A pumping unit well standard diagram calibration method based on unary distance clustering

Abstract

The present application relates to the technical field of crude oil production, and particularly relates to a pumping unit well standard dynamogram calibration method based on unary distance clustering, which comprises collecting real-time dynamogram data of the oil well, performing abnormal data cleaning on the real-time dynamogram data, the abnormal data comprising zero value, null value and negative value, determining the most stable and highest frequency dynamogram through clustering analysis, and calibrating the most stable and highest frequency dynamogram as a standard dynamogram. The present application can improve the accuracy of the standard dynamogram, reduce the setting time of the standard dynamogram, and improve the work efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crude oil production, in particular to a pumping unit well standard diagram calibration method based on one-dimensional distance clustering. BACKGROUND

[0002] A dynamometer card is a diagram that reflects the working condition of a deep well pump, is measured by a special instrument, and is drawn on a coordinate graph. The area enclosed by the closed line segment represents the work done by the pumping unit in one reciprocating motion.

[0003] In the field of crude oil production, the dynamometer card of a pumping unit well directly reflects the working condition of the oil pump. Because of different geological conditions and production systems, the shapes of the dynamometer cards of each well are different, and they are all direct manifestations of the working condition of the oil well.

[0004] By analyzing the dynamometer card, the working state of the oil pump and whether the parameters of the oil pumping equipment are normal can be understood. In order to monitor the working condition of the oil well, a standard diagram needs to be set for each well as a comparison basis to analyze the changes in the diagram at different times, thereby providing a basis for production management personnel.

[0005] At present, the setting of the standard diagram is manually set, and the standard diagram is manually selected and calibrated from a large number of real-time diagrams. Because of different technical experiences, different diagrams are set. At the same time, when the dynamometer card changes greatly, the standard diagram cannot be adjusted in time, which will affect the analysis of the production management personnel, and problems cannot be accurately and timely found. SUMMARY

[0006] The purpose of the present application is to solve the problem of inaccurate standard diagram in the prior art, and to provide a pumping unit well standard diagram calibration method based on one-dimensional distance clustering.

[0007] To solve the problem of inaccurate standard diagram, the present application uses big data method to analyze the data rule from real-time time series data, and realizes the technical problem of intelligent calibration of standard diagram.

[0008] The technical scheme is as follows:

[0009] A pumping unit well standard diagram calibration method based on one-dimensional distance clustering, comprising collecting real-time diagram data of the oil well, further comprising the following steps:

[0010] a: cleaning the abnormal data of the real-time diagram data, the abnormal data including zero value, null value and negative value;

[0011] b: determining the most stable and highest frequency diagram through clustering analysis, and calibrating it as a standard diagram.

[0012] Further, the step a performs abnormal cleaning on the discrete data through box plot analysis method.

[0013] Further, the discrete data includes a hystogram area, a maximum load or a minimum load.

[0014] Further, the step b includes:

[0015] b1: hystogram parameter disassembly, each hystogram is disassembled into a hystogram area S, a maximum load Zmax and a minimum load Zmin;

[0016] b2: dimensionless processing, each hystogram is dimensionless processed:

[0017] each hystogram is dimensionless processed:

[0018] Formula 1

[0019] a dimensionless array D=[,, …] is formed; , ,……];

[0020] b3: distance difference between each data point and the initial point is calculated, taking the initial point as the starting point, the distance difference between and is calculated in turn, and a distance difference array ΔD is formed: to, and a distance difference array ΔD is formed:

[0021] ΔD=[,, … , ,…… ] Formula 2;

[0022] b4: the cluster number K is defined, and the cluster number K is 3-5;

[0023] b5: a traversal radius R is calculated;

[0024] b6: the number of points within R range of each dimensionless point is traversed;

[0025] b7: the maximum value in N is taken:

[0026] Formula 5;

[0027] b8: the data index in the array D is backtracked, and the points in the original data are determined.

[0028] Further, the cluster number K is 3.

[0029] Further, the step b5 specifically includes:

[0030] R= Formula 3.

[0031] Further, the step b6 specifically includes:

[0032] ​​If , the xth point is included in the counting point , forming a counting point array N:

[0033] Formula 4.

[0034] Further, the step b8 specifically includes:

[0035] By backtracking The subscript in N, the subscript index of the calibrated work graph in D is determined.

[0036] The present application has the following advantages:

[0037] Taking the Shengli Oilfield Management Area of Sinopec as an example, an average of 200 pumping wells are managed in each management area. The effects achieved by the present application mainly include:

[0038] 1) Improve the accuracy of standard work graph.

[0039] Instead of manual setting, based on real-time data, the accuracy is improved by 50%.

[0040] 2) Reduces the standard work graph setting time and improves the work efficiency.

[0041] The original manual setting needs to be updated regularly, and it takes 2 hours to manually set the standard work graph for a management area. After using the intelligent algorithm, it can be completed within 1 minute, and the efficiency is improved by more than 99%. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 is a technical architecture diagram;

[0043] Figure 2 is a work graph data cleaning flowchart;

[0044] Figure 3 is a box plot analysis method principle diagram;

[0045] Figure 4 is a one-dimensional distance clustering flowchart;

[0046] DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the present application clearer and more clear, the present application will be further described in detail below in combination with specific embodiments and with reference to the drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present application. In addition, in the following description, the description of known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present application.

