An ultra-deep production string leak detection system and method

Abstract

The present application relates to the technical field of oil and gas exploitation, and particularly relates to a kind of ultra-deep production string leakage detection system and method.The technical scheme is: through production string pressure detection unit and annulus pressure detection unit, real-time pressure data is obtained, and through first electric throttle control valve and second electric throttle control valve, the pressure of string side and annulus side is adjusted, so that wellbore forms controllable differential pressure condition under the premise of meeting safety conditions;The downhole detection sub is connected by armored cable and lowered into the production string, moves in the direction of well depth for continuous detection or fixed-point detection, and the multi-band signals detected are gathered through data transmission sub and then uploaded to the ground control and analysis system to complete the processing of detection data analysis and leakage result output.The present application improves the accuracy, stability and engineering applicability of leakage type, leakage location and leakage intensity discrimination, and provides technical support for ultra-deep gas well production string leakage diagnosis, workover decision and wellbore integrity management.

Description

Technical Field

[0001] This invention relates to the field of oil and gas extraction technology, and in particular to a leak detection system and method for ultra-deep production tubing. Background Technology

[0002] Existing methods for detecting leaks in oil and gas well production tubing often rely on indirect assessments based on changes in wellhead pressure, flow rate, and production rates, or employ single physical quantity detection methods to identify downhole anomalies. While these methods have some application value in conventional wells or under relatively stable conditions, in ultra-deep gas well scenarios, due to long downhole transmission paths, multiple sources of disturbance, and significant formation-wellbore coupling, changes in wellhead parameters often exhibit lag, weak sensitivity, and multiple solutions, making it difficult to accurately reflect the location, type, and intensity of leaks. Furthermore, traditional single physical quantity detection methods have limited anti-interference capabilities, and are prone to problems such as missed detections, false detections, and unstable judgment results under conditions of minor leaks, complex noise backgrounds, and multi-point leaks.

[0003] With the development of downhole monitoring and data processing technologies, intelligent identification methods based on acoustic signals, pressure signals, and multi-source sensor information are gradually becoming an important direction for leak diagnosis. However, most existing intelligent identification methods are still based on single operating conditions, single sampling, or static sample classification. They typically involve directly inputting the collected acoustic signals or multi-source signals into convolutional neural networks, recurrent neural networks, or attention networks for training and classification. While these methods can achieve a certain level of recognition accuracy on specific datasets, they generally suffer from the following shortcomings: First, they lack effective characterization mechanisms for background noise, downhole environmental disturbances, and operating condition drift. The models often learn the absolute amplitude characteristics of the signal rather than the actual changes induced by leakage. Second, they fail to fully utilize the dynamic evolution characteristics of the leakage response, which gradually increases with the change in oil-casing pressure differential, making it difficult to establish a discrimination criterion consistent with the leakage mechanism. Third, they typically use only simple splicing or static fusion methods between multi-source signals, failing to fully reflect the coupling response relationship between different frequency bands and different physical quantities under complex downhole operating conditions. Fourth, after the model output, there is a lack of result calibration, similarity verification, and retesting confirmation mechanisms for engineering applications, resulting in insufficient interpretability, stability, and repeatability of the recognition results.

[0004] Furthermore, existing technologies for leak detection typically focus on "passively receiving and classifying signals," lacking an active identification approach that integrates with controlled downhole conditions. For leaks in ultra-deep production tubing, the leak response is not constant but exhibits varying degrees of acoustic, pressure, and temperature disturbances depending on the pressure differential between the production tubing and the annulus. Judging based solely on instantaneous samples under a single condition is insufficient to effectively distinguish leak signals from background noise and other non-leakage disturbances, and it also hinders the improvement of identification reliability for minor leaks and complex conditions. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned deficiencies in existing technologies by providing an ultra-deep production tubing leakage detection system and method. This system not only enables the collaborative acquisition and fusion analysis of multi-source information such as ultrasonic signals, low-frequency acoustic signals, pressure signals, and temperature signals, but also constructs a leakage response enhancement mechanism by combining background operating conditions and graded differential pressure conditions. Furthermore, it utilizes the evolution law of leakage response with differential pressure to calibrate and confirm the identification results, thereby improving the accuracy, stability, and engineering applicability of leakage type, leakage location, and leakage intensity determination.

[0006] The present invention discloses an ultra-deep production tubing leakage detection system, the technical solution of which includes a production tubing and a surface control and analysis system. The system further includes an annular pressure detection unit, a production tubing pressure detection unit, a first electric throttle control valve, a second electric throttle control valve, and a downhole detection sub. The annular pressure detection unit and the production tubing pressure detection unit are installed on the wellhead device and connected to the second and first electric throttle control valves, respectively. Pressure data is acquired in real time through the production tubing pressure detection unit and the annular pressure detection unit. The pressure on the tubing side and the annular side is adjusted through the first and second electric throttle control valves to create a controllable differential pressure condition in the wellbore while meeting safety requirements. The downhole detection sub is connected to and lowered into the production tubing via an armored cable, moving along the well depth direction for continuous or fixed-point detection. The detected multi-frequency signals are aggregated by the data transmission sub and uploaded to the surface control and analysis system for processing, analysis, and leakage result output.

