Reservoir inherent frequency downhole in-situ measurement method based on neural network filtering

Abstract

The invention provides a reservoir inherent frequency downhole in-situ measurement method based on neural network filtering. The method comprises the steps that (1) pulse shock waves are generated through a continuous high-voltage pulse discharging device to induce reservoir rock mass vibration; (2) acquiring natural vibration noise signal data of the continuous high-voltage pulse discharge device during pulse discharge at the wellhead; (3) acquiring a background noise signal of the continuous high-voltage pulse discharge device in a static state in a shaft and feedback vibration signal data during pulse discharge; (4) filtering and noise reduction are carried out on the high-noise background data; (5) carrying out vibration signal analysis; (6) obtaining a Hilbert marginal spectrum, and determining a correlation coefficient between each intrinsic mode and an original vibration signal; and (7) determining the inherent frequency of the target reservoir. According to the method, the inherent frequency measurement of the target reservoir can be directly carried out underground, the problem that the underground inherent frequency measurement is difficult or the indoor inherent frequency test is inaccurate is effectively solved, and the underground in-situ efficient acquisition of the inherent frequency of the reservoir is realized.

Description

Technical Field

[0001] The invention belongs to the technical field of rock physics and well logging, and in particular relates to a downhole in-situ measurement method of reservoir natural frequency based on neural network filtering. Background Art

[0002] Unconventional oil and gas, such as shale oil and gas, is an important area for increasing oil and gas reserves and production. Since this type of reservoir is dense, low-porosity and low-permeability, it is necessary to improve production capacity through reservoir transformation. In order to overcome the problems of complex process, high cost and high environmental pressure of traditional hydraulic fracturing reservoir transformation technology, new water-free reservoir transformation technology is constantly innovating and developing. At present, a reservoir resonance transformation technology based on pulse vibration and resonance effect has been proposed. This technology has the characteristics of high efficiency, low cost and environmental protection, and has broad application prospects. Taking shale reservoirs as an example, since the displacement response amplitude of the rock mass is the largest in the resonance state, by continuously applying a certain frequency of pulse force to the formation around the well and exciting the reservoir to reach a resonance state, the tensile stress of the rock can overcome the ground stress and tensile strength under the condition of a small single excitation energy or force, so that the original cracks in the formation can be expanded or new cracks can be generated, the permeability of the shale reservoir can be improved, and then the single well production capacity can be improved.

[0003] In reservoir resonance transformation technology, it is necessary to set an external excitation frequency according to the natural frequency of the target reservoir to stimulate the reservoir to achieve resonance. Therefore, how to accurately determine the natural frequency of the rock mass within the transformation target range is the key to implementing reservoir resonance transformation. At present, the main means of testing the natural frequency of an object are the "hammer method" and the "sweep frequency method". Among them, the "hammer method" has the advantages of high efficiency and strong universality. It is mostly used for vibration characteristic analysis of buildings, bridges and mechanical structures. By monitoring the surface vibration signal of the object under excitation, the structural response data is obtained, and the natural frequency of the object under test is extracted using signal processing methods. The "sweep frequency method" is intuitive and highly accurate. By collecting the response changes of the object under test at different excitation frequencies, the resonance frequency with a more obvious displacement response is directly determined based on the amplitude-frequency curve, and the vibration characteristics of the object under the natural frequency can also be intuitively obtained.

[0004] The natural frequency of an object is determined by the size and properties of the material. Small-scale rock samples differ greatly from the original reservoir rock mass in terms of size and physical properties. Therefore, the natural frequency test results of small-scale rock samples cannot be used as the natural frequency of the reservoir rock mass. The latter measurement should be carried out in situ underground to accurately obtain the natural frequency of the rock mass in a specific range under original conditions.

[0005] The current downhole in-situ measurement method of reservoir natural frequency is similar to the "hammer method", which induces reservoir vibration by applying pulse excitation, and analyzes the monitored reservoir vibration signal to extract the characteristic frequency. However, due to the great influence of structures such as casing, perforation, cement ring, etc. on vibration propagation, coupled with the heterogeneity of the formation and liquid interference in the wellbore, the vibration signal collected downhole is very complex, and data analysis is difficult. It is difficult to quickly and accurately extract the reservoir natural frequency. Therefore, it is necessary to innovate the measurement and analysis method of reservoir natural frequency to achieve efficient acquisition of natural frequency. Summary of the invention

[0006] In response to the technical problems existing in the above-mentioned background technology, the present invention proposes a method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering. The method has a reasonable concept and can directly carry out the measurement of target reservoir natural frequency downhole, effectively solving the problem of difficulty in downhole natural frequency measurement or inaccurate indoor natural frequency testing, and realizing efficient in-situ acquisition of reservoir natural frequency downhole.

