Microseismic recognition method based on neural network and computer readable storage medium

Abstract

The invention belongs to the field of microseism signal identification and positioning, and discloses a microseism identification method based on a neural network and a computer readable storage medium, and the method comprises the steps: enabling original seismic data to be used for initial seismic phase identification; opening a time window for the arrival time of the seismic phase identified for the first time on the seismic trace data, and calculating an amplitude value in the time window; when the amplitude value in the time window is abnormal, inputting the corresponding original seismic trace data into a validity judgment neural network, and identifying whether the original seismic trace data belongs to noise data or valid data; if the original seismic trace data is noise data, eliminating the original seismic trace data from the original seismic data to obtain noise-reduced original data, and using the noise-reduced original data for secondary seismic phase identification; and if the original seismic trace data are valid data, the original seismic trace data are reserved in the original seismic data, and the original seismic data are used for secondary seismic phase identification. According to the method, strong energy noise events are eliminated, the workload of processing personnel is greatly reduced, the labor cost is reduced, and cost reduction and efficiency improvement of microseism observation are realized.

Description

Technical Field

[0001] The present invention relates to the field of microseismic signal recognition and positioning, and more specifically, to a microseismic recognition method based on a neural network and a computer-readable storage medium. Background Art

[0002] Microseismic monitoring plays an increasingly important role in oil and gas development and production. During hydraulic fracturing, microseismic positioning results can reflect the morphology of the fracture network and the effect of reservoir transformation; in the long-term monitoring during oil and gas development and production and gas storage injection, microseismic technology can be used to monitor gas migration patterns and disaster warnings.

[0003] Microseismic event identification is the basis of microseismic monitoring and positioning. Currently, a common method is to apply the long-short time window method (STA / LTA) on the stacked trace, mainly based on the sudden change of amplitude on the stacked trace to identify microseismic events. This method can enhance weak signals with low signal-to-noise ratio and suppress irregular noise by in-phase stacking. However, in actual microseismic data, due to interference from industrial or civil facilities, seismic traces or stations near the interference source will generate strong amplitude noise, generate strong energy in the stacked trace, and be misjudged as microseismic events; this part of the noise interference reduces the recognition rate of microseismic events and produces erroneous positioning results, so it needs to be manually identified and eliminated by quality control in real time, which greatly increases the workload of processing personnel.

[0004] In recent years, intelligent processing technology in the field of natural earthquake and microseismic signal processing has developed rapidly. Taking the picking technology as an example, the convolutional neural network trained based on artificial picking labels has learned a series of characteristics of microseismic events, which can efficiently distinguish effective signals from noise signals, with an accuracy comparable to that of professional processing personnel. However, intelligent picking technology can often only pick up the arrival time information of microseismic signals, and cannot enhance and identify weak signals by waveform in-phase superposition, and its recognition ability for low signal-to-noise ratio data is insufficient. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems existing in the prior art and to provide a microseismic identification method based on a neural network.

[0006] The present invention is implemented by the following technical scheme: a microseismic identification method based on neural network, comprising the following steps:

[0007] The original seismic data are used for initial seismic phase identification;

[0008] A time window is opened for the arrival time of the first identified seismic phase on the seismic trace data, and the amplitude value within the time window is calculated;

[0009] When the amplitude value in the time window is abnormal, the corresponding original seismic trace data is input into the validity judgment neural network to identify whether the original seismic trace data is noise data or valid data;

[0010] If it is noise data, the original seismic trace data is removed from the original seismic data to obtain noise-reduced original data, and the noise-reduced original data is used for secondary seismic phase identification;

[0011] If it is valid data, the original seismic trace data is retained in the original seismic data, and the original seismic data is used for secondary seismic phase identification.

[0012] Furthermore, the same method is used to perform primary and secondary seismic phase identification: using the short / long time window method to perform seismic phase identification in the superposition trace.

[0013] Furthermore, the stacked trace is obtained after dynamic correction and horizontal stacking.

[0014] Furthermore, the seismic phase is identified according to the short / long time window amplitude R. When the short / long time window amplitude ratio is greater than a threshold, it is considered that a seismic phase exists in the short time window.

[0015] Furthermore, the calculation formula of the short / long time window amplitude R is as follows:

[0016]

[0017] Where M and N are the number of samples in the long and short time windows respectively, and X i , Y i , Z i They are the amplitude values ​​of each sample point of the three components of the seismic station in the long and short time windows respectively.