[0007] Preferably, the aforementioned downhole detection sub includes an armored cable, an upper centralizing connector, a data transmission sub, an ultrasonic sensor, an ultrasonic detection sub, a temperature sensor, a temperature detection sub, a pressure sensor, a pressure detection sub, a low-frequency acoustic sensor, a low-frequency acoustic detection sub, and a lower centralizing connector. The upper centralizing connector is connected to the armored cable at its upper end and to the data transmission sub at its lower end. Below the data transmission sub, the ultrasonic detection sub, temperature detection sub, pressure detection sub, and low-frequency acoustic detection sub are connected. An ultrasonic sensor is installed within the ultrasonic detection sub for collecting data on leaks in the production tubing. The high-frequency acoustic signal caused by the leak is collected by a temperature sensor installed inside the temperature detection sub, which is used to acquire the downhole ambient temperature and the fluid temperature inside the production tubing to reflect the local temperature disturbance characteristics caused by the leak, and to perform condition compensation and drift correction on the acoustic signal; the pressure detection sub is installed inside the pressure sensor to acquire the fluid pressure change signal and pressure pulsation signal inside the production tubing; the low-frequency acoustic wave detection sub is installed inside the low-frequency acoustic wave sensor to acquire the low-frequency acoustic signal and fluid disturbance noise signal caused by the leak, and the lower part of the low-frequency acoustic wave detection sub is connected to the lower centering joint.

[0008] Preferably, the above-mentioned upper straightening connector includes an upper guide head and a straightening body. The upper end of the upper guide head is connected to the armored cable, and the lower end of the straightening body is connected to the upper end of the data transmission sub. The outer diameter of the straightening body is larger than the outer diameter of the data transmission sub, the ultrasonic detection sub, the temperature detection sub, the pressure detection sub, and the low-frequency acoustic wave detection sub.

[0009] Preferably, the annular pressure detection unit and the production tubing pressure detection unit mentioned above use pressure gauges.

[0010] Preferably, the above-mentioned ground control and analysis system includes a cable working vehicle, a cable winch, a PLC control system, a PC, and a data acquisition card. The cable working vehicle drives the armored cable to move up and down through the cable winch, causing the downhole inspection section to move up and down along the inner cavity of the production tubing for inspection. The downhole inspection section uploads the inspection data to the PC through the data acquisition card, and then the PLC control system controls the operation of the first electric throttle control valve, the second electric throttle control valve, and the cable winch.

[0011] Preferably, the annular pressure detection unit and the production tubing pressure detection unit monitor the pressure of the annulus and the production tubing, respectively, and transmit the detection data to the PC for analysis and judgment. The PC generates control commands based on the target pressure difference and sends them to the PLC control system. The PLC control system controls the opening of the first electric throttle control valve and the second electric throttle control valve to establish and maintain the preset pressure difference between the production tubing and the annulus.

[0012] The detection method of the ultra-deep production tubing leakage detection system mentioned in this invention includes the following process:

[0013] 1. Under the condition that no leakage pressure differential is established or leakage flow is induced, the ultrasonic signal, low-frequency sound wave signal, pressure signal and temperature signal in the production string are collected by the downhole detection sub to obtain the background signal set, which is used for subsequent comparison and analysis with signals under different pressure differential conditions.

[0014] 2. Based on the safety constraints and operation strategy of the production tubing, the opening of the first electric throttle control valve and the second electric throttle control valve are adjusted to establish and stabilize the corresponding target pressure difference between the production tubing and the annulus. After the pressure difference level is stabilized, the armored cable is moved up and down by the cable winch to drive the downhole detection sub to collect multi-source operating condition signals and send them to the PC to obtain the set of operating condition signals under the corresponding pressure difference level.

[0015] Third, by comparing and analyzing the background signal set and the operating condition signal set using a PC, a background-operating condition difference enhancement feature reflecting the incremental response induced by leakage is constructed to improve the contrast between the leakage signal and the background noise.

[0016] The difference enhancement feature is represented as follows:

[0017] (1),

[0018] in, B (n) Indicates the first n One background sample, S i (n) Indicates the first i Under the first pressure differential level n One working condition sample, F (.) is a statistical function of energy, root mean square, spectral energy, pressure fluctuation amplitude, temperature disturbance amplitude, or a combination thereof; through the above comparative analysis, the regularity of leakage response with pressure difference is enhanced, and a basis is provided for subsequent cross-pressure difference evolution analysis, trend calibration and result confirmation;

[0019] IV. Synchronization or quasi-synchronization alignment, filtering and noise reduction, smoothing compensation, normalization and multi-source sample construction of multi-source signals;

[0020] Among them, the synchronization or quasi-synchronization alignment, filtering and denoising, smoothing compensation and normalization of multi-source signals are performed by using the synchronization trigger signal and / or time stamp information provided by the data transmission section to align each channel; after alignment, the ultrasonic signal and the low-frequency sound wave signal are processed by bandpass filtering, bandstop filtering and / or adaptive denoising, the pressure signal and the temperature signal are processed by smoothing, and each channel is normalized.

[0021] Multi-source sample construction refers to slicing the signal sequences of each channel using a sliding window method to construct multi-source samples. Let the window length be L and the step size be S. For any channel signal sequence, multiple time window segments are extracted. After aligning each channel within the same time window, corresponding multi-source sample pairs are formed. Sample labels include at least a leak type label, a leak location interval label, and a leak intensity level label. The leak type label includes at least one or more of the following: coupling thread seal failure, tubing corrosion perforation or cracking, and packer failure. The leak location interval label is segmented according to well depth or tubing structure segments. The leak intensity level label is graded according to acoustic energy indicators and / or differential pressure response characteristics.

[0022] V. Construct cross-pressure difference evolution characteristics based on the background-operating condition difference enhancement characteristics under different pressure difference levels; then organize the difference response characteristics of each level according to the pressure difference level order to form cross-pressure difference evolution characteristics used to characterize the leakage response as a function of pressure difference, expressed as:

[0023] (2),

[0024] Among them, G (n) Indicates the first n Transpressure gradient evolution characteristics corresponding to each sample M Given the target number of differential pressure levels, Concat(.) indicates that the levels are assembled in a predetermined order.