[0007] In order to solve the above technical problems, the present invention provides a downhole in-situ measurement method of reservoir natural frequency based on neural network filtering, which mainly includes the following steps:

[0008] (1) First, a pulse shock wave is generated by a continuous high-voltage pulse discharge device to induce reservoir rock vibration;

[0009] (2) Based on distributed optical fiber acoustic wave sensing, the self-vibration noise signal data of the continuous high-voltage pulse discharge device during pulse discharge at the wellhead is obtained;

[0010] (3) Based on distributed optical fiber acoustic wave sensing, the background noise signal of the continuous high-voltage pulse discharge device in the wellbore at rest and the feedback vibration signal data during pulse discharge are obtained;

[0011] (4) filtering and denoising the self-vibration noise signal data, background noise signal and feedback vibration signal data during pulse discharge based on a deep residual shrinkage neural network;

[0012] (5) Use the improved adaptive complete empirical mode decomposition method and Fourier transform method to carry out vibration signal analysis;

[0013] (6) Performing Hilbert transform again to obtain the Hilbert marginal spectrum reflecting the changing trend of the eigenmode, and combining the Pearson correlation coefficient method to determine the correlation coefficient between each eigenmode and the original vibration signal;

[0014] (7) Finally, the inherent frequency of the target reservoir is determined based on the Hilbert marginal spectrum and the correlation coefficient.

[0015] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein the specific process of step (1) is as follows:

[0016] (1.1) Locate the perforation position of the oil and gas well based on the logging device, and record the perforation boundary position, target reservoir position and non-perforation section position;

[0017] (1.2) Bundling the armored optical fiber with the cable of the continuous high-voltage pulse discharge device to achieve synchronous retraction and release of the optical fiber and the cable of the continuous high-voltage pulse discharge device;

[0018] (1.3) Use a winch to lower a continuous high-voltage pulse discharge device to the wellhead and required locations underground and perform single or frequency-increasing continuous pulse discharges to generate pulse vibrations and stimulate vibrations in the formations around the well.

[0019] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein: the optical fiber used in the step (1.2) is a distributed acoustic wave sensing optical fiber.

[0020] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein: the cable of the continuous high-voltage pulse discharge device in step (1.2) can also use an optoelectronic composite cable containing optical fiber.

[0021] The method for measuring the reservoir natural frequency downhole in situ based on neural network filtering, wherein the specific process of step (2) is: using a winch to lift the continuous high-voltage pulse discharge device and suspend it at the wellhead, controlling the continuous high-voltage pulse discharge device to perform single discharge and continuous discharge with increasing frequency, and collecting the continuous high-voltage pulse discharge device's own vibration signal under this condition through optical fiber, and this signal is called the continuous high-voltage pulse discharge device's self-vibration noise.

[0022] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein the specific process of step (3) is as follows:

[0023] (3.1) A continuous high-voltage pulse discharge device is lowered by a winch, and the optical fiber is lowered synchronously with the cable of the continuous high-voltage pulse discharge device. When the continuous high-voltage pulse discharge device reaches the perforation section, the lowering is stopped. After the continuous high-voltage pulse discharge device is judged to be stationary based on the vibration signal monitored by the optical fiber, the optical fiber is used to monitor and record the downhole background vibration signal data, and the signal is named as the perforation section background noise; then the continuous high-voltage pulse discharge device is controlled to perform a single discharge and continuous discharge with increasing frequency at this location, and the generated pulse shock wave excites the vibration of the formation rock mass around the well, and the downhole vibration signal is collected through the optical fiber during the discharge process;

[0024] (3.2) Continue to lower the continuous high-voltage pulse discharge device, and stop lowering the continuous high-voltage pulse discharge device when it reaches the target reservoir position, monitor the downhole environment based on the optical fiber, and start recording data after the continuous high-voltage pulse discharge device stops and names it as the target reservoir background noise; then control the continuous high-voltage pulse discharge device to perform a single discharge and continuous discharge with increasing frequency at that location, and collect vibration signals through the optical fiber during the discharge process;