[0018] Furthermore, a time window is opened for the arrival time in the following manner: with the arrival time as the center, a window is opened for Δt seconds before and after.

[0019] Furthermore, the box plot method is used to determine whether the amplitude value within the time window is abnormal.

[0020] Furthermore, when there is no abnormality in the amplitude value within the time window, it is determined whether a microearthquake occurs based on the seismic phase obtained by the initial seismic phase identification.

[0021] Furthermore, when a micro-earthquake is identified, a spatial search method is used to scan and locate the stratum where the micro-earthquake occurs.

[0022] The present invention also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in the present invention.

[0023] Compared with the prior art, the beneficial effects of the present invention include:

[0024] 1. The present invention performs preliminary seismic phase identification, and then uses a neural network to eliminate the seismic trace data that are easily misidentified due to noise interference. After eliminating these seismic trace data from the original data, the remaining data is used for secondary seismic phase identification, eliminating high-energy noise events, greatly reducing the workload of processing personnel, reducing labor costs, and achieving cost reduction and efficiency improvement of microseismic observation.

[0025] 2. Based on the recognition results of the long and short time window method (STA / LTA) of the superimposed channels, the box plot method and intelligent neural network in statistics are used to distinguish each channel. After eliminating the strong energy noise events, the long and short time window method (STA / LTA) is used again for event discrimination.

[0026] 3. When there is no microseismic signal in the time window, this method is equivalent to an automatic quality control method, which improves the accuracy of event recognition; when strong amplitude noise in the time window submerges the effective signal, the application of this method can improve the accuracy of event recognition and positioning.

[0027] 4. The recognition point of the STA / LTA method may not be the point with the maximum amplitude, so the energy in the time window is calculated instead of directly calculating the AMP of the recognition point to improve the recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart;

[0029] Figure 2 Strong amplitude noise interference on some seismic traces provided for examples;

[0030] Figure 3 Schematic diagram of the principle of identifying outliers provided for the example. DETAILED DESCRIPTION

[0031] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0032] refer to Figure 1 As shown, a microseismic identification method based on a neural network includes the following steps:

[0033] The original seismic data are used for initial seismic phase identification;

[0034] A time window is opened for the arrival time of the first identified seismic phase on the seismic trace data, and the amplitude value within the time window is calculated;

[0035] When the amplitude value in the time window is abnormal, the corresponding original seismic trace data is input into the validity judgment neural network to identify whether the original seismic trace data is noise data or valid data;

[0036] If it is noise data, the original seismic trace data is removed from the original seismic data to obtain noise-reduced original data, and the noise-reduced original data is used for secondary seismic phase identification;

[0037] If it is valid data, the original seismic trace data is retained in the original seismic data, and the original seismic data is used for secondary seismic phase identification.

[0038] When there is no abnormality in the amplitude value within the time window, the seismic phase obtained by the initial seismic phase identification is used to determine whether a microearthquake has occurred.

[0039] Preferably, the same method is used to perform the primary seismic phase identification and the secondary seismic phase identification: the short / long time window method is used to perform the seismic phase identification in the superposition trace.

[0040] Preferably, the stacked trace is obtained after dynamic correction and horizontal stacking.

[0041] Preferably, the seismic phase is identified based on the short / long time window amplitude R, and when the short / long time window amplitude ratio is greater than a threshold, it is considered that a seismic phase exists in the short time window.

[0042] Preferably, the calculation formula of the short / long time window amplitude R is as follows:

[0043]

[0044] Where M and N are the number of samples in the long and short time windows respectively, and X i , Y i , Z i They are the amplitude values ​​of each sample point of the three components of the seismic station in the long and short time windows respectively.

[0045] Preferably, the time window is opened for the arrival time in the following manner: with the arrival time as the center, each window is opened for Δt seconds before and after.

[0046] Preferably, a box plot method is used to determine whether the amplitude value within the time window is abnormal.

[0047] Preferably, after a micro-earthquake is identified, a spatial search method is used to scan and locate the stratum where the micro-earthquake occurred.