[0025] 6. Based on the background-condition difference enhancement features, the cross-pressure difference evolution features and / or the multi-source samples, perform fusion identification to output the leakage type, leakage location range and / or leakage intensity level;

[0026] The fusion identification is implemented using a one-dimensional multi-scale convolutional network, attention network, Transformer module or a combination thereof, to extract and fusion discriminate features from multi-source signals, differential response features and / or cross-pressure difference evolution features. Through the above fusion identification, the leakage type, leakage location range and / or leakage intensity level are output, and the corresponding confidence or probability value is output at the same time.

[0027] Preferably, to enhance the interpretability of the output results and reduce the false alarm rate under complex operating conditions, a trend calibration rule based on differential pressure levels is introduced: the obtained background signal set is compared with the operating condition signal set obtained at different differential pressure levels to obtain the difference response characteristics at different differential pressure levels, and based on this, it is determined whether the leakage-related response shows a preset trend as the differential pressure increases; further, the trends of ultrasonic signal energy, low-frequency sound wave energy, pressure pulsation amplitude, and temperature disturbance amplitude are evaluated; the trend consistency can be expressed as:

[0028] (3),

[0029] in, Indicates the first n The trend consistency score of each sample, where 1(.) is the indicator function. τ To allow for fluctuation thresholds; when the trend consistency score reaches a preset threshold, or when the indicator shows monotonic enhancement, stage enhancement and / or reaches a preset threshold relative to the background signal at at least two differential pressure levels, it is determined that it is consistent with the leakage mechanism, thereby improving the confidence of the identification result and serving as one of the criteria for result confirmation; conversely, when the trend does not meet the preset rules, the confidence of the corresponding identification result is reduced, or the result is marked as pending review or retesting.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] This invention uses a data acquisition card to collect ultrasonic signals, low-frequency acoustic signals, pressure signals, and temperature signals output from a downhole detection sub. A PC then performs signal preprocessing, differential response construction, multi-source sample construction, fusion identification, trend calibration, similarity verification, and closed-loop confirmation of retesting, thereby enabling the identification of leakage type, leakage location range, and leakage intensity level.

[0032] This invention does not merely perform static classification of raw signals acquired under a single operating condition. Instead, it constructs a leakage response enhancement mechanism by combining background operating condition signals and graded differential pressure operating condition signals. The difference in response between the background signal and operating condition signals at different differential pressure levels characterizes the leakage-induced change law. Based on the evolution relationship of the difference in response at different differential pressure levels, it constructs cross-differential pressure evolution characteristics. Furthermore, it combines the cross-differential pressure evolution characteristics and their changing trends to calibrate and confirm the identification results. This improves the accuracy, stability, and engineering applicability of identifying micro-leakage, multi-point leakage, and weak-response leakage in complex downhole environments, providing technical support for leakage diagnosis of production tubing in ultra-deep gas wells, well workover decisions, and wellbore integrity management. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall construction process of the present invention;

[0034] Figure 2 This is a schematic diagram of the downhole inspection sub.

[0035] Figure 3 This is a schematic diagram of the detection method of the present invention;

[0036] In the diagram: Packer 101, Annular medium 102, Casing 103, Production tubing 104, Wellhead device 105, Annular pressure detection unit 106, Production tubing pressure detection unit 107, First electric throttle control valve 108, Second electric throttle control valve 109, Downhole inspection sub 200, Armored cable 201, Upper straightening connector 202, Data transmission sub 203, Ultrasonic sensor 204, Ultrasonic detection sub 205, Temperature sensor 206, Temperature detection sub 207, Pressure sensor 208, Pressure detection sub 209, Low-frequency acoustic wave sensor 210, Low-frequency acoustic wave detection sub 211, Lower straightening connector 212, Cable handling vehicle 301, Cable winch 302, PLC control system 303, PC 304, Data acquisition card 305, Upper guide head 202.1, Straightening body 202.2. Detailed Implementation

[0037] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0038] Example 1, referring to Figures 1-2 The present invention discloses an ultra-deep production tubing leakage detection system, comprising a production tubing 104 and a surface control and analysis system. The system further includes an annular pressure detection unit 106, a production tubing pressure detection unit 107, a first electric throttle control valve 108, a second electric throttle control valve 109, and a downhole detection sub 200. The annular pressure detection unit 106 and the production tubing pressure detection unit 107 are installed on the wellhead device 105 and connected to the second electric throttle control valve 109 and the first electric throttle control valve 108, respectively. Pressure data is acquired in real time through the production tubing pressure detection unit 107 and the annular pressure detection unit 106, and the data is transmitted via the first electric throttle control valve 108 and the second electric throttle control valve 109. The dynamic throttling control valve 109 regulates the pressure on the tubing side and the annulus side, enabling the wellbore to form a controllable differential pressure condition while meeting safety requirements. The downhole detection sub 200 is connected to and lowered into the production tubing 104 via an armored cable 201. The production tubing 104 is located inside the casing 103 downhole. The annulus between the production tubing 104 and the casing 103, and located above the packer 101, is filled with annulus medium 102, which is water. The downhole detection sub 200 moves along the well depth direction to perform continuous or fixed-point detection. The detected multi-frequency signals are aggregated by the data transmission sub 203 and uploaded to the surface control and analysis system to complete the processing, detection data analysis, and leakage result output.

