An optimized method for diagnosing electrical submersible pump faults based on complex data feature extraction

Abstract

The present invention discloses an optimized method for electric submersible pump fault diagnosis based on complex data feature extraction, which relates to the technical field of electric submersible pumps and includes: forming a collection scheme for at least one detection data index of the electric submersible pump and the numerical value of the detection data index; determining whether there is room for optimization in the current diagnosis scheme of the sample fault; when optimizing, forming the correlation coefficient between the detection data index and the sample fault; forming a detection data index set; forming at least one pending detection scheme for the sample fault; verifying the pending detection scheme; forming at least one overall detection scheme; calculating the detection efficiency coefficient of the overall detection scheme, and taking the overall detection scheme with the largest detection efficiency coefficient as the target detection scheme. By forming the correlation coefficient between the detection data index and the sample fault and calculating the detection efficiency coefficient of the overall detection scheme, the overall detection efficiency can be improved, and thus the effect of optimizing the electric submersible pump fault diagnosis can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric submersible pumps, and more specifically, to an optimized method for fault diagnosis of electric submersible pumps based on complex data feature extraction. Background Art

[0002] Electric submersible pumps have become one of the most widely used artificial lift methods in oil and gas fields. Therefore, studying the working state of electric submersible pumps, ensuring the long-term and efficient operation of the units, thereby extending the pump inspection period and reducing production costs have become an important topic in the production technology field of electric pump wells.

[0003] The fault detection of electric submersible pumps involves many aspects, and its detection efficiency for electric submersible pumps may not be the best. When detecting multiple faults, there may be too many detection indicators, which will increase the detection time. Summary of the Invention

[0004] To solve the above technical problems, an optimized method for fault diagnosis of electric submersible pumps based on complex data feature extraction is provided, and this technical solution solves the problems raised in the above background art.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] An optimized method for fault diagnosis of electric submersible pumps based on complex data feature extraction, comprising:

[0007] Based on historical detection data, obtain at least one sample fault of the electric submersible pump, and obtain the current diagnosis scheme of the sample fault. The current diagnosis scheme is composed of at least one current detection indicator and the fault identification range of the current detection indicator;

[0008] Form a collection scheme for at least one detection data indicator of the electric submersible pump and the value of the detection data indicator. The detection data indicators are extracted from image data, electrical data, and vibration data;

[0009] Judge whether there is room for optimization in the current diagnosis scheme of the sample fault. If not, do nothing. If so, optimize the current diagnosis scheme;

[0010] When optimizing, form the correlation coefficient between the detection data indicator and the sample fault;

[0011] Summarize the detection data indicators whose correlation coefficients with the sample fault are greater than 0 to form a detection data indicator set, and pair the detection data indicator set with the sample fault;

[0012] Based on the detection data indicator set, form at least one pending detection scheme for the sample fault;

[0013] Verify the pending detection scheme, use the pending detection scheme that passes the verification as the preliminary detection scheme, and match the preliminary detection scheme to the corresponding sample fault;

[0014] Form at least one overall detection scheme, which consists of the sample fault and the preliminary detection scheme;

[0015] Calculate the detection efficiency coefficient of the overall detection scheme, use the overall detection scheme with the largest detection efficiency coefficient as the target detection scheme, and use the target detection scheme to replace the current diagnosis scheme for all sample faults.

[0016] Preferably, the formation of at least one detection data index of the electric submersible pump and the acquisition scheme of the numerical value of the detection data index include the following steps:

[0017] Set at least one sampling point at the cable, impeller, bearing, and wearing ring of the electric submersible pump, and obtain the image data at the sampling point. The image data consists of the sample image and the actual image, and the sample image is the image when the sampling point is normal;

[0018] Take the difference between the actual image and the sample image at the sampling point as the detection data index, and the numerical value of the detection data index is the ratio of the difference between the actual image and the sample image at the sampling point;

