A method for extracting and diagnosing power diagram features based on Freeman chain code

Abstract

The present invention proposes a method for extracting and diagnosing the power diagram feature based on the Freeman chain code. The power diagram on the surface of the oil well is analyzed, the pump power diagram is converted, and the chain code diagram corresponding to the power diagram is established based on the Freeman chain code. The key information in the chain code is extracted by feature extraction of the chain code diagram, and a total of 12 feature vectors are extracted. The common working conditions of the oil field are combined with the extracted feature vectors to establish a typical working condition-feature pattern set. Through the pattern set, the judgment rules of each typical working condition are given, and a working condition evaluation system is established to give the occurrence degree of each feature of the power diagram. Rapid diagnostic modeling and accurate analysis of working conditions are realized, which is convenient for the management of oil field personnel and realizes the refined management of pumping wells.

Description

Technical Field

[0001] The invention relates to the field of oilfield pumping well production modes, in particular to a freeman chain code-based work diagram feature extraction and diagnosis method. Background Art

[0002] With the deepening of oil field exploitation, the increase in the number of oil wells, and the emergence of complex situations such as high viscosity, high sand, water flooding and strong corrosion, the probability of accidents in rod pumping systems is increasing. Therefore, it is of great significance to develop a fault diagnosis and evaluation system for rod pumping wells. The system monitors the data and parameters fed back by the pumping system, realizes real-time analysis and judgment of the data, diagnoses the cause of the system failure, and takes timely measures for the oil wells where accidents are found, thereby reducing the economic losses caused by the accidents.

[0003] At present, some achievements have been made in the research on computer intelligent diagnosis methods for downhole working conditions of rod pumping systems, such as expert systems, artificial neural networks, rough set theory, support vector machines, spectrum analysis and filtering technology. However, all of the above methods need to be based on working condition model libraries. It is difficult to obtain enough learning samples in the early stage, resulting in a long period of establishing working condition model libraries, which leads to low early diagnosis accuracy. In addition, traditional working condition analysis methods simply give a diagnosis result, and do not make an overall and detailed evaluation of the collected power diagram. On-site operators can only make judgments based on the results given, lacking a self-learning process. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a method for extracting and diagnosing a power diagram feature based on the Freeman chain code. The surface power diagram of the oil well is analyzed and converted into a pump power diagram. A chain code diagram corresponding to the power diagram is established based on the Freeman chain code. Feature extraction is performed on the chain code diagram to extract key information in the chain code. A total of 12 feature vectors are extracted. The common working conditions of the oil field are combined with the extracted feature vectors to establish a typical working condition-feature pattern set. Through the pattern set, judgment rules for each typical working condition are given, a working condition evaluation system is established, and the occurrence degree of each feature of the power diagram is given.

[0005] The technical solution adopted by the present invention is as follows: a method for extracting and diagnosing the power diagram features based on Freeman chain code, comprising the following steps:

[0006] Collect oil well data through the dynamometer to obtain the dynamometer diagram of each oil well;

[0007] Extracting power diagram features from the power diagram;

[0008] The operating conditions of the oil well are evaluated based on the judgment rules and operating conditions of the occurrence degree of each characteristic.

[0009] The extracting of the power diagram features from the power diagram is to extract the power diagram features through the Freeman chain code.

[0010] The power diagram features include: the degree of jaggedness, the degree of protrusion of the lower left corner, the degree of protrusion of the upper right corner, the degree of flatness, the degree of missing of the upper left corner, the degree of missing of the upper right corner, the degree of missing of the lower right corner, the degree of missing of the lower left corner, the smoothness of the upstroke, the degree of sharp unloading of the lower right corner, the smoothness of the downstroke, and the degree of obesity.