[0039] Reference Figure 2The downhole detection sub 200 mentioned in this invention includes an armored cable 201, an upper straightening connector 202, a data transmission sub 203, an ultrasonic sensor 204, an ultrasonic detection sub 205, a temperature sensor 206, a temperature detection sub 207, a pressure sensor 208, a pressure detection sub 209, a low-frequency acoustic wave sensor 210, a low-frequency acoustic wave detection sub 211, and a lower straightening connector 212. The upper end of the upper straightening connector 202 is connected to the armored cable 201, and the lower end is connected to the data transmission sub 203. The ultrasonic detection sub 205, temperature detection sub 207, pressure detection sub 209, and low-frequency acoustic wave detection sub 211 are connected below the data transmission sub 203. The ultrasonic detection sub 205... An ultrasonic sensor 204 is installed inside the production tubing to collect high-frequency acoustic signals caused by leakage. A temperature sensor 206 is installed inside the temperature detection subsection 207 to obtain the downhole ambient temperature and the fluid temperature inside the production tubing 104, so as to reflect the local temperature disturbance characteristics caused by leakage and to perform condition compensation and drift correction on the acoustic signals. A pressure sensor 208 is installed inside the pressure detection subsection 209 to collect the fluid pressure change signal and pressure pulsation signal inside the production tubing. A low-frequency acoustic wave sensor 210 is installed inside the low-frequency acoustic wave detection subsection 211 to collect the low-frequency acoustic signal and fluid disturbance noise signal caused by leakage. The lower part of the low-frequency acoustic wave detection subsection 211 is connected to the lower centering connector 212.

[0040] The aforementioned upper straightening connector 202 includes an upper guide head 202.1 and a straightening body 202.2. The upper end of the upper guide head 202.1 is connected to the armored cable 201, and the lower end of the straightening body 202.2 is connected to the upper end of the data transmission sub-section 203. The outer diameter of the straightening body 202.2 is larger than the outer diameter of the data transmission sub-section 203, the ultrasonic detection sub-section 205, the temperature detection sub-section 207, the pressure detection sub-section 209, and the low-frequency acoustic wave detection sub-section 211. The lower straightening connector 212 is similar to the upper straightening connector 202 and will not be described further.

[0041] Furthermore, the aforementioned annular pressure detection unit 106 and production tubing pressure detection unit 107 employ pressure gauges.

[0042] Reference Figure 1The ground control and analysis system mentioned in this invention includes a cable handling vehicle 301, a cable winch 302, a PLC control system 303, a PC 304, and a data acquisition card 305. The cable handling vehicle 301 drives the armored cable 201 to move up and down through the cable winch 302, causing the downhole inspection sub 200 to move up and down along the inner cavity of the production tubing 104 for inspection. The downhole inspection sub 200 uploads the inspection data to the PC 304 through the data acquisition card 305, and then the PLC control system 303 controls the operation of the first electric throttle control valve 108, the second electric throttle control valve 109, and the cable winch 302.

[0043] Preferably, the annular pressure detection unit 106 and the production string pressure detection unit 107 monitor the pressure of the annulus and the pressure of the production string 104, respectively, and transmit the detection data to the PC 304 for analysis and judgment. The PC 304 generates control commands based on the target pressure difference and sends them to the PLC control system 303. The PLC control system 303 controls the opening of the first electric throttle control valve 108 and the second electric throttle control valve 109 to establish and maintain the preset pressure difference between the production string 104 and the annulus.

[0044] The detection method of the ultra-deep production tubing leakage detection system mentioned in this invention includes the following process:

[0045] 1. Under the condition that no leakage pressure difference is established or leakage flow is induced, the downhole detection sub 200 collects ultrasonic signals, low-frequency acoustic signals, pressure signals and temperature signals in the production tubing 104 to obtain a background signal set, which is used for subsequent comparative analysis with signals under different pressure difference conditions.

[0046] 2. Based on the safety constraints and operation strategy of the production tubing 104, the opening of the first electric throttle control valve 108 and the second electric throttle control valve 109 is adjusted to establish and stabilize the corresponding target pressure difference between the production tubing 104 and the annulus. After the pressure difference level is stabilized, the armored cable 201 is moved up and down by the cable winch 302, which drives the downhole detection sub 200 to collect multi-source operating condition signals and send them to the PC 304 to obtain the set of operating condition signals under the corresponding pressure difference level.

[0047] 3. By comparing and analyzing the background signal set and the operating condition signal set using a PC 304, a background-operating condition difference enhancement feature reflecting the incremental response induced by leakage is constructed to improve the contrast between the leakage signal and the background noise.

[0048] The difference enhancement feature is represented as follows:

[0049] (1),

[0050] in,B (n) Indicates the first n One background sample, S i (n) Indicates the first i Under the first pressure differential level n One working condition sample, F (.) is a statistical function of energy, root mean square, spectral energy, pressure fluctuation amplitude, temperature disturbance amplitude, or a combination thereof; through the above comparative analysis, the regularity of leakage response with pressure difference is enhanced, and a basis is provided for subsequent cross-pressure difference evolution analysis, trend calibration and result confirmation;

[0051] IV. Synchronization or quasi-synchronization alignment, filtering and noise reduction, smoothing compensation, normalization and multi-source sample construction of multi-source signals;

[0052] Among them, the synchronization or quasi-synchronization alignment, filtering and denoising, smoothing compensation and normalization of multi-source signals are performed by using the synchronization trigger signal and / or time stamp information provided by the data transmission section 203 to align each channel; after alignment, the ultrasonic signal and the low-frequency sound wave signal are processed by bandpass filtering, bandstop filtering and / or adaptive denoising, the pressure signal and the temperature signal are processed by smoothing, and each channel is normalized.