[0019] Collect the electrical data during the operation of the electric submersible pump. The electrical data includes the operating load, operating current, and operating voltage. The operating load is collected using a load meter, the operating current is collected using an ammeter, and the operating voltage is collected using a voltmeter;

[0020] Take the operating load, operating current, and operating voltage in the electrical data as the detection data index respectively;

[0021] Collect the overall vibration intensity of the electric submersible pump. Obtain at least one basic component of the electric submersible pump in advance. The basic components consist of the cable, impeller, bearing, wearing ring, stator winding, filter, valve, shaft seal, and housing;

[0022] Obtain the vibration frequency of the basic component, and obtain the vibration intensity equal to the vibration frequency of the basic component in the overall vibration intensity to obtain the vibration intensity of the basic component;

[0023] Take the vibration intensity of the basic component as the detection data index.

[0024] Preferably, the judgment of whether there is room for optimization in the current diagnosis scheme of the sample fault includes the following steps:

[0025] Under the same conditions, obtain at least one first detection value of the detection data index as the electric submersible pump operates in the preset time interval, and obtain at least one second detection value of the current detection index as the operation time of the electric submersible pump changes;

[0026] Fit the first detection value with respect to time to obtain a first fitting function, and take the derivative of the first fitting function to obtain a first derivative function;

[0027] Fit the second detection value with respect to time to obtain a second fitting function, and take the derivative of the second fitting function to obtain a second derivative function;

[0028] Integrate the absolute value of the difference between the first derivative function and the second derivative function over a preset time interval to obtain a difference coefficient between the detection data index and the current detection index;

[0029] If the difference coefficient between the detection data index and the current detection index is less than a preset value and the detection data index is different from the current detection index, then there is room for optimization in the current diagnosis scheme for the sample fault; otherwise, there is no room for optimization in the current diagnosis scheme for the sample fault.

[0030] Preferably, the formation of the correlation coefficient between the detection data index and the sample fault includes the following steps:

[0031] Obtain at least one fault degree of the sample fault, and obtain the value of the detection data index under the condition of the fault degree of the sample fault;

[0032] Select one of at least one fault degree of the sample fault as the reference fault degree, and use the value of the detection data index corresponding to the reference fault degree as the reference value;

[0033] Use the correlation formula to calculate the correlation coefficient between the detection data index and the sample fault;

[0034] The correlation formula is as follows: ,

[0035] where A is the correlation coefficient between the detection data index and the sample fault, i is the subscript, n is the number of at least one fault degree of the sample fault, is the reference value, is the reference fault degree, is the value of the detection data index corresponding to the i-th fault degree of the sample fault, is the i-th fault degree of the sample fault.

[0036] Preferably, the formation of at least one pending detection scheme for the sample fault based on the detection data index set includes the following steps:

[0037] Summarize the detection data indexes included in the subset of the detection data index set corresponding to the sample fault as the pending detection scheme;

[0038] All subsets of the detection data index set corresponding to the sample fault form at least one pending detection scheme.

[0039] Preferably, the verification of the pending detection scheme comprises the following steps:

[0040] Use the pending detection scheme to detect the corresponding sample faults, and calculate the judgment accuracy of the pending detection scheme;

[0041] When the difference between the judgment accuracy and 1 is less than the preset difference, the pending detection scheme passes the verification; otherwise, the pending detection scheme fails the verification.

[0042] Preferably, forming at least one overall detection scheme comprises the following steps:

[0043] taking one of at least one preliminary detection scheme for the sample fault as a candidate detection scheme;

[0044] Summarize the candidate detection schemes for all sample faults to form an overall detection scheme;

[0045] When the candidate detection scheme traverses at least one preliminary detection scheme of the sample fault, at least one overall detection scheme is formed.