[0011] The power diagram features include:

[0012] (1) The degree of sawtooth shape is used to determine the degree of change in the load transfer point between the floating and fixed valve opening and closing points, that is:

[0013]

[0014] Among them, C i is the boundary chain code of the pump performance diagram, and the jagged degree p1 is:

[0015]

[0016] Where n is the total number of chain code pump work diagrams;

[0017] (2) The protrusion degree of the lower left corner is used to judge the slope of the displacement load between the middle of the lower stroke and the bottom dead center of the stroke. When the chain code is continuously greater than or equal to 5 and exceeds the set value, the lower left corner is considered to be protruding; the protrusion degree p2 of the lower left corner is as follows:

[0018]

[0019] Among them, c0~c7 are the numbers corresponding to the 0-7 order chain codes of the pump work diagram. If the number of chain codes that are continuously greater than or equal to 5 is greater than 10, f2 is set to 1, otherwise it is 0. f2 is the flag bit indicating whether the number of chain codes that are continuously greater than or equal to 5 is greater than 10;

[0020] (3) The protrusion degree of the upper right corner is used to judge the slope of the displacement load between the middle of the upper stroke and the top dead center of the stroke. When the chain code is less than or equal to 3 consecutive times and exceeds the set value, it is considered that the upper left corner is protruding; the protrusion degree p3 of the upper right corner is:

[0021]

[0022] Among them, if the number of chain codes that are consecutively less than or equal to 3 is greater than 10, f3 is set to 1, otherwise it is 0. f3 is the identification bit that indicates whether the number of chain codes that are consecutively less than or equal to 3 is less than 10;

[0023] (4) Flatness: It is used to determine the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram. The flatness p4 is:

[0024]

[0025] Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, L lmin is the minimum load of the theoretical work diagram;

[0026] (5) The degree of missing upper left corner is used to judge the rising curvature from the middle of the upstroke load to the closing point of the floating valve. When the 0th and 1st order chain codes exceed the threshold, it is considered that there is a missing in the upper left corner. The degree of missing upper left corner p5 is defined as:

[0027] p5 = (c0 + 0.5 * c1 - 0.5 * c2) / n;

[0028] (6) The degree of missing in the upper right corner is used to judge the descending curvature from the middle of the upper stroke to the fixed valve closing point. When the 6th and 7th order chain codes exceed the threshold, it is considered that there is a missing in the upper right corner. The degree of missing in the upper right corner p6 is defined as:

[0029] p6=(1.5*c6+c7+0.01*c0-c1-1.5*c2) / n

[0030] (7) The degree of missing in the lower right corner is used to judge the slope of the curve from the top dead center to the opening point of the floating valve. When the 4th and 5th order chain codes exceed the threshold, it is considered that there is a missing in the lower right corner. The degree of missing in the lower right corner p7 is defined as:

[0031] p7=(1.5*c4+0.5*c5-0.5*c6-c2-0.5*c3) / n

[0032] (8) The missing degree of the lower left corner is used to judge the slope of the curve from the middle of the downstroke to the fixed valve opening point. When the 2nd and 3rd order chain codes exceed the threshold, it is considered that there is a missing in the lower left corner. The missing degree of the lower left corner p8 is defined as:

[0033] p8=(1.5*c2+c3+0.01*c4-c5-1.5*c6) / n

[0034] (9) The smoothness of the upstroke is used to determine the degree of change of the curve from the floating valve closing point to the fixed valve closing point. When the chain code of other orders exceeds the threshold except the 0th order chain code, the upstroke is considered to be not smooth. The smoothness of the upstroke p9 is:

[0035]

[0036] (10) The degree of rapid unloading at the lower right is used to determine the degree of unloading from the top dead center to the opening point of the floating valve. When the 2nd and 3rd order chain codes exceed the threshold, it is considered that the lower right is rapidly unloaded. The degree of rapid unloading at the lower right is p10 Defined as:

[0037] p 10 =(3*c3+3*c4-0.5*c2) / n

[0038] (11) Downstroke smoothness is used to determine the degree of change of the curve from the floating valve opening point to the fixed valve opening point. When the chain code other than the 0th order chain code exceeds the threshold, the downstroke is considered to be not smooth. The downstroke smoothness p9 is:

[0039]

[0040] (12) Obesity level, used to determine the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram. The obesity level p3 is:

[0041]

[0042] Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, L lmin It is the minimum load of the theoretical power diagram.