[0053] Multi-source sample construction refers to slicing the signal sequences of each channel using a sliding window method to construct multi-source samples. Let the window length be L and the step size be S. For any channel signal sequence, multiple time window segments are extracted. After aligning each channel within the same time window, corresponding multi-source sample pairs are formed. Sample labels include at least a leak type label, a leak location interval label, and a leak intensity level label. The leak type label includes at least one or more of the following: coupling thread seal failure, tubing corrosion perforation or cracking, and packer failure. The leak location interval label is segmented according to well depth or tubing structure segments. The leak intensity level label is graded according to acoustic energy indicators and / or differential pressure response characteristics.

[0054] V. Construct cross-pressure difference evolution characteristics based on the background-operating condition difference enhancement characteristics under different pressure difference levels; then organize the difference response characteristics of each level according to the pressure difference level order to form cross-pressure difference evolution characteristics used to characterize the leakage response as a function of pressure difference, expressed as:

[0055] (2),

[0056] Among them, G (n) Indicates the first n Transpressure gradient evolution characteristics corresponding to each sample M Given the target number of differential pressure levels, Concat(.) indicates that the levels are assembled in a predetermined order.

[0057] 6. Based on the background-condition difference enhancement features, the cross-pressure difference evolution features and / or the multi-source samples, perform fusion identification to output the leakage type, leakage location range and / or leakage intensity level;

[0058] The fusion identification is implemented using a one-dimensional multi-scale convolutional network, attention network, Transformer module or a combination thereof, to extract and fusion discriminate features from multi-source signals, differential response features and / or cross-pressure difference evolution features. Through the above fusion identification, the leakage type, leakage location range and / or leakage intensity level are output, and the corresponding confidence or probability value is output at the same time.

[0059] Example 2, the detection method of the ultra-deep production tubing leakage detection system mentioned in this invention includes the following detailed process:

[0060] I. Acquisition Strategy and Signal Enhancement

[0061] S21) Background condition establishment and background signal acquisition:

[0062] Under conditions where no leakage pressure differential is established or leakage flow is induced, the downhole detection sub 200 synchronously or quasi-synchronously acquires ultrasonic signals, low-frequency acoustic signals, pressure signals, and temperature signals within the production tubing 104, obtaining a background signal set. This background signal set is then written into a signal database as background samples for subsequent comparative analysis with signals under different pressure differential conditions. Each sample in the background signal set is a one-dimensional time-domain sequence combination corresponding to the ultrasonic channel, low-frequency acoustic channel, pressure channel, and temperature channel within the same sampling window. The number of background samples is denoted as... N b ;

[0063] S22) Target differential pressure level setting:

[0064] Based on the safety constraints of the production tubing, downhole operation strategies, and target detection sensitivity, at least two target differential pressure levels are set; the differential pressure levels adopt an incremental sequence to progressively enhance the response to acoustic disturbances, pressure disturbances, and temperature disturbances induced by leakage flow; each differential pressure level is preset or adjusted online according to well depth conditions, wellbore integrity status, and allowable operation window;

[0065] S23) Differential pressure establishment and operating condition signal acquisition:

[0066] For each target differential pressure level, the opening degrees of the first electric throttle control valve 108 and the second electric throttle control valve 109 are adjusted to establish and stabilize the production tubing 104 and the annulus at the corresponding target differential pressure. After the differential pressure stabilizes, the downhole detection sub 200 performs multi-source signal acquisition to obtain the set of operating condition signals at the corresponding differential pressure level, and writes it into the signal database as leakage condition samples. The number of samples corresponding to the i-th differential pressure level is denoted as . N i ;

[0067] S24) Combined acquisition of continuous and fixed-point measurements:

[0068] To improve signal stability and repeatability, the downhole detection sub 200 employs a combined "continuous measurement + fixed-point measurement" acquisition strategy along the well depth direction; the PLC control system 303 controls the rotational speed and tension of the cable winch 302, controlling the downhole detection sub 200 to move downwards or upwards along the well depth direction at a preset travel speed v; and after each travel to a preset distance Δ... h Rear station preset time T s Fixed-point data acquisition is performed, with signals acquired during both the continuous travel and stationary phases. The stationary phase is used to obtain more stable multi-source signal segments to improve subsequent identification accuracy. In addition, during the detection process, the upper straightening connector 202 and the lower straightening connector 212 are used to center the detection instrument and avoid collision with the pipe wall, which would introduce non-leakage noise.

[0069] S25) Background—Comparison Analysis of Operating Conditions and Enhancement of Differences:

[0070] To address the challenges of strong downhole background noise and numerous operational interferences during ultra-deep well operations, this step employs a "background signal—graded differential pressure signal—contrast enhancement" approach to improve the contrast between the leakage signal and the background noise. Specifically, the background signal set obtained in step S21 is compared and analyzed with the operational signal sets under different differential pressure levels obtained in step S23 to construct a difference enhancement feature reflecting the incremental response induced by the leakage. This difference enhancement feature is expressed as follows:

[0071] (1),

[0072] in, B (n) Indicates the first n One background sample, S i (n) Indicates the first i Under the first pressure differential level n One working condition sample, F(.) is a statistical function of energy, root mean square, spectral energy, pressure pulsation amplitude, temperature disturbance amplitude, or a combination thereof; through the above comparative analysis, the regularity of leakage response with pressure difference is enhanced, and a basis is provided for subsequent cross-pressure difference evolution analysis, trend calibration and result confirmation.