[0048] Based on the recognition results of the long and short time window method (STA / LTA) of the superimposed traces, this specific implementation method uses the box plot method in statistics and the intelligent neural network to distinguish each trace, removes the strong energy noise events, and then uses the long and short time window method (STA / LTA) again for event discrimination. This intelligent quality control method can improve the event recognition rate, greatly reduce the workload of processing personnel, reduce the employment cost, and achieve cost reduction and efficiency improvement of microseismic observation.

[0049] A more specific description is given below in conjunction with embodiments.

[0050] Example 1

[0051] The same method is used to identify the primary and secondary seismic phases: the short / long time window method is used to identify the seismic phases in the stacked traces. The stacked traces are obtained after dynamic correction and horizontal stacking.

[0052] After the microseismic profile is subjected to dynamic correction and horizontal stacking, the STA / LTA method is used to identify the seismic phases in the stacked traces, and the arrival time of the strong energy stacked traces that meet the threshold (the value can be set empirically) is stored. Figure 2 shown.

[0053] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0054] Example 2

[0055] The seismic phase is identified based on the short / long time window amplitude R. When the short / long time window amplitude ratio is greater than the threshold, it is considered that there is a seismic phase in the short time window. The calculation formula of the short / long time window amplitude R is as follows:

[0056]

[0057] Where M and N are the number of samples in the long and short time windows respectively, and X i , Y i , Z i They are the amplitude values ​​of each sample point of the three components of the seismic station in the long and short time windows respectively.

[0058] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0059] Example 3

[0060] The time window is opened as follows: with the time of arrival as the center, each window is opened for Δt seconds before and after, where Δt is 1 to 5 seconds, preferably 2 seconds.

[0061] At the identification point, that is, before and after the arrival time, a 2s time window is opened (not belonging to the time window in STA / LTA), and the amplitude value AMP within the time window is calculated for each seismic channel corresponding to each station. The calculation formula is as follows:

[0062]

[0063] AMP is the amplitude value in each time window (2s before and after the identification point).

[0064] The identification point of the STA / LTA method may not be the point of maximum amplitude, so the energy within the time window is calculated instead of directly calculating the AMP of the identification point.

[0065] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0066] Example 4

[0067] refer to Figure 3 As shown, the box plot method is used to determine whether the amplitude value in the time window is abnormal. The box plot method is a very effective way to determine abnormal values ​​in statistics. The principle of the box plot method belongs to the prior art and will not be repeated here.

[0068] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0069] Example 5

[0070] When the amplitude value in the time window is abnormal, the phase obtained by secondary phase identification is used to determine whether a microseismic event has occurred. For time windows with abnormal amplitude values, the corresponding original data (seismic trace data without correction and superposition) are used. The internal parameters of the PhaseNet neural network we selected are trained from unprocessed seismic data; in microseismic data processing, signal processing such as bandpass filtering and gain is usually required to increase its signal-to-noise ratio, but these processes will change the original characteristics of the microseismic signal and reduce the recognition power of the neural network, so the original data without signal processing is still used. Input validity judgment neural network (the neural network has been trained in advance and has the ability to distinguish between noise and valid signals) If the output probability value is less than the preset value, the data is discarded.

[0071] When a micro-earthquake is identified, a spatial search method is used to scan and locate the stratum where the micro-earthquake occurs. The spatial search algorithm belongs to the prior art and will not be described in detail here.

[0072] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0073] Example 6

[0074] When the amplitude value in the time window is normal, the seismic phase obtained by the initial seismic phase identification is used to determine whether a micro-earthquake has occurred. When a micro-earthquake is identified, a spatial search method is used to scan and locate the stratum where the micro-earthquake has occurred. The spatial search algorithm belongs to the prior art and will not be described in detail here.

[0075] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0076] Example 7

[0077] This embodiment provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps of the neural network-based microseismic identification method described in the above embodiment.

[0078] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0079] Example 8

[0080] This embodiment combines the advantages of the long and short time window method (STA / LTA) and the intelligent neural network, which can improve the event recognition rate, greatly reduce the workload of processing personnel, reduce labor costs, and achieve cost reduction and efficiency improvement of microseismic observation.

[0081] Specifically include:

[0082] In the first step, after dynamic correction and horizontal stacking of the microseismic profile, the STA / LTA method is used to identify the seismic phases in the stacked traces, and the arrival time of the strong energy stacked traces that meet the threshold is stored ( Figure 2 );

[0083]

[0084] Among them, M and N are the number of samples in the long and short time windows respectively, and X i , Y i , Z i They are the amplitude values ​​of each sample point of the three components of the seismic station in the long and short time windows respectively. The three components are usually orthogonal.