[0025] (3.3) Continue to lower the continuous high-voltage pulse discharge device. When the continuous high-voltage pulse discharge device reaches the non-perforation section, stop lowering the continuous high-voltage pulse discharge device. Monitor the downhole environment based on the optical fiber. After the continuous high-voltage pulse discharge device stops, start recording data and name it as the non-perforation section background noise. Then control the continuous high-voltage pulse discharge device to perform a single discharge and continuous discharge with increasing frequency at this location. During the discharge process, collect vibration signals through the optical fiber.

[0026] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein the specific process of step (4) is: based on the deep residual shrinkage neural network, the vibration signal background noise is filtered out, the self-vibration noise of the continuous high-voltage pulse discharge device, the perforation section background noise, the target reservoir background noise, and the non-perforation section background noise are used as training samples, and the vibration signals collected during the single discharge and the continuous discharge with increasing frequency at the corresponding position are subjected to noise reduction filtering processing based on the deep residual shrinkage neural network.

[0027] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein the specific process of step (5) is as follows:

[0028] (5.1) Based on the improved adaptive complete empirical mode decomposition method, modal analysis is performed on the vibration signal after noise reduction filtering, the intrinsic mode components obtained by signal decomposition are Fourier transformed, the eigenmodes within the required frequency range are extracted, and the extracted eigenmodes are reconstructed into the vibration signal;

[0029] (5.2) Based on the improved adaptive complete empirical mode decomposition method, modal analysis is performed on the reconstructed vibration signal. The eigenmode components obtained by signal decomposition are subjected to Fourier transform. The eigenmodes with main frequencies lower than the required frequency range are subjected to Hilbert transform. The occurrence time of the main frequencies of each mode is analyzed to obtain the Hilbert marginal spectrum reflecting the change trend of the eigenmodes.

[0030] (5.3) The shock wave velocity is calculated based on the difference in the starting time of the shock response between different channels of the optical fiber and the distance between each channel. Then, the time it takes for the shock wave to reach the wall of each structure is calculated based on the size of each foundation structure. The time required for the stress wave caused by the vibration of each foundation structure and the target reservoir rock mass to be transmitted to the optical fiber is inverted based on the stress wave velocity in the medium of each foundation structure.

[0031] The method for measuring reservoir natural frequency downhole in situ based on neural network filtering, wherein the specific process of step (6) is: statistically analyzing and dividing the main frequencies of the intrinsic modes extracted from all vibration signals into intervals, and comparing the main frequency occurrence time in the Hilbert marginal spectrum of the samples in each interval with the theoretical transmission time, so as to provide a reference for the identification of the natural frequencies of each structure, and then extracting the natural frequency of the target reservoir rock mass.

[0032] The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering, wherein the specific process of step (7) is: based on the strong tracking filter, the data points in the natural frequency interval corresponding to the target reservoir are estimated to obtain the universal natural frequency of the target reservoir.

[0033] By adopting the above technical solution, the present invention has the following beneficial effects:

[0034] The in-situ downhole measurement method of reservoir natural frequency based on neural network filtering of the present invention is reasonably conceived and can directly carry out the measurement of target reservoir natural frequency in the downhole, effectively solving the problem of difficult downhole natural frequency measurement or inaccurate indoor natural frequency testing, and realizing efficient in-situ acquisition of reservoir natural frequency in the downhole.

[0035] The present invention also has the following characteristics and advantages:

[0036] (1) The present invention is the first in-situ measurement method of reservoir natural frequency that combines a "tapping method" with a machine learning algorithm. The present invention uses field optical fiber monitoring to collect in-situ measurement data of natural frequency for the first time, extracts characteristic frequencies from complex vibration signals based on measured data sets and deep learning methods, and identifies reservoir rock natural frequencies based on empirical values ​​and theoretical solutions, thereby achieving in-situ, efficient and accurate measurement of reservoir natural frequency.

[0037] (2) The reservoir vibration signal is complex, including different structural vibration modes. At the same time, different shock wave excitation positions induce different structural vibration modes. The present invention can construct a database based on deep learning theory to realize dynamic monitoring and synchronous measurement of reservoir natural frequency in reservoir transformation projects, which is more in line with actual application needs.