[0046] Preferably, the calculation of the detection efficiency coefficient of the overall detection scheme comprises the following steps:

[0047] Acquire at least one sample fault set, where the at least one sample fault set constitutes all combinations of sample faults;

[0048] Get the detection time of the detection data indicator;

[0049] Aggregate the preliminary detection schemes corresponding to the sample faults in the sample fault set into a preliminary detection scheme set;

[0050] Aggregating candidate detection schemes of the overall detection scheme into a candidate detection scheme set;

[0051] Take the intersection of the candidate detection scheme set and the preliminary detection scheme set to obtain a feature set;

[0052] Summarize and remove duplicate detection data indicators of the preliminary detection scheme in the feature set to obtain a feature detection scheme, which is a scheme for detecting all sample faults in the sample fault set;

[0053] The detection data indicators in the feature detection scheme are classified into image indicators, electrical indicators and vibration indicators according to the attributes of the indicators. The attributes of the indicators are image, electrical and vibration.

[0054] The detection time of the image index is superimposed to obtain the first time, the detection time of the electrical index is superimposed to obtain the second time, and the detection time of the vibration index is superimposed to obtain the third time;

[0055] Take the maximum value among the first time, the second time, and the third time as the total detection time of the sample fault set;

[0056] Add up the total detection times of at least one sample fault set and take the reciprocal to obtain the detection efficiency coefficient of the overall detection scheme.

[0057] Preferably, the steps for obtaining the detection time of the detection data index include the following:

[0058] When the attribute of the detection data index is an image, then count the time for obtaining the value of the detection data index through image analysis as the detection time of the detection data index;

[0059] When the attribute of the detection data index is electrical, then count the time for the instrument to measure the value of the detection data index as the detection time of the detection data index;

[0060] When the attribute of the detection data index is vibration, then count the time for obtaining the value of the detection data index through vibration frequency analysis as the detection time of the detection data index.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] By judging whether there is room for optimization in the current diagnosis scheme of the sample fault, forming the correlation coefficient between the detection data index and the sample fault, and calculating the detection efficiency coefficient of the overall detection scheme, it is possible to evaluate the current diagnosis scheme of the sample fault of the existing electric submersible pump. Thus, it is possible to determine whether the current diagnosis scheme needs to be optimized according to the evaluation result. And when optimizing, select the overall detection scheme with the largest detection efficiency coefficient as the target detection scheme. Then, when detecting multiple sample faults, it can ensure that the repetition degree of the detection data index adopted by the sample fault is relatively large and the overall detection time consumed is relatively small. Thus, it can improve the overall detection efficiency and further achieve the effect of optimizing the fault diagnosis of the electric submersible pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flow chart of the electric submersible pump fault diagnosis optimization method based on complex data feature extraction of the present invention;

[0064] Figure 2 It is a schematic flow chart of the acquisition scheme for forming at least one detection data index and the value of the detection data index of the electric submersible pump of the present invention;

[0065] Figure 3 It is a schematic flow chart of judging whether there is room for optimization in the current diagnosis scheme of the sample fault of the present invention;

[0066] Figure 4Schematic flow diagram of forming the correlation coefficient between the detection data index and the sample fault of the present invention;

[0067] Figure 5 Schematic flow diagram of forming at least one pending detection scheme for sample faults based on the detection data index set of the present invention;

[0068] Figure 6 Schematic flow diagram of verifying the pending detection scheme of the present invention;

[0069] Figure 7 Schematic flow diagram of forming at least one overall detection scheme of the present invention;

[0070] Figure 8 Schematic flow diagram of calculating the detection efficiency coefficient of the overall detection scheme of the present invention;

[0071] Figure 9 Schematic flow diagram of obtaining the detection time of the detection data index of the present invention. Detailed implementation manners

[0072] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0073] Refer to Figure 1 As shown, an optimization method for electric submersible pump fault diagnosis based on complex data feature extraction includes:

[0074] Based on historical detection data, obtain at least one sample fault of the electric submersible pump, obtain the current diagnosis scheme of the sample fault, and the current diagnosis scheme is composed of at least one current detection index and the fault identification range of the current detection index;

[0075] Form a collection scheme for at least one detection data index of the electric submersible pump and the value of the detection data index, and the detection data index is extracted from image data, electrical data, and vibration data;

