 is a schematic diagram of the interface of the dynamometer card inversion program of the present invention;

[0035] Figure 4 is a schematic diagram of the measured motor input power curve of the present invention;

[0036] Figure 5 is a schematic diagram of the comparison between the inverted result and the measured result of the polished rod dynamometer card of the present invention;

[0037] Figure 6 is a schematic diagram of the measured motor input power curve of the present invention;

[0038] Figure 7 is a schematic diagram of the comparison between the inverted result and the measured result of the polished rod dynamometer card of the present invention;

[0039] Figure 8 is a schematic diagram of the measured motor input power curve of the present invention;

[0040] Figure 9 is a schematic diagram of the comparison between the inverted result and the measured result of the polished rod dynamometer card of the present invention.

[0041] Specific embodiments In the embodiments of the present invention

[0042] Next, in conjunction with the accompanying drawings, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. Embodiment

[0043] As Figures 1-3 shown, the embodiments of the present invention provide a working condition diagnosis method combining analytical solutions and exact solutions, which is mainly a combined diagnosis method for oil well working conditions, including the following diagnosis steps:

[0044] S1. First, establish a training sample library based on system dynamic simulation, and comprehensively record and store oil well working condition data;

[0045] S2. Establish a numerical simulation mathematical model of oil well working conditions according to the data in the sample database. The numerical values of oil well working conditions include oil well pump fullness working conditions, traveling valve leakage working conditions, fixed valve leakage working conditions, plunger leakage working conditions, plunger out of pump barrel working conditions, upper bump pump working conditions, lower bump pump working conditions, sucker rod string breakage and disconnection working conditions, oil well wax deposition working conditions, gas influence working conditions, and insufficient liquid supply working condition numerical values;

[0046] S3. Analyze the measured motor input power curve according to the simulation mathematical model, establish simulation models of motor instantaneous output power, input shaft torque, and crank load torque, and perform normalization of load torque TF and crank load torque factor M z and optimization and noise reduction of crank load torque M z In step S3, when normalizing load torque TF and crank load torque factor M z at the three zero points of torque factor TF, crank load torque M z should also be zero. The three zero points of the crank load torque and torque factor curve do not necessarily exactly correspond. In order to avoid affecting the fitting accuracy of the polished rod load, it is necessary to optimize and reduce the noise of the crank load torque;

[0047] S4. After optimizing and reducing the noise of crank load torque M z in step S4, the analytical solution inverse dynamometer diagram mathematical model includes the following formula:

[0048]

[0049] where a is the threshold of the load torque factor; f is the radial basis neural network model; TF represents the length of the front arm of the walking beam; M Z'The rod string representing the Z level; Z represents the level of the oil well to which it belongs; B represents the pumping speed of the oil pump; Q y represents the mass of a single crank block; g represents the length of the connecting rod; l represents the length of the rear arm of the walking beam; p represents the number of counterweights on the crankshaft of the oil well pump; q represents the offset angle of the crankshaft;

[0050] S5. According to the simulation model of the relationship between ground node force and energy parameters and the hook load and crank load torque, the mathematical model for determining and obtaining the exact solution (TF > a) of the dynamometer card inversion can be determined. In the S5 step, the mathematical model for the exact solution (TF > a) of the dynamometer card inversion includes the following formula:

[0051]

[0052] where i MB represents the horizontal projection of the MB base rod; P out represents the return work of the oil well pump;

[0053] S6. When the torque factor is greater than a, the exact solution model is used to invert the hook dynamometer card. When the torque factor is less than or equal to a, the radial basis neural network analytical solution model is used to invert the hook dynamometer card;

[0054] S7. Use Matlab to compile the mathematical model for the exact solution (TF > a) of the dynamometer card inversion into an application software;

[0055] S8. According to the established application software, perform the inversion process. In the S8 step: The inversion process includes

[0056] the following steps:

[0057] I. Select the well number on the software interface, then select the data test date corresponding to the well number and the serial number of the test corresponding to that date, retrieve the structural parameters, mass parameters, oil well parameters, and rod string parameters of the walking beam pumping unit of this well from the database, and read the measured data corresponding to the test serial number at this time;

[0058] II. Substitute the input parameters into the simulation model of the relationship between ground node force and energy parameters and the motion simulation model of the walking beam pumping unit to obtain the load torque M Z at the crank and the motion law of the hook and the torque factor TF curve;

[0059] III. Use the crank load torque optimization model to optimize and reduce the noise of the load torque at the crank calculated in step II, as well as the torque factor threshold a;

[0060] IV. Segment the crank load torque curve according to the torque factor threshold a, and use the exact model to calculate the polished rod load corresponding to this part. For the part where |TF| ≤ a, input the determined sensitive parameters into the trained radial basis neural network to obtain the corresponding polished rod load, and then obtain the polished rod load curve for the entire cycle.

[0061] V. Draw the polished rod dynamometer card based on the calculated polished rod load and polished rod displacement, and calculate the final polished rod load and polished rod displacement, and draw the polished rod dynamometer card.