[0038] (3) This method can be applied to various reservoir downhole continuous dynamic load transformation technologies and devices, and has the advantages of wide applicability, strong timeliness and high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 The present invention is a flow chart of the downhole in-situ measurement method of reservoir natural frequency based on neural network filtering. DETAILED DESCRIPTION

[0041] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The present invention is further explained below in conjunction with specific implementation modes.

[0043] Example 1

[0044] like Figure 1 As shown, the present embodiment 1 provides an in-situ downhole measurement method of reservoir natural frequency based on neural network filtering, which is to first generate a pulse shock wave through a continuous high-voltage pulse discharge device to induce reservoir rock vibration, then obtain feedback vibration signal data in the wellbore based on distributed optical fiber acoustic wave sensing, filter and reduce noise on the high-noise background data based on a deep residual shrinkage neural network, and then use an improved adaptive complete empirical mode decomposition-Fourier transform method to carry out vibration signal analysis, and then perform Hilbert transform to obtain the Hilbert marginal spectrum reflecting the change trend of the intrinsic mode, and combine the Pearson correlation coefficient method to clarify the correlation coefficient between each intrinsic mode and the original vibration signal, and finally determine the target reservoir natural frequency based on the Hilbert marginal spectrum and the correlation coefficient.

[0045] The present invention provides a method for measuring reservoir natural frequency downhole in-situ based on neural network filtering, which specifically comprises the following steps:

[0046] S010. Conduct wellbore depth calibration based on the γ logging device, locate the target drilling perforation section, and record the perforation boundary position, target reservoir position, and non-perforation section position. Bundle the armored optical fiber with the cable of the continuous high-voltage pulse discharge device (hereinafter referred to as the discharge device) to achieve synchronous retraction and release of the optical fiber and the discharge device cable. The optical fiber used is a distributed acoustic wave sensor optical fiber.

[0047] S020. Use a winch to lift the discharge device and suspend it at the wellhead, control the discharge device to perform single discharge and continuous discharge with increasing frequency, and collect the device's own vibration signal under this condition through optical fiber. This signal is called the device's self-vibration noise.

[0048] S030. Use a winch to lower the discharge device. The optical fiber is lowered into the well synchronously with the cable of the discharge device. When the discharge device reaches the perforation section, the lowering stops. After judging that the device is stationary based on the vibration signal monitored by the optical fiber, the optical fiber is used to monitor and record the downhole background vibration signal data. This signal is named as the perforation section background noise. Then, the discharge device is controlled to perform a single discharge and continuous discharge with increasing frequency at this location. The generated pulse shock wave excites the vibration of the formation rock mass around the well. The downhole vibration signal is collected through the optical fiber during the discharge process.

[0049] S040, continue to lower the discharge device, and stop lowering the device when it reaches the target reservoir position, monitor the downhole environment based on the optical fiber, and start recording data after the device is still and name it as the target reservoir background noise; then control the discharge device to perform a single discharge and continuous discharge with increasing frequency at this location, and collect vibration signals through the optical fiber during the discharge process.

[0050] S050, continue to lower the discharge device, and stop lowering the device when it reaches the non-perforation section. Monitor the downhole environment based on the optical fiber. After the device stops, start recording data and name it as non-perforation section background noise. Then control the discharge device to perform a single discharge and continuous discharge with increasing frequency at this location. During the discharge process, collect vibration signals through the optical fiber.

[0051] S060. Filter and reduce noise on high-noise background data (i.e., self-vibration noise signal data, background noise signal, and feedback vibration signal data during pulse discharge) based on a deep residual shrinkage neural network. The device self-vibration noise, perforation section background noise, target reservoir background noise, and non-perforation section background noise are used as training samples. The vibration signals collected during a single discharge and continuous discharge with increasing frequency at the corresponding position are subjected to noise reduction filtering processing based on a deep residual shrinkage neural network.

[0052] S070. Based on the improved adaptive complete empirical mode decomposition method, modal analysis is performed on the vibration signal after noise reduction filtering, the intrinsic mode components obtained by signal decomposition are subjected to Fourier transform, the intrinsic modes with a main frequency value lower than 50 Hz are extracted, and the extracted eigenmodes are reassembled into a reconstructed vibration signal.