[0043] The evaluation of the working condition of the oil well is as follows:

[0044] List all working conditions and the eigenvectors corresponding to each working condition;

[0045] Construct a typical working condition-feature pattern set, and query the feature vectors that satisfy the pattern set in each working condition in turn;

[0046] Based on the judgment rules satisfied by the samples and combined with the typical working condition-characteristic pattern set, the sample working conditions are diagnosed and evaluated.

[0047] The present invention has the following effects and advantages:

[0048] 1. Under current technical conditions, the acquisition of oil well surface performance diagrams has been realized in real time, so the use of performance diagrams for diagnosis can ensure accurate analysis of single well conditions;

[0049] 2. The Freeman chain code-based method is used to analyze the working conditions of a single well, which changes the disadvantages of long period and heavy workload of establishing an expert database in the past, and low diagnostic accuracy in a short period of time, and realizes rapid and accurate judgment of the working conditions of a single well;

[0050] 3. Based on the Freeman chain code, the characteristic vector of the performance diagram is re-established, and the specific diagnosis and evaluation results of the single well performance diagram are given through quantitative analysis of the characteristic vector, which is convenient for on-site oilfield staff to view. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figures 1 to 12 The 12 characteristic vector analysis diagrams of the present invention are, in order, the degree of sawtooth, the degree of lower left protrusion, the degree of upper right protrusion, the degree of flatness, the degree of missing upper left corner, the degree of missing upper right corner, the degree of missing lower right corner, the degree of missing lower left corner, the degree of smoothness of the upstroke, the degree of sharp unloading of the lower right corner, the degree of smoothness of the downstroke, and the degree of obesity;

[0052] Fig.13a The normalized pump performance diagram sample 1 is obtained after the performance diagram of the pumping well to be diagnosed is converted;

[0053] Fig.13b The normalized pump performance diagram sample 2 is obtained after the performance diagram of the pumping well to be diagnosed is converted;

[0054] Fig.13c The normalized pump performance diagram sample 3 after the conversion of the performance diagram of the pumping well to be diagnosed;

[0055] Fig.14a It is the boundary chain code conversion diagram of sample 1;

[0056] Fig.14b This is the boundary chain code conversion diagram of sample 2;

[0057] Fig.14c This is the boundary chain code conversion diagram of sample 3. DETAILED DESCRIPTION

[0058] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0059] Aiming at the technical requirements of power diagram diagnosis, the present invention designs a new oil field single well diagnosis and analysis method. The present invention collects oil well data through a power indicator to obtain the power diagram of each oil well; analyzes the surface power diagram of the oil well, converts the pump power diagram, establishes a chain code diagram corresponding to the power diagram based on the Freeman chain code, extracts key information from the chain code by feature extraction of the chain code diagram, extracts 12 feature vectors in total, and combines the common working conditions of the oil field with the extracted feature vectors to establish a typical working condition-feature pattern set. Through the pattern set, the judgment rules of each typical working condition are given, the working condition evaluation system is established, and the occurrence degree of each feature of the power diagram is given. It changes the previous traditional mode that the working condition diagnosis needs to rely on the expert knowledge base, and the establishment of the expert knowledge base is a long-term limitation, which leads to a large error in the early stage of the working condition diagnosis. The present invention can realize the accurate analysis of the working condition, wherein the evaluation system can more intuitively reflect the actual situation of the power diagram, facilitate the management of oil field personnel, and realize the refined management of pumping wells.

[0060] 1. Establish a work graph feature extraction strategy based on Freeman chain code, such as Figures 1 to 12 shown.

[0061] (1) Degree of serration

[0062] It mainly determines the degree of change of the load transfer point between the movable and fixed valve opening and closing points. Here, when the degree of change is greater than 90°, the chain code is considered to be sawtooth, that is:

[0063]

[0064] Among them, C i is the boundary chain code of the pump performance diagram, and the jagged degree p1 is defined as:

[0065]

[0066] Where n is the total number of chain code pump diagrams.