[0076] Judge whether there is room for optimization in the current diagnosis scheme of the sample fault. If not, do nothing. If so, optimize the current diagnosis scheme;

[0077] When optimizing, form the correlation coefficient between the detection data index and the sample fault;

[0078] Summarize the detection data indexes with a correlation coefficient greater than 0 with the sample fault, form a detection data index set, and pair the detection data index set with the sample fault;

[0079] Based on the detection data index set, form at least one pending detection scheme for the sample fault;

[0080] Verify the to-be-detected scheme, use the to-be-detected scheme that passes the verification as the preliminary detection scheme, and match the preliminary detection scheme to the corresponding sample fault;

[0081] Form at least one overall detection scheme, where the overall detection scheme consists of sample faults and preliminary detection schemes;

[0082] Calculate the detection efficiency coefficient of the overall detection scheme, use the overall detection scheme with the largest detection efficiency coefficient as the target detection scheme, and use the target detection scheme to replace the current diagnosis scheme for all sample faults.

[0083] When performing detection, it is not necessarily for one sample fault, but may be for multiple sample faults. When the coincidence degree of the detection data indicators of multiple sample faults is small, the number of detection data indicators to be detected is relatively large. At the same time, the detection time of the detection data indicators used may not be optimal. Therefore, the overall detection time will be prolonged. In this solution, for this situation, the diagnostic detection is optimized to ensure that the coincidence degree of the detection data indicators of multiple sample faults is large, and at the same time, the total time of the detection time of the detection data indicators used is minimized, thereby improving the detection efficiency.

[0084] Refer to Figure 2 As shown, the formation of at least one detection data indicator of the electric submersible pump and the acquisition scheme of the values of the detection data indicators includes the following steps:

[0085] Set at least one sampling point at the cable, impeller, bearing, and wearing ring of the electric submersible pump, and obtain the image data at the sampling point. The image data consists of a sample image and an actual image, and the sample image is the image when the sampling point is normal;

[0086] Take the difference between the actual image and the sample image at the sampling point as the detection data indicator, and the value of the detection data indicator is the ratio of the difference between the actual image and the sample image at the sampling point;

[0087] Collect the electrical data when the electric submersible pump is running. The electrical data includes running load, running current, and running voltage. The running load is collected using a load meter, the running current is collected using an ammeter, and the running voltage is collected using a voltmeter;

[0088] Take the running load, running current, and running voltage in the electrical data as the detection data indicators respectively;

[0089] Collect the overall vibration intensity of the electric submersible pump. Obtain at least one basic component of the electric submersible pump in advance. The basic components consist of a cable, impeller, bearing, wearing ring, stator winding, filter, valve, shaft seal, and housing;

[0090] Obtain the vibration frequency of the basic component, obtain the vibration intensity equal to the vibration frequency of the basic component in the overall vibration intensity, and obtain the vibration intensity of the basic component;

[0091] Take the vibration intensity of the basic component as the detection data index.

[0092] Refer to Figure 3 As shown, determining whether there is room for optimization in the current diagnosis plan for the sample fault includes the following steps:

[0093] Under the same conditions, obtain at least one first detection value of the detection data index as the electric submersible pump operates in the preset time interval, and obtain at least one second detection value of the current detection index as the operation time of the electric submersible pump changes;

[0094] Fit the first detection value with respect to time to obtain a first fitting function, and take the derivative of the first fitting function to obtain a first derivative function;

[0095] Fit the second detection value with respect to time to obtain a second fitting function, and take the derivative of the second fitting function to obtain a second derivative function;

[0096] Integrate the absolute value of the difference between the first derivative function and the second derivative function over the preset time interval to obtain the difference coefficient between the detection data index and the current detection index;

[0097] If the difference coefficient between the detection data index and the current detection index is less than the preset value and the detection data index is different from the current detection index, then there is room for optimization in the current diagnosis plan for the sample fault; otherwise, there is no room for optimization in the current diagnosis plan for the sample fault.