[0062] As Figures 4-9 shown, the embodiment of the present invention provides a working condition diagnosis method combining analytical solution and exact solution, taking the graphical comparison and data analysis of the dynamometer card inversed from on-site measured electrical parameters and the measured dynamometer card as an example, and at the same time giving the measured data:

[0063]

[0064]

[0065]

[0066] Data comparison and analysis of the measured dynamometer card and the inverse dynamometer card

[0067] Conclusion: The measured electric power, the measured polished rod dynamometer card and the inverse polished rod dynamometer card are plotted for comparison. The average error of the maximum polished rod load is 2.55%, and the maximum error is 9.6%; the average error of the minimum polished rod load is 6.1%, and the maximum error is 13.22%; the average error of the dynamometer card area is 9.88%, and the maximum error is 19.91%. The accuracy of the software is verified through on-site analysis, and the data inversion results all meet the technical requirements and satisfy the engineering applications.

[0068] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A working condition diagnosis method combining analytical solution and exact solution, characterized in that: Regarding the oil well working condition as the main combined diagnosis method, it includes the following diagnosis steps: S1. First, establish a training sample library based on system dynamic simulation, and comprehensively record and store oil well working condition data; S2. Establish a numerical simulation mathematical model of oil well working conditions according to the data in the sample database; S3. Analyze the measured motor input power curve according to the simulation mathematical model, establish the simulation models of the motor instantaneous output power, input shaft torque, and crank load torque, and perform the normalization of the load torque TF and the crank load torque factor M z and the optimized noise reduction of the crank load torque M z ; S4. For the crank load torque M z After noise reduction optimization, an analytical solution inverse demonstration work diagram mathematical model can be established; In the S4 step, the analytical solution inverse dynamometer card mathematical model includes the following formula: ； Among them, a is the threshold value of the load torque factor; f is the radial basis neural network model; TF represents the length of the front arm of the walking beam; M Z ' represents the rod string of level Z; Z represents the level of the oil well to which it belongs; B represents the pumping speed of the oil pump; Q y represents the mass of a single crank block; g represents the length of the connecting rod; l represents the length of the rear arm of the walking beam; p represents the number of crankshaft balance blocks of the oil well pump; q represents the offset angle of the crankshaft; S5. According to the ground node force and energy parameter relationship simulation model and the polished rod load and crank load torque, the mathematical model of the exact solution of the inverse dynamometer card can be determined and obtained; In the S5 step, the mathematical model of the exact solution of the inverse dynamometer card includes the following formula: ； Among them, i MB represents the horizontal projection of the MB base rod; P out represents the useful return work of the oil well pump; S6. When the torque factor is greater than a, use the exact solution model to invert the polished rod dynamometer card. When the torque factor is less than or equal to a, use the radial basis neural network analytical solution model to invert the polished rod dynamometer card; S7. Use Matlab to compile the mathematical model of the exact solution of the inverse dynamometer card established into an application software; S8. According to the established application software, perform the inversion process, calculate the final polished rod load and polished rod displacement, and draw the polished rod dynamometer card.

2. A working condition diagnosis method combining analytical solution and exact solution according to claim 1, characterized in that: In the S2 step, the oil well working condition values include the oil well pump full condition, traveling valve leakage condition, fixed valve leakage condition, plunger leakage condition, plunger out of pump barrel condition, upper bump pump condition, lower bump pump condition, sucker rod string breakage condition, oil well wax deposition condition, gas influence condition, and insufficient liquid supply condition values.

3. A working condition diagnosis method combining analytical solution and exact solution according to claim 1, characterized in that: In the step S3, the load torque TF and the crank load torque factor M z When normalized, at the three zero points of the torque factor TF, the crank load torque M z should also be zero. The three zero points of the crank load torque and torque factor curve do not necessarily exactly correspond. In order to avoid affecting the fitting accuracy of the polished rod load, it is necessary to optimize and reduce the noise of the crank load torque.

4. A working condition diagnosis method combining analytical solution and exact solution according to claim 1, characterized in that: In the S8 step: The inversion process includes the following steps: I. Select the well number on the software interface, then select the data test date corresponding to the well number and the test serial number corresponding to that date, retrieve the beam pumping unit structure parameters, mass parameters, oil well parameters, and rod string parameters of the well from the database, and read the measured data corresponding to the test serial number at that time; II. Substitute the input parameters into the simulation model of the relationship between the force and energy parameters of the ground node and the kinematic simulation model of the walking beam pumping unit to obtain the load torque M at the crank z and the motion law of the polished rod and the torque factor TF curve; III. Optimize and reduce the noise of the load torque at the crank calculated in step II using the crank load torque optimization model, as well as the torque factor threshold a; IV. Segment the crank load torque curve according to the torque factor threshold a, use the exact model to calculate the corresponding polished rod load for this part. For the part less than the threshold, i.e., |TF|≤a, input the determined sensitive parameters into the trained radial basis neural network to obtain the corresponding polished rod load, and then obtain the polished rod load curve for the entire cycle; V. Draw the polished rod dynamometer card according to the calculated polished rod load and polished rod displacement.

5. A working condition diagnosis method combining analytical solution and exact solution according to claim 4, characterized in that: In the I step of the inverse calculation process, when reading the measured data corresponding to the test serial number at that time, it is necessary to check whether the parameters have been changed. For the parameters that need to be corrected, directly input the corrected values in the input box.

6. A working condition diagnosis method combining analytical solution and exact solution according to claim 4, characterized in that: In step IV of the back-calculation process, the part greater than the torque factor threshold a is |TF| > a, and the part less than the torque factor threshold a is |TF| ≤ a.

Images