[0067] (2) Prominence of the lower left corner

[0068] It is mainly determined by judging the slope of the displacement load between the middle of the lower stroke and the lower dead point of the stroke. Here, when there are many chain codes that are continuously greater than or equal to 5, it is considered that the lower left corner is protruding. The degree of protrusion of the lower left corner is defined as:

[0069]

[0070] Among them, c1~c7 are the numbers corresponding to the 1-7 order chain codes of the pump work diagram. If the number of chain codes that are consecutively greater than or equal to 5 is greater than 10, f2 is set to 1, otherwise it is 0.

[0071] (3) Prominence of the upper right corner

[0072] It is mainly determined by judging the slope of the displacement load between the middle of the upper stroke and the top dead center of the stroke. Here, it is considered that when there are many chain codes less than or equal to 3, the upper left corner is considered to be protruding. The degree of protrusion of the upper left corner is defined as:

[0073]

[0074] Among them, if the number of chain codes that are consecutively less than or equal to 3 is greater than 10, f3 is set to 1, otherwise it is 0.

[0075] (4) Flatness

[0076] The flatness is mainly determined by judging the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram. The flatness is defined as:

[0077]

[0078] Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, Llmin It is the minimum load of the theoretical power diagram.

[0079] (5) Degree of missing upper left corner

[0080] It is mainly determined by judging the rising curvature from the middle of the upstroke load to the closing point of the floating valve. Here, it is considered that when there are more 0-order and 1-order chain codes, it is considered that there is a missing in the upper left corner. The degree of missing in the upper left corner is defined as:

[0081] p5=(c0+0.5*c1-0.5*c2) / n

[0082] (6) Degree of missing upper right corner

[0083] It is mainly determined by judging the descending curvature from the middle of the upper stroke to the fixed valve closing point. Here, the influence of the 0th order chain code needs to be eliminated. When there are more 6th and 7th order chain codes, it is considered that there is a missing in the upper right corner. The missing degree of the upper right corner is defined as:

[0084] p6=(1.5*c6+c7+0.01*c0-c1-1.5*c2) / n

[0085] (7) Degree of missing lower right corner

[0086] It is mainly determined by judging the slope of the curve from the top dead center to the opening point of the floating valve. Here, it is considered that when there are more 4th and 5th order chain codes, it is considered that there is a missing in the lower right corner. The missing degree of the lower right corner is defined as:

[0087] p7=(1.5*c4+0.5*c5-0.5*c6-c2-0.5*c3) / n

[0088] (8) Degree of missing lower left corner

[0089] It is mainly determined by judging the slope of the curve from the middle of the downstroke to the fixed valve opening point. Here, when there are more 2nd and 3rd order chain codes, it is considered that there is a missing in the lower left corner. The missing degree of the lower left corner is defined as:

[0090] p8=(1.5*c2+c3+0.01*c4-c5-1.5*c6) / n

[0091] (9) Upstroke smoothness

[0092] It is mainly determined by judging the degree of change of the curve from the floating valve closing point to the fixed valve closing point. Here, it is considered that when there are more chain codes of other orders except the 0th order chain code, the upstroke is not smooth. The degree of smoothness of the upstroke is defined as:

[0093]

[0094] (10) Rapid unloading degree at the lower right

[0095] It is mainly determined by judging the degree of unloading from the top dead center to the opening point of the floating valve. Here, it is considered that when there are more 2nd and 3rd order chain codes, it is considered that the lower right is unloaded sharply. The degree of lower right sharp unloading is defined as:

[0096] p 10 =(3*c3+3*c4-0.5*c2) / n

[0097] (11) Downstroke smoothness

[0098] It is mainly determined by judging the degree of change of the curve from the floating valve opening point to the fixed valve opening point. Here, it is considered that when there are more chain codes except the 0th order chain code, the downstroke is not smooth. The degree of downstroke smoothness is defined as:

[0099]

[0100] (12) Obesity level

[0101] Mainly by judging the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram, the degree of obesity is defined as:

[0102]

[0103] Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, L lmin It is the minimum load of the theoretical power diagram.