[0098] The basis is that when the change situation of the detection data index as the electric submersible pump operates is very similar to the change situation of the current detection index, it indicates that they are similar indicators, that is, the detection effects of the two are the same, so they can be interchanged. Then, when the detection data index is different from the current detection index, the effect after interchange may be better, so there is room for optimization.

[0099] Refer to Figure 4 As shown, forming the correlation coefficient between the detection data index and the sample fault includes the following steps:

[0100] Obtain at least one fault degree of the sample fault, and obtain the value of the detection data index under the condition of the fault degree of the sample fault;

[0101] Select one of at least one fault degree of the sample fault as the reference fault degree, and take the value of the detection data index corresponding to the reference fault degree as the reference value;

[0102] Use the correlation formula to calculate the correlation coefficient between the detection data index and the sample fault;

[0103] The correlation formula is as follows: ,

[0104] where A is the correlation coefficient between the detection data index and the sample fault, i is the subscript, n is the number of at least one fault degree of the sample fault, is the reference value, is the reference fault degree, is the value of the detection data index corresponding to the i-th fault degree of the sample fault, is the i-th fault degree of the sample fault.

[0105] The calculation of the correlation coefficient depends on the change of the value of the detection data index caused by the change of the sample fault. When there is no correlation, the change of the sample fault hardly affects the value of the detection data index, and the value of the detection data index will not change. When the correlation coefficient between the detection data index and the sample fault is not 0, it indicates that it may be used for the identification and judgment of the sample fault, but a series of subsequent steps are required for verification.

[0106] Referring to Figure 5 shown, based on the detection data index set, forming at least one pending detection scheme for the sample fault includes the following steps:

[0107] Summarize the detection data indexes included in the subset of the detection data index set corresponding to the sample fault as the pending detection scheme;

[0108] All subsets of the detection data index set corresponding to the sample fault form at least one pending detection scheme.

[0109] The detection of the sample fault depends on the detection data indexes in its corresponding detection data index set. Any subset of the detection data index set can be used as a pending detection scheme, but the pending detection scheme is not a scheme that can accurately identify the sample fault, that is, the detection value of the detection data index depending on the pending detection scheme is not sufficient to determine the existence or non-existence of the sample fault. Therefore, it is necessary to verify the pending detection scheme, and the verification depends on the correct rate of the identification of the pending detection scheme. When using the detection data index, the normal range of the detection data index can be obtained based on historical experience. When the value of the detection data index is not within the normal range, the detection data index is abnormal. The method for the pending detection scheme to identify the sample fault is: when all the detection data indexes in the pending detection scheme are abnormal, it is determined that the sample fault exists according to the pending detection scheme.

[0110] Referring to Figure 6 shown, verifying the pending detection scheme includes the following steps:

[0111] Use the pending detection scheme to detect the corresponding sample faults, and calculate the judgment accuracy of the pending detection scheme;

[0112] When the difference between the judgment accuracy and 1 is less than the preset difference, the pending detection scheme passes the verification; otherwise, the pending detection scheme fails the verification.

[0113] Reference Figure 7 As shown, forming at least one overall detection scheme includes the following steps:

[0114] taking one of at least one preliminary detection scheme for the sample fault as a candidate detection scheme;

[0115] Summarize the candidate detection schemes for all sample faults to form an overall detection scheme;

[0116] When the candidate detection scheme traverses at least one preliminary detection scheme of the sample fault, at least one overall detection scheme is formed.

[0117] Since a sample fault has at least one preliminary detection scheme to detect it, and the overall detection scheme is a scheme for detecting all sample faults, it is necessary to determine the preliminary detection scheme adopted for the sample fault in the overall detection scheme. Therefore, each selection of at least one preliminary detection scheme for the sample fault will form an overall detection scheme.