[0073] II. Multi-source signal alignment, preprocessing, and multi-source sample construction

[0074] S26) Synchronous or quasi-synchronous alignment and preprocessing:

[0075] To ensure comparability between different channels and consistency of subsequent identification inputs, multi-source signals are synchronized or quasi-synchronized and aligned. The synchronization trigger signal and / or time stamp information provided by data transmission section 203 are used to align each channel. After alignment, ultrasonic signals and low-frequency sound wave signals are denoised using bandpass or bandstop filtering, pressure signals and temperature signals are smoothed, and each channel is normalized. Normalization employs zero-mean unit variance normalization or minimum-maximum normalization to reduce the impact of dimensional differences between different channels on model training and inference.

[0076] S27) Construction of multi-source samples:

[0077] The one-dimensional time-domain sequences of each channel processed in step S26 are sliced ​​using a sliding window method to construct multi-source samples. The window length is L, and the step size is S. For any channel signal sequence, multiple time window segments are extracted. After each channel is aligned within the same time window, corresponding multi-source sample pairs are formed. The sample labels include at least one or a combination of leakage type, leakage location interval, and leakage intensity level labels. The leakage type includes at least one or more of the following: coupling thread seal failure, tubing corrosion perforation or cracking, and packer failure. The leakage location interval is segmented according to well depth or tubing structure segments. The leakage intensity level is graded according to acoustic energy indicators and / or differential pressure response characteristics.

[0078] III. Characterization of Differential Response, Construction and Fusion Identification of Transpressure Difference Evolution Features

[0079] S28) Characterization of differential response, construction and fusion identification of cross-pressure difference evolution features:

[0080] Furthermore, the present invention does not only identify the original multi-source time-domain samples under a single pressure difference level, but combines the difference enhancement results constructed in step S25 to characterize and organize the multi-source responses under different pressure difference levels, thereby forming an input representation that is more consistent with the leakage mechanism.

[0081] Specifically, firstly, based on the comparison results between the background signal and the operating condition signals at each differential pressure level, multi-source differential response characteristics are constructed; then, the differential response characteristics at each level are organized according to the differential pressure level order to form a cross-differential pressure evolution characteristic used to characterize the variation of leakage response with differential pressure, expressed as:

[0082] (2),

[0083] Among them, G (n) Indicates the first n Transpressure gradient evolution characteristics corresponding to each sample M The target number of differential pressure levels is defined by `Concat(.)`, which indicates that the data is concatenated in a predetermined order. The cross-differential pressure evolution characteristics are constructed using a sequential concatenation method.

[0084] The fusion identification is implemented using a one-dimensional multi-scale convolutional network, attention network, Transformer module or a combination thereof, to extract and fusion discriminate features from multi-source time-domain signals, differential response features and / or cross-pressure difference evolution features; through the above fusion identification, the leakage type, leakage location range and / or leakage intensity level are output, and the corresponding confidence or probability value is output simultaneously;

[0085] The fusion recognition method specifically includes: employing an end-to-end fusion modeling approach for one-dimensional time-domain data to avoid complex manual feature extraction; and using a concatenation method for multi-source fusion input, as shown in the following formula:

[0086] ,

[0087] Wherein, Concat means concatenating the input vector / tensor according to the channel dimension or feature dimension to obtain the fused input vector / tensor;

[0088] The discriminant model is a one-dimensional multi-scale convolutional neural network (1D-MCNN), which can be combined with a Transformer self-attention fusion module; the multi-scale convolution of 1D-MCNN can be achieved by setting different convolutional kernel lengths. r 1 ,r 2 ,…,r j For fused input Parallel convolutional feature extraction is performed, and the outputs at each scale are concatenated to obtain multi-scale local features, as shown in the following formula:

[0089] ,

[0090] When a Transformer is introduced, self-attention can be used to perform weighted fusion of time segments and different features. The self-attention calculation can be expressed as follows:

[0091] ,

[0092] Where Q, K, and V are the query, key, and value matrices, respectively. d k The key vector dimension is used; the output of the above model includes at least the leakage type, leakage location range and leakage intensity level, and can output the corresponding confidence level or probability value.

[0093] IV. Result calibration, similarity verification, and closed-loop confirmation of retesting based on pressure difference evolution trends

[0094] S29) Online identification and trend calibration based on differential pressure levels:

[0095] After receiving the multi-source signals uploaded by the downhole detection sub 200, the ground control and analysis system completes alignment, preprocessing, sample construction, differential response characterization, cross-pressure difference evolution feature construction and fusion identification according to steps S26 to S28, and performs online inference on the current sample to be identified;

[0096] To enhance the interpretability of the output results and reduce the false alarm rate under complex operating conditions, this step further introduces a trend calibration rule based on differential pressure levels: the background signal set obtained in step S21 is compared with the operating condition signal sets obtained in steps S23-S24 at different differential pressure levels to obtain the difference response characteristics under different differential pressure levels, and based on this, it is determined whether the leakage-related response shows a preset trend as the differential pressure increases; in addition, the trends of ultrasonic signal energy, low-frequency sound wave energy, pressure pulsation amplitude, and temperature disturbance amplitude are evaluated; the trend consistency can be expressed as:

[0097] (3),

[0098] in, Indicates the first n The trend consistency score of each sample, where 1(.) is the indicator function. τ To allow for fluctuation thresholds; when the trend consistency score reaches a preset threshold, or when the indicator shows monotonic enhancement, stage enhancement and / or reaches a preset threshold relative to the background signal at at least two differential pressure levels, it is determined that it is consistent with the leakage mechanism, thereby improving the confidence of the identification result and serving as one of the criteria for result confirmation; conversely, when the trend does not meet the preset rules, the confidence of the corresponding identification result is reduced, or the result is marked as pending review or retesting;