[0104] 2. Give the judgment rules for the occurrence degree of each feature

[0105] The judgment rules for the occurrence degree of the 12 eigenvectors are established as follows:

[0106]

[0107] When the power diagram feature vector satisfies the judgment rule z1 in the table, it indicates that there is a very high probability of the feature appearing. When the power diagram feature vector satisfies the judgment rule z2, it indicates that there is a certain probability of the feature appearing.

[0108] 3. Give typical working conditions-characteristic mode set

[0109] The pattern set for establishing typical operating conditions and characteristics is as follows:

[0110]

[0111] Among them, u1,u2,...,u 12 are 12 eigenvectors, g1,g2,...,g 13These are 13 typical working conditions commonly seen in oil fields, namely: gas impact, insufficient fluid supply, sucker rod breakage, thick oil, floating valve loss, fixed valve loss, double valve loss, pump collision, pump collision, sand production, gap loss, wax deposition, and plunger disengagement. 0 represents that the working condition will not have this feature, and 1 represents that the working condition will have this feature. Taking floating valve loss g5 as an example, its main features are missing upper left corner, missing upper right corner, and uneven upstroke, that is, u5, u6, and u9 in the feature vector are all 1.

[0112] 4. Establish a working condition evaluation system

[0113] Example:

[0114] like Figure 13a to Figure 13c , which are the normalized pump performance diagrams after conversion of the three pumping well performance diagrams to be diagnosed.

[0115] Convert it to boundary chain code and get the result as follows Figure 14a to Figure 14c .

[0116] Extract its 12 eigenvectors, the specific values ​​are as follows:

[0117]

[0118] Combined with the judgment rules of the occurrence degree of each feature, typical operating conditions and characteristic pattern sets, the actual operating conditions with diagnostic samples are given: the characteristic vectors of sample 1 are not in the possible judgment rules, so the sample operating condition is normal operation; sample 2 features 7, 10, and 11 meet the judgment rule z1, combined with the typical operating condition-characteristic pattern set, it can be determined that the sample operating condition is insufficient fluid supply; sample 3 features 5, 6, and 9 meet the judgment rule z1, combined with the typical operating condition-characteristic pattern set, it can be determined that the sample operating condition is floating valve leakage.

[0119] The working condition evaluation system is established as follows:

[0120]