[0118] Reference Figure 8 As shown, calculating the detection efficiency coefficient of the overall detection scheme includes the following steps:

[0119] Obtain at least one sample fault set, where the at least one sample fault set constitutes all combinations of sample faults;

[0120] Get the detection time of the detection data indicator;

[0121] Aggregate the preliminary detection schemes corresponding to the sample faults in the sample fault set into a preliminary detection scheme set;

[0122] Aggregating candidate detection schemes of the overall detection scheme into a candidate detection scheme set;

[0123] Take the intersection of the candidate detection scheme set and the preliminary detection scheme set to obtain a feature set;

[0124] Summarize and remove duplicate detection data indicators of the preliminary detection scheme in the feature set to obtain a feature detection scheme, which is a scheme for detecting all sample faults in the sample fault set;

[0125] The detection data indicators in the feature detection scheme are classified into image indicators, electrical indicators and vibration indicators according to the attributes of the indicators. The attributes of the indicators are image, electrical and vibration.

[0126] Overlay the detection times of the image metrics to obtain a first time, overlay the detection times of the electrical metrics to obtain a second time, and overlay the detection times of the vibration metrics to obtain a third time;

[0127] Take the maximum value among the first time, the second time, and the third time as the total detection time of the sample fault set;

[0128] Overlay the total detection times of at least one sample fault set and then take the reciprocal to obtain the detection efficiency coefficient of the overall detection scheme.

[0129] Because in actual detection, it is necessary to detect all possible combinations of sample faults. Therefore, the calculation of the detection efficiency coefficient depends on all possible combinations of sample faults, that is, the sample fault set. In the overall detection scheme, candidate detection schemes are determined for all sample faults, and at least one preliminary detection scheme corresponding to the sample fault is also known. Therefore, in order to determine all the preliminary detection schemes adopted by the sample fault set, it is necessary to take the intersection of the candidate detection scheme set and the preliminary detection scheme set to obtain a feature set. Then, the preliminary detection schemes in the feature set are the schemes adopted by the sample faults in the sample fault set. It should be noted that here the candidate detection scheme is also a preliminary detection scheme, so the intersection can be taken. However, there will be overlapping situations in the detection data metrics of the preliminary detection schemes in the feature set. Therefore, it is necessary to de-duplicate and summarize them, and calculate the time-consuming for identifying these metrics. However, due to the different attributes of these metrics, the identification of metrics with different attributes is parallel because different modules are used for identification. Therefore, the first time, the second time, and the third time are classified, and the total detection time of the sample fault set is obtained. Since it is necessary to detect all sample fault sets, the total detection times of at least one sample fault set are overlaid and then the reciprocal is taken to obtain the detection efficiency coefficient of the overall detection scheme. The reason for taking the reciprocal is that the longer the time, the lower the detection efficiency.

[0130] Refer to Figure 9 As shown, obtaining the detection time of the detection data metrics includes the following steps:

[0131] When the attribute of the detection data metric is an image, then count the time for obtaining the value of the detection data metric through image analysis as the detection time of the detection data metric;

[0132] When the attribute of the detection data metric is electrical, then count the time for the instrument to measure the value of the detection data metric as the detection time of the detection data metric;

[0133] When the attribute of the detection data metric is vibration, then count the time for obtaining the value of the detection data metric through vibration frequency analysis as the detection time of the detection data metric.

[0134] Furthermore, the present solution also proposes a storage medium, on which a computer-readable program is stored. When the computer-readable program is called, it executes the above-mentioned optimized method for diagnosing ESP faults based on complex data feature extraction.

[0135] It can be understood that the storage medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid-state disk (SSD).

[0136] In summary, the advantages of the present invention are as follows: By determining whether there is room for optimization in the current diagnosis solution for sample faults, forming the correlation coefficient between the detection data index and the sample faults, and calculating the detection efficiency coefficient of the overall detection solution, the current diagnosis solution for the sample faults of the existing ESP can be evaluated. Thus, it can be determined whether the current diagnosis solution needs to be optimized according to the evaluation results. And when optimizing, the overall detection solution with the largest detection efficiency coefficient is selected as the target detection solution. Then, when detecting multiple sample faults, it can ensure that the repetition degree of the detection data indexes used for the sample faults is relatively large and the overall detection time consumed is relatively small. Thus, the overall detection efficiency can be improved, and further the effect of optimizing the diagnosis of ESP faults can be achieved.