[0099] S30) Database similarity retrieval review and retest closed-loop confirmation:

[0100] This step combines similarity retrieval and verification with the signal database: the current sample to be identified is compared with the background samples, leaked samples, and interference samples in the signal database; if the similarity between the current sample and the historical leaked samples is higher than the preset threshold, and the similarity with the background samples and / or interference samples is lower than the preset threshold, then the current identification result is confirmed as the final detection conclusion; otherwise, if the confidence of the identification result is insufficient, the trend calibration does not meet the preset rules, or the similarity with the background samples and / or interference samples is high, then the result is marked as "to be retested" or "to be manually verified";

[0101] Furthermore, when the result is marked as "to be retested", the ground control and analysis system outputs retesting suggestions and triggers the retesting process. The retesting suggestions include: adjusting the target differential pressure level between the production tubing 104 and the annulus, extending the static acquisition time Ts during the fixed-point measurement stage, reducing the travel speed v of the downhole detection sub to change the interval venting strategy, or re-executing the operating condition acquisition and identification steps. Through the above-mentioned "graded differential pressure induction - background / operating condition difference enhancement - fusion identification - differential pressure trend calibration - database retrieval verification - retesting closed-loop confirmation" method, the accuracy, stability, and repeatability of leak identification are improved under ultra-deep and complex operating conditions, and the interpretability and engineering applicability of the detection conclusions are enhanced.

[0102] The above description is merely a partial preferred embodiment of the present invention. Any person skilled in the art can modify the above-described technical solutions or modify them into equivalent technical solutions. Therefore, any simple modifications or equivalent transformations made based on the technical solutions of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. An ultra-deep production string leak detection system comprising a production string (104) and a surface control and analysis system, characterized by: It also includes an annular pressure detection unit (106), a production tubing pressure detection unit (107), a first electric throttle control valve (108), a second electric throttle control valve (109), and a downhole detection sub (200). The annular pressure detection unit (106) and the production tubing pressure detection unit (107) are installed on the wellhead device (105) and connected to the second electric throttle control valve (109) and the first electric throttle control valve (108), respectively. The pressure is obtained in real time through the production tubing pressure detection unit (107) and the annular pressure detection unit (106). Pressure data is collected, and the pressure on the tubing side and the annulus side is adjusted through the first electric throttle control valve (108) and the second electric throttle control valve (109) to form a controllable differential pressure condition in the wellbore under the premise of meeting safety conditions; the downhole detection sub (200) is connected to the production tubing (104) through the armored cable (201) and moved along the well depth direction to perform continuous detection or fixed-point detection. The multi-frequency signals detected are collected by the data transmission sub (203) and uploaded to the ground control and analysis system to complete the processing, detection data analysis and leakage result output; The downhole detection sub (200) includes an armored cable (201), an upper straightening connector (202), a data transmission sub (203), an ultrasonic sensor (204), an ultrasonic detection sub (205), a temperature sensor (206), a temperature detection sub (207), a pressure sensor (208), a pressure detection sub (209), a low-frequency acoustic sensor (210), a low-frequency acoustic detection sub (211), and a lower straightening connector (212). The upper straightening connector (202) is connected to the armored cable (201) at its upper end and to the data transmission sub (203) at its lower end. The ultrasonic detection sub (205), temperature detection sub (207), pressure detection sub (209), and low-frequency acoustic detection sub (211) are connected below the data transmission sub (203). An ultrasonic sensor (204) is installed inside the acoustic detection section (205) to collect high-frequency acoustic signals caused by leakage in the production tubing. A temperature sensor (206) is installed inside the temperature detection section (207) to obtain the downhole ambient temperature and the fluid temperature inside the production tubing (104) to reflect the local temperature disturbance characteristics caused by leakage and to perform condition compensation and drift correction on the acoustic signals. A pressure sensor (208) is installed inside the pressure detection section (209) to collect the fluid pressure change signal and pressure pulsation signal inside the production tubing. A low-frequency acoustic wave sensor (210) is installed inside the low-frequency acoustic wave detection section (211) to collect the low-frequency acoustic signal and fluid disturbance noise signal caused by leakage. The lower part of the low-frequency acoustic wave detection section (211) is connected to the lower centering connector (212).

2. The ultra-deep production string leak detection system of claim 1, wherein: The upper straightening connector (202) includes an upper guide head (202.1) and a straightening body (202.2). The upper end of the upper guide head (202.1) is connected to the armored cable (201), and the lower end of the straightening body (202.2) is connected to the upper end of the data transmission sub (203). The outer diameter of the straightening body (202.2) is larger than the outer diameter of the data transmission sub (203), the ultrasonic detection sub (205), the temperature detection sub (207), the pressure detection sub (209), and the low-frequency acoustic wave detection sub (211).

3. The ultra-deep production string leak detection system of claim 2, wherein: The annular pressure detection unit (106) and the production tubing pressure detection unit (107) use pressure gauges.

4. The ultra-deep production tubing leakage detection system according to claim 3, characterized in that: The ground control and analysis system includes a cable handling vehicle (301), a cable winch (302), a PLC control system (303), a PC (304), and a data acquisition card (305). The cable handling vehicle (301) drives the armored cable (201) to move up and down through the cable winch (302), so that the downhole inspection sub (200) moves up and down along the inner cavity of the production string (104) for inspection. The downhole inspection sub (200) uploads the inspection data to the PC (304) through the data acquisition card (305), and then controls the operation of the first electric throttle control valve (108), the second electric throttle control valve (109), and the cable winch (302) through the PLC control system (303).