Claims

1. A method for extracting and diagnosing power diagram features based on Freeman chain code, characterized in that: The following steps are involved: Collect oil well data through the dynamometer to obtain the dynamometer diagram of each oil well; Extracting power diagram features from the power diagram; Diagnose and evaluate the working condition of the oil well according to the judgment rules and working conditions of the occurrence degree of each characteristic; The power diagram features include: jaggedness, protrusion of the lower left corner, protrusion of the upper right corner, flatness, missing degree of the upper left corner, missing degree of the upper right corner, missing degree of the lower right corner, missing degree of the lower left corner, smoothness of the upstroke, sharp unloading degree of the lower right corner, smoothness of the downstroke, and fatness, specifically: (1) The degree of sawtooth shape is used to determine the degree of change in the load transfer point between the floating and fixed valve opening and closing points, that is: Among them, C i is the boundary chain code of the pump performance diagram, and the jagged degree p1 is: Where n is the total number of chain code pump work diagrams; (2) The protrusion degree of the lower left corner is used to judge the slope of the displacement load between the middle of the lower stroke and the bottom dead center of the stroke. When the chain code is continuously greater than or equal to 5 and exceeds the set value, the lower left corner is considered to be protruding; the protrusion degree p2 of the lower left corner is as follows: Among them, c0~c7 are the numbers corresponding to the 0-7 order chain codes of the pump work diagram. If the number of chain codes that are continuously greater than or equal to 5 is greater than 10, f2 is set to 1, otherwise it is 0. f2 is the flag bit indicating whether the number of chain codes that are continuously greater than or equal to 5 is greater than 10; (3) The protrusion degree of the upper right corner is used to judge the slope of the displacement load between the middle of the upper stroke and the top dead center of the stroke. When the chain code is less than or equal to 3 consecutive times and exceeds the set value, it is considered that the upper left corner is protruding; the protrusion degree p3 of the upper right corner is: If the number of chain codes that are consecutively less than or equal to 3 is greater than 10, f3 is set to 1, otherwise it is 0. f3 is the flag bit that indicates whether the number of chain codes that are consecutively less than or equal to 3 is less than 10. (4) Flatness: It is used to determine the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram. The flatness p4 is: Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, L lmin is the minimum load of the theoretical work diagram; (5) The degree of missing upper left corner is used to judge the rising curvature from the middle of the upstroke load to the closing point of the floating valve. When the 0th and 1st order chain codes exceed the threshold, it is considered that there is a missing in the upper left corner. The degree of missing upper left corner p5 is defined as: p5 = (c0 + 0.5 * c1 - 0.5 * c2) / n; (6) The degree of missing in the upper right corner is used to judge the descending curvature from the middle of the upper stroke to the fixed valve closing point. When the 6th and 7th order chain codes exceed the threshold, it is considered that there is a missing in the upper right corner. The degree of missing in the upper right corner p6 is defined as: p6=(1.5*c6+c7+0.01*c0-c1-1.5*c2) / n (7) The degree of missing in the lower right corner is used to judge the slope of the curve from the top dead center to the opening point of the floating valve. When the 4th and 5th order chain codes exceed the threshold, it is considered that there is a missing in the lower right corner. The degree of missing in the lower right corner p7 is defined as: p7=(1.5*c4+0.5*c5-0.5*c6-c2-0.5*c3) / n (8) The missing degree of the lower left corner is used to judge the slope of the curve from the middle of the downstroke to the fixed valve opening point. When the 2nd and 3rd order chain codes exceed the threshold, it is considered that there is a missing in the lower left corner. The missing degree of the lower left corner p8 is defined as: p8=(1.5*c2+c3+0.01*c4-c5-1.5*c6) / n (9) The smoothness of the upstroke is used to determine the degree of change of the curve from the floating valve closing point to the fixed valve closing point. When the chain code of other orders exceeds the threshold except the 0th order chain code, the upstroke is considered to be not smooth. The smoothness of the upstroke p9 is: (10) The degree of rapid unloading at the lower right is used to determine the degree of unloading from the top dead center to the opening point of the floating valve. When the 2nd and 3rd order chain codes exceed the threshold, it is considered that the lower right is rapidly unloaded. The degree of rapid unloading at the lower right is p 10 Defined as: p 10 =(3*c3+3*c4-0.5*c2) / n (11) Downstroke smoothness is used to determine the degree of change of the curve from the floating valve opening point to the fixed valve opening point. When the chain code other than the 0th order chain code exceeds the threshold, the downstroke is considered to be not smooth. The downstroke smoothness p9 is: (12) Obesity level, used to determine the ratio of the maximum and minimum load difference of the indicator diagram to the maximum and minimum load difference of the theoretical diagram. The obesity level p3 is: Among them, L fmax The maximum load of the acquisition diagram, L fmin To collect the minimum load of the power diagram, L lmax is the maximum load of the theoretical power diagram, L lmin It is the minimum load of the theoretical power diagram.

2. The method for extracting and diagnosing the power diagram features based on Freeman chain code according to claim 1 is characterized in that: The extracting of the power diagram features from the power diagram is to extract the power diagram features through the Freeman chain code.

3. The method for extracting and diagnosing the power diagram features based on Freeman chain code according to claim 1 is characterized in that: The diagnostic evaluation of the working condition of the oil well is as follows: List all working conditions and the eigenvectors corresponding to each working condition; Construct a typical working condition-feature pattern set, and query the feature vectors that satisfy the pattern set in each working condition in turn; Based on the judgment rules satisfied by the samples and combined with the typical working condition-characteristic pattern set, the sample working conditions are diagnosed and evaluated.

Images