[0048] Example 1:​

[0049] Standard diagram intelligence calibration, need to use a large number of power diagram real-time data, from which to select the highest frequency, the most stable power diagram. As a graph, the power diagram needs to analyze the parameters of the power diagram, and use the relationship between the parameters to calibrate the standard power diagram or stable power diagram. Because of the different production conditions of oil wells, it cannot be calibrated by standard power diagram in the standard sense, but needs to be different for each well, and needs to be calibrated separately according to the production status of each well.

[0050] A standard power diagram calibration method for pumping unit wells, comprising the following steps:

[0051] Collecting real-time power diagram data of several single wells;

[0052] Decomposing the parameters of single well real-time power diagram data into power diagram area, maximum load and minimum load;

[0053] Abnormal data cleaning, abnormal data mainly includes

[0054] Power diagram abnormal data cleaning zero value, null value, negative value or other abnormal data;

[0055] Through linear clustering analysis method, the most stable and highest frequency power diagram is taken as the standard power diagram.

[0056] Example two:

[0057] Please refer to Figure 1 A standard power diagram calibration method for pumping unit wells based on one-dimensional distance clustering, including collecting real-time power diagram data of oil wells and the following steps:

[0058] 1. Power diagram abnormal data cleaning.

[0059] 1) Cleaning of unmeasured abnormal data.

[0060] For the power diagram area parameter in the power diagram data, data cleaning is performed for various unmeasured conditions, including zero value, null value and negative value.

[0061] 2) Discrete data cleaning.

[0062] Please refer to Figures 2-3 Discrete data refers to data with large discrete degree in power diagram parameters such as power diagram area, maximum load and minimum load. Box plot analysis method is used to realize abnormal cleaning.

[0063] Through cleaning, the power diagram in which the abnormal value in the power diagram parameter data is cleaned, and the stable running data is retained.

[0064] 2. One-dimensional distance clustering calibration.

[0065] Please refer to Figure 4The ratio of the area of the power diagram and the difference between the maximum and minimum load in the power diagram can be used respectively.

[0066] 1) Power diagram parameter disassembly

[0067] Each power diagram is disassembled into the power diagram area S, the maximum load Zmax, and the minimum load Zmin.

[0068] 2) Dimensionless processing

[0069] Each power diagram is dimensionless processed:

[0070] Formula 1

[0071] A dimensionless array D=[ is formed. , ,…… ].

[0072] 3) Calculate the distance difference between each data point and the initial point

[0073] Take the initial point as the starting point, and calculate the distance difference between to and , forming a distance difference array ΔD.

[0074] =[ ΔD , ,…… ] Formula 2

[0075] 4) Define the number of clusters K

[0076] Input the number of clusters K, generally 3-5, and the default is 3.

[0077] 5) Calculate the traversal radius R

[0078] R = Formula 3

[0079] 6) Traverse the number of points within R range for each dimensionless point

[0080] The judgment rule is that, taking as an example, < , the xth point is included in the counting point . Form a counting point array N.

[0081] Formula 4

[0082] 7) Take the maximum value in N​

[0083] Formula 5

[0084] 8) Backtrack the data index in array D to determine the point in the original data

[0085] By backtracking The index in D, determine the index of the calibrated work graph in D.

[0086] Through the above steps, the most stable and highest frequency work graph is determined based on the original work graph data set, and the standard work graph is calibrated.

[0087] Although embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims.

Claims

1. A pumping unit well standard dynamometer card calibration method based on a unary distance clustering, comprising collecting real-time dynamometer card data of the oil well, characterized in that, Further comprising the following steps: a: abnormal data cleaning on real-time work graph data, the abnormal data including zero value, null value and negative value; b: determining the most stable and highest frequency work graph through cluster analysis, and marking as standard work graph; The step a performs abnormal cleaning on discrete data through box plot analysis method; The step b includes: b1: work graph parameter disassembly, decomposing each work graph into work graph area S, maximum load Zmax and minimum load Zmin; b2: dimensionless treatment, performing dimensionless treatment on each work graph: Performing dimensionless treatment on each work graph: d = S / (Z max - Z min ) Equation 1 Form an array of dimensionless numbers, D = [d1, d2,... d n ] b3: Calculate the distance difference between each data point and the initial point, starting from d1, and then calculate the distance difference between d2 and d n and d1, forming a distance difference array ΔD: ΔD = [Δd1, Δd2,... Δdn] Formula 2 n ] Formula 2 b4: defining cluster number K, the cluster number K being 3-5; b5: calculating traversal radius R; b6: traversing the number of points of each dimensionless point within R range; b7: taking the maximum value in N: n max = max (N) Equation 5; b8: backtracking the data index in array D to determine the point in original data; The step b5 specifically includes: R = (Ad n - Ad1) / K Equation 3.

2. The method of claim 1, wherein, The discrete data includes work graph area, maximum load or minimum load.

3. The method of claim 2, wherein, The cluster number K is 3.

4. The method of claim 3, wherein, The step b6 specifically includes: determination rule, if |d x -d1|<|d1+R|, the xth point is included in the counting point n1, forming a counting point array N: N = [n1, n2,... n m ] Equation 4.

5. The method of claim 3, wherein, The step b8 specifically includes: By backtracking n max The subscript in N, determines the subscript index of the calibrated work graph in D.

Images