5. The ultra-deep production string leak detection system of claim 4, wherein: The annular pressure detection unit (106) and the production string pressure detection unit (107) respectively monitor the pressure of the annulus and the pressure of the production string (104), and transmit the detection data to the PC (304) for analysis and judgment. The PC (304) generates control commands based on the target pressure difference and sends them to the PLC control system (303). The PLC control system (303) controls the opening of the first electric throttle control valve (108) and the second electric throttle control valve (109) to establish and maintain the preset pressure difference between the production string (104) and the annulus.

6. A method of detecting a leak in a super-deep production string leak detection system as defined in claim 5, characterized by Includes the following processes:

1. Under the condition that no leakage pressure difference is established or leakage flow is induced, the ultrasonic signal, low frequency sound wave signal, pressure signal and temperature signal in the production string (104) are collected by the downhole detection sub (200) to obtain the background signal set, which is used for subsequent comparison and analysis with signals under different pressure difference conditions.

2. Based on the safety constraints and operation strategy of the production tubing (104), the opening of the first electric throttle control valve (108) and the second electric throttle control valve (109) is adjusted to establish and stabilize the corresponding target pressure difference between the production tubing (104) and the annulus. After the pressure difference level is stabilized, the armored cable (201) is moved up and down by the cable winch (302) to drive the downhole detection sub (200) to collect multi-source operating condition signals and send them to the PC (304) to obtain the set of operating condition signals under the corresponding pressure difference level.

3. By comparing and analyzing the background signal set and the operating condition signal set through a PC (304), a background-operating condition difference enhancement feature reflecting the incremental response induced by leakage is constructed to improve the contrast between the leakage signal and the background noise. The difference enhancement feature is represented as follows: （1）， in, B (n) Indicates the first n One background sample, S i (n) Indicates the first i Under the first pressure differential level n One working condition sample, F (.) is a statistical function of energy, root mean square, spectral energy, pressure fluctuation amplitude, temperature disturbance amplitude, or a combination thereof; through the above comparative analysis, the regularity of leakage response with pressure difference is enhanced, and a basis is provided for subsequent cross-pressure difference evolution analysis, trend calibration and result confirmation; IV. Synchronization or quasi-synchronization alignment, filtering and noise reduction, smoothing compensation, normalization and multi-source sample construction of multi-source signals; Among them, the synchronization or quasi-synchronization alignment, filtering and noise reduction, smoothing compensation and normalization of multi-source signals are performed by using the synchronization trigger signal and / or time stamp information provided by the data transmission section (203) to align each channel; after alignment, the ultrasonic signal and the low-frequency sound wave signal are processed by bandpass filtering, bandstop filtering and / or adaptive noise reduction, the pressure signal and the temperature signal are processed by smoothing, and each channel is normalized. Multi-source sample construction refers to slicing the signal sequences of each channel using a sliding window method to construct multi-source samples. Let the window length be L and the step size be S. For any channel signal sequence, multiple time window segments are extracted. After aligning each channel within the same time window, corresponding multi-source sample pairs are formed. Sample labels include at least a leak type label, a leak location interval label, and a leak intensity level label. The leak type label includes at least one or more of the following: coupling thread seal failure, tubing corrosion perforation or cracking, and packer failure. The leak location interval label is segmented according to well depth or tubing structure segments. The leak intensity level label is graded according to acoustic energy indicators and / or differential pressure response characteristics. V. Construct cross-pressure difference evolution characteristics based on the background-operating condition difference enhancement characteristics under different pressure difference levels; then organize the difference response characteristics of each level according to the pressure difference level order to form cross-pressure difference evolution characteristics used to characterize the leakage response as a function of pressure difference, expressed as: （2）， wherein G (n) represents the cross-pressure difference evolution feature corresponding to the i n th sample, M is the target number of pressure difference levels, and Concat(.) represents concatenation in a predetermined order.

6. Based on the background-condition difference enhancement features, the cross-pressure difference evolution features and / or the multi-source samples, perform fusion identification to output the leakage type, leakage location range and / or leakage intensity level; The fusion identification is implemented using a one-dimensional multi-scale convolutional network, attention network, Transformer module or a combination thereof, to extract and fusion discriminate features from multi-source signals, differential response features and / or cross-pressure difference evolution features. Through the above fusion identification, the leakage type, leakage location range and / or leakage intensity level are output, and the corresponding confidence or probability value is output at the same time.

7. The method of claim 6, wherein the method further comprises: To enhance the interpretability of the output results and reduce the false alarm rate under complex operating conditions, a trend calibration rule based on differential pressure levels is introduced: the obtained background signal set is compared with the operating condition signal set obtained at different differential pressure levels to obtain the difference response characteristics under different differential pressure levels, and based on this, it is determined whether the leakage-related response shows a preset trend as the differential pressure increases; further, the trends of ultrasonic signal energy, low-frequency sound wave energy, pressure pulsation amplitude, and temperature disturbance amplitude are evaluated; the trend consistency can be expressed as: （3）， in, Indicates the first n The trend consistency score of each sample, where 1(.) is the indicator function. τ To allow for fluctuation thresholds; when the trend consistency score reaches a preset threshold, or when the indicator shows monotonic enhancement, stage enhancement and / or reaches a preset threshold relative to the background signal at at least two differential pressure levels, it is determined that it is consistent with the leakage mechanism, thereby improving the confidence of the identification result and serving as one of the criteria for result confirmation; conversely, when the trend does not meet the preset rules, the confidence of the corresponding identification result is reduced, or the result is marked as pending review or retesting.

Images