[0137] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, various changes and improvements will occur to the present invention, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An electric submersible pump fault diagnosis optimization method based on complex data feature extraction, characterized in that: include: Based on the historical detection data, at least one sample fault of the electric submersible pump is obtained, and a current diagnosis scheme of the sample fault is obtained, wherein the current diagnosis scheme is composed of at least one current detection index and a fault identification range of the current detection index; forming a collection scheme for at least one detection data indicator of the electric submersible pump and a value of the detection data indicator, the detection data indicator being extracted from the image data, the electrical data, and the vibration data; Determine whether there is room for optimization in the current diagnostic scheme of the sample fault. If not, do nothing. If yes, optimize the current diagnostic scheme. During optimization, the correlation coefficient between the detection data index and the sample fault is formed; Aggregate the detection data indicators whose correlation coefficients with the sample faults are greater than 0 to form a detection data indicator set, and pair the detection data indicator set with the sample faults; Based on the set of detection data indicators, forming at least one pending detection scheme for the sample fault; Verify the pending detection plan, use the pending detection plan that has passed the verification as the preliminary detection plan, and match the preliminary detection plan to the corresponding sample fault; Forming at least one overall detection plan, the overall detection plan consisting of sample faults and preliminary detection plans; The detection efficiency coefficient of the overall detection scheme is calculated, and the overall detection scheme with the largest detection efficiency coefficient is used as the target detection scheme. The target detection scheme is used to replace the current diagnosis scheme of all sample faults.

2. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 1 is characterized in that: The scheme for collecting at least one detection data indicator and the value of the detection data indicator of the electric submersible pump comprises the following steps: At least one sampling point is set at the cable, impeller, bearing, and mouth ring of the electric submersible pump to obtain image data at the sampling point, wherein the image data is composed of a sample image and an actual image, and the sample image is an image when the sampling point is normal; The difference between the actual image and the sample image at the sampling point is used as a detection data index, and the value of the detection data index is the ratio of the difference between the actual image and the sample image at the sampling point; Collect electrical data when the electric submersible pump is running. The electrical data includes operating load, operating current and operating voltage. The operating load is collected using a load meter, the operating current is collected using an ammeter, and the operating voltage is collected using a voltmeter; The operating load, operating current and operating voltage in the electrical data are respectively used as detection data indicators; Collect the overall vibration intensity of the electric submersible pump, and obtain in advance at least one basic component of the electric submersible pump, the basic component consisting of a cable, an impeller, a bearing, a mouth ring, a stator winding, a filter, a valve, a shaft seal and a casing; Obtaining the vibration frequency of the basic component, obtaining the vibration intensity equal to the vibration frequency of the basic component in the overall vibration intensity, and obtaining the vibration intensity of the basic component; The vibration intensity of basic components is used as the detection data indicator.

3. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 2 is characterized in that: The method of determining whether there is room for optimization of the current diagnostic scheme for the sample fault comprises the following steps: Under the same conditions, obtaining at least one first detection value of the detection data index as the electric submersible pump runs in a preset time interval, and obtaining at least one second detection value of the current detection index as the electric submersible pump runs for a certain period of time; Fitting the first detection value with respect to time to obtain a first fitting function, and taking the derivative of the first fitting function to obtain a first derivative function; Fitting the second detection value with respect to time to obtain a second fitting function, and taking the derivative of the second fitting function to obtain a second derivative function; Integrate the absolute value of the difference between the first derivative function and the second derivative function over a preset time interval to obtain a difference coefficient between the detection data index and the current detection index; When the difference coefficient between the detection data indicator and the current detection indicator is less than the preset value, and the detection data indicator is different from the current detection indicator, there is room for optimization of the current diagnostic scheme for the sample fault; otherwise, there is no room for optimization of the current diagnostic scheme for the sample fault.

4. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 3 is characterized in that: The forming of the correlation coefficient between the detection data index and the sample fault comprises the following steps: Obtaining at least one fault degree of the sample fault, and obtaining a value of the detection data indicator under the condition of the fault degree of the sample fault; Selecting at least one of the fault levels of the sample fault as a reference fault level, and using the value of the detection data indicator corresponding to the reference fault level as a reference value; Use the correlation formula to calculate the correlation coefficient between the detection data index and the sample fault; The associated formula is as follows: , Where A is the correlation coefficient between the detection data index and the sample fault, i is the subscript, and n is the number of at least one fault degree of the sample fault. is the benchmark value, is the baseline fault level, is the value of the detection data indicator corresponding to the i-th fault degree of the sample fault, is the i-th fault degree of the sample fault.

5. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 4 is characterized in that: The forming of at least one pending detection scheme for the sample fault based on the detection data indicator set comprises the following steps: Summarize the detection data indicators contained in the subset of the detection data indicator set corresponding to the sample fault as a pending detection plan; All subsets of the detection data indicator set corresponding to the sample fault form at least one pending detection scheme.

6. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 5 is characterized in that: The verification of the pending detection scheme comprises the following steps: Use the pending detection scheme to detect the corresponding sample faults, and calculate the judgment accuracy of the pending detection scheme; When the difference between the judgment accuracy and 1 is less than the preset difference, the pending detection scheme passes the verification; otherwise, the pending detection scheme fails the verification.

7. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 6 is characterized in that: The forming of at least one overall detection scheme comprises the following steps: taking one of at least one preliminary detection scheme for the sample fault as a candidate detection scheme; Summarize the candidate detection schemes for all sample faults to form an overall detection scheme; When the candidate detection scheme traverses at least one preliminary detection scheme of the sample fault, at least one overall detection scheme is formed.

8. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 7 is characterized in that: The calculation of the detection efficiency coefficient of the overall detection scheme comprises the following steps: Acquire at least one sample fault set, where the at least one sample fault set constitutes all combinations of sample faults; Get the detection time of the detection data indicator; Aggregate the preliminary detection schemes corresponding to the sample faults in the sample fault set into a preliminary detection scheme set; Aggregating candidate detection schemes of the overall detection scheme into a candidate detection scheme set; Take the intersection of the candidate detection scheme set and the preliminary detection scheme set to obtain a feature set; Summarize and remove duplicate detection data indicators of the preliminary detection scheme in the feature set to obtain a feature detection scheme, which is a scheme for detecting all sample faults in the sample fault set; The detection data indicators in the feature detection scheme are classified into image indicators, electrical indicators and vibration indicators according to the attributes of the indicators. The attributes of the indicators are image, electrical and vibration. The detection time of the image index is superimposed to obtain the first time, the detection time of the electrical index is superimposed to obtain the second time, and the detection time of the vibration index is superimposed to obtain the third time; Taking the maximum value among the first time, the second time and the third time as the total detection time of the sample fault set; The total detection time of at least one sample fault set is superimposed and the reciprocal is taken to obtain the detection efficiency coefficient of the overall detection scheme.

9. The electric submersible pump fault diagnosis optimization method based on complex data feature extraction according to claim 8 is characterized in that: The method of obtaining the detection time of the detection data indicator includes the following steps: When the attribute of the detection data indicator is an image, the time for obtaining the value of the detection data indicator through image analysis is counted as the detection time of the detection data indicator; When the attribute of the detection data indicator is electrical, the time when the statistical instrument measures the value of the detection data indicator is used as the detection time of the detection data indicator; When the attribute of the detection data indicator is vibration, the time for obtaining the value of the detection data indicator through vibration frequency analysis is counted as the detection time of the detection data indicator.

Images