[0053] S080. Perform modal analysis on the reconstructed vibration signal based on the improved adaptive complete empirical mode decomposition method, perform Fourier transform on the eigenmode components obtained by signal decomposition, perform Hilbert transform on the eigenmodes with a main frequency lower than 50 Hz, and analyze the occurrence time of the main frequency of each mode.

[0054] S090. The shock wave velocity is solved based on the difference in the starting time of the shock response between different channels of the optical fiber and the distance between each channel, and then the time for the shock wave to reach the wall of each structure is obtained in combination with the size of each foundation structure. The time required for the stress wave caused by the vibration of each foundation structure and the target reservoir rock mass to be transmitted to the optical fiber is inverted based on the stress wave velocity in the medium of each foundation structure.

[0055] S100, the main frequencies of the intrinsic modes extracted from all vibration signals are counted and divided into intervals, and the main frequency appearance time in the Hilbert spectrum of the samples in each interval is compared with the theoretical transmission time, so as to provide a reference for the identification of the natural frequency of each structure, and then extract the natural frequency of the target reservoir rock mass.

[0056] S110, estimating data points in a natural frequency interval corresponding to the target reservoir based on a strong tracking filter to obtain a universal natural frequency of the target reservoir.

[0057] Example 2

[0058] The structure of this embodiment is basically the same as that of embodiment 1, except that it is not limited to performing Hilbert transform on the eigenmode with a main frequency lower than 50 Hz, and the occurrence time of the main frequency of each mode is analyzed, and the specific frequency is adjusted according to the actual situation.

[0059] The present invention has a reasonable concept and can directly carry out the measurement of the natural frequency of the target reservoir underground, effectively solving the problem of difficult natural frequency measurement underground or inaccurate natural frequency testing indoors, and realizing efficient in-situ acquisition of the natural frequency of the reservoir underground.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring reservoir natural frequency downhole in situ based on neural network filtering, characterized in that: The main steps include: (1) First, a pulse shock wave is generated by a continuous high-voltage pulse discharge device to induce reservoir rock vibration; (2) Based on distributed optical fiber acoustic wave sensing, the self-vibration noise signal data of the continuous high-voltage pulse discharge device during pulse discharge at the wellhead is obtained; (3) Based on distributed optical fiber acoustic wave sensing, the background noise signal of the continuous high-voltage pulse discharge device in the wellbore at rest and the feedback vibration signal data during pulse discharge are obtained; (4) filtering and denoising the self-vibration noise signal data, background noise signal and feedback vibration signal data during pulse discharge based on a deep residual shrinkage neural network; (5) Use the improved adaptive complete empirical mode decomposition method and Fourier transform method to carry out vibration signal analysis; (6) Performing Hilbert transform again to obtain the Hilbert marginal spectrum reflecting the changing trend of the eigenmode, and combining the Pearson correlation coefficient method to determine the correlation coefficient between each eigenmode and the original vibration signal; (7) Finally, the inherent frequency of the target reservoir is determined based on the Hilbert marginal spectrum and the correlation coefficient.

2. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 1, characterized in that: The specific process of step (1) is as follows: (1.1) Locate the perforation position of the oil and gas well based on the logging device, and record the perforation boundary position, target reservoir position and non-perforation section position; (1.2) Bundling the armored optical fiber with the cable of the continuous high-voltage pulse discharge device to achieve synchronous retraction and release of the optical fiber and the cable of the continuous high-voltage pulse discharge device; (1.3) Use a winch to lower a continuous high-voltage pulse discharge device to the wellhead and required locations underground and perform single or increasing frequency continuous pulse discharges to generate pulse vibrations and stimulate vibrations in the formations around the well.

3. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 3 is characterized in that: The optical fiber used in the step (1.2) is a distributed acoustic wave sensing optical fiber.

4. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 3 is characterized in that: The cable of the continuous high-voltage pulse discharge device in the step (1.2) can also be an optoelectronic composite cable containing optical fiber.

5. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 1, characterized in that: The specific process of step (2) is: use a winch to lift the continuous high-voltage pulse discharge device and suspend it at the wellhead, control the continuous high-voltage pulse discharge device to perform single discharge and continuous discharge with increasing frequency, and collect the self-vibration signal of the continuous high-voltage pulse discharge device under this condition through optical fiber. This signal is called the self-vibration noise of the continuous high-voltage pulse discharge device.

6. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 1, characterized in that: The specific process of step (3) is as follows: (3.1) A continuous high-voltage pulse discharge device is lowered by a winch, and the optical fiber is lowered synchronously with the cable of the continuous high-voltage pulse discharge device. When the continuous high-voltage pulse discharge device reaches the perforation section, the lowering is stopped. After the continuous high-voltage pulse discharge device is judged to be stationary based on the vibration signal monitored by the optical fiber, the optical fiber is used to monitor and record the downhole background vibration signal data, and the signal is named as the perforation section background noise; then the continuous high-voltage pulse discharge device is controlled to perform a single discharge and continuous discharge with increasing frequency at this location, and the generated pulse shock wave excites the vibration of the formation rock mass around the well, and the downhole vibration signal is collected through the optical fiber during the discharge process; (3.2) Continue to lower the continuous high-voltage pulse discharge device, and stop lowering the continuous high-voltage pulse discharge device when it reaches the target reservoir position, monitor the downhole environment based on the optical fiber, and start recording data after the continuous high-voltage pulse discharge device stops and names it as the target reservoir background noise; then control the continuous high-voltage pulse discharge device to perform a single discharge and continuous discharge with increasing frequency at that location, and collect vibration signals through the optical fiber during the discharge process; (3.3) Continue to lower the continuous high-voltage pulse discharge device. When the continuous high-voltage pulse discharge device reaches the non-perforation section, stop lowering the continuous high-voltage pulse discharge device. Monitor the downhole environment based on the optical fiber. After the continuous high-voltage pulse discharge device stops, start recording data and name it as the non-perforation section background noise. Then control the continuous high-voltage pulse discharge device to perform a single discharge and continuous discharge with increasing frequency at this location. During the discharge process, collect vibration signals through the optical fiber.

7. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 6, characterized in that: The specific process of step (4) is as follows: based on the deep residual shrinkage neural network, the vibration signal background noise is filtered out, the self-vibration noise of the continuous high-voltage pulse discharge device, the perforation section background noise, the target reservoir background noise, and the non-perforation section background noise are used as training samples, and the vibration signals collected during the single discharge and the continuous discharge with increasing frequency at the corresponding position are subjected to noise reduction filtering processing according to the deep residual shrinkage neural network.

8. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 1, characterized in that: The specific process of step (5) is as follows: (5.1) Based on the improved adaptive complete empirical mode decomposition method, modal analysis is performed on the vibration signal after noise reduction filtering, the intrinsic mode components obtained by signal decomposition are Fourier transformed, the eigenmodes within the required frequency range are extracted, and the extracted eigenmodes are reconstructed into the vibration signal; (5.2) Based on the improved adaptive complete empirical mode decomposition method, modal analysis is performed on the reconstructed vibration signal. The eigenmode components obtained by signal decomposition are subjected to Fourier transform. The eigenmodes with main frequencies lower than the required frequency range are subjected to Hilbert transform. The occurrence time of the main frequencies of each mode is analyzed to obtain the Hilbert marginal spectrum reflecting the change trend of the eigenmodes. (5.3) The shock wave velocity is calculated based on the difference in the starting time of the shock response between different channels of the optical fiber and the distance between each channel. Then, the time it takes for the shock wave to reach the wall of each structure is calculated based on the size of each foundation structure. The time required for the stress wave caused by the vibration of each foundation structure and the target reservoir rock mass to be transmitted to the optical fiber is inverted based on the stress wave velocity in the medium of each foundation structure.

9. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 8, characterized in that: The specific process of step (6) is as follows: statistics and interval division are performed on the main frequencies of the intrinsic modes extracted from all vibration signals, and the main frequency occurrence time in the Hilbert marginal spectrum of the samples in each interval is compared with the theoretical transmission time, so as to provide a reference for the identification of the natural frequencies of each structure, and then extract the natural frequencies of the target reservoir rock mass.

10. The method for downhole in-situ measurement of reservoir natural frequency based on neural network filtering according to claim 1, characterized in that: The specific process of step (7) is: based on the strong tracking filter, the data points in the natural frequency interval corresponding to the target reservoir are estimated to obtain the universal natural frequency of the target reservoir.

Images