[0085] R is the amplitude ratio of the short / long time window. If the value is large (usually determined by setting a threshold), that is, there is a sudden change in amplitude, it is considered that there is an earthquake phase in the short time window.

[0086] The stacking is performed to enhance the signal-to-noise ratio, and the STA / LTA method is used for phase identification.

[0087] The second step is to open a 3s time window before and after the arrival time, calculate the amplitude value within the time window for each seismic trace corresponding to each station, and use the box plot method to determine the abnormal amplitude value;

[0088]

[0089] AMP is the amplitude energy value in each time window (3 seconds before and after the identification point).

[0090] The third step is to remove the original data (seismic trace data without correction and superposition) corresponding to the time window with abnormal amplitude value. The internal parameters of the PhaseNet neural network we selected are trained by unprocessed seismic data; in the processing of microseismic data, it is usually necessary to perform signal processing such as bandpass filtering and gain to increase its signal-to-noise ratio, but these processes will change the original characteristics of the microseismic signal and reduce the recognition ability of the neural network, so the original data without signal processing is still used. Input validity judgment neural network (the neural network has been trained in advance and has the ability to distinguish between noise and valid signals) If the output probability value is less than the preset value, the data of this trace is discarded.

[0091] In the case of "no abnormal amplitude", it is possible that microseismic events exist in all seismic traces, but their amplitude values ​​are close.

[0092] The fourth step is to re-stack the seismic data in the time window (the time window in step 2), identify the STA / LTA phases, and use the spatial search method for scanning and positioning.

[0093] This embodiment also provides a computer-readable storage medium, which stores at least one computer-executable program. When the at least one program is executed by the computer, the computer executes the steps in the neural network-based microseismic identification method described in this embodiment.

[0094] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0095] In the description of the present invention, unless otherwise specified, the terms "upper", "lower", "left", "right", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they cannot be understood as limitations on the present invention.

[0096] The above technical scheme is only a specific implementation method of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the principles disclosed in the present invention, and it is not limited to the technical scheme described in the above specific embodiments of the present invention. Therefore, the above description is only preferred and does not have a restrictive meaning.

Claims

1. A microseismic identification method based on neural network, characterized in that: The following steps are involved: The original seismic data are used for initial seismic phase identification; A time window is opened for the arrival time of the first identified seismic phase on the seismic trace data, and the amplitude value within the time window is calculated; When the amplitude value in the time window is abnormal, the corresponding original seismic trace data is input into the validity judgment neural network to identify whether the original seismic trace data is noise data or valid data; If it is noise data, the original seismic trace data is removed from the original seismic data to obtain noise-reduced original data, and the noise-reduced original data is used for secondary seismic phase identification; If it is valid data, the original seismic trace data is retained in the original seismic data, and the original seismic data is used for secondary seismic phase identification.

2. The microseismic identification method based on neural network according to claim 1, characterized in that: The same method is used for primary and secondary phase identification: the short / long time window method is used to identify the phases in the superposition trace.

3. The microseismic identification method based on neural network according to claim 2 is characterized in that: The stacked trace is obtained after dynamic correction and horizontal stacking.

4. The microseismic identification method based on neural network according to claim 2 is characterized in that: The seismic phase is identified based on the short / long time window amplitude R. When the short / long time window amplitude ratio is greater than the threshold, it is considered that there is a seismic phase in the short time window.

5. The microseismic identification method based on neural network according to claim 4 is characterized in that: The calculation formula of short / long time window amplitude R is as follows: Where M and N are the number of samples in the long and short time windows respectively, and X i , Y i , Z i They are the amplitude values ​​of each sample point of the three components of the seismic station in the long and short time windows respectively.

6. The microseismic identification method based on neural network according to claim 1, characterized in that: The time window is opened for the arrival time as follows: with the arrival time as the center, open Δt seconds before and after.

7. The microseismic identification method based on neural network according to claim 1, characterized in that: The box plot method is used to determine whether the amplitude value within the time window is abnormal.

8. The microseismic identification method based on neural network according to claim 1, characterized in that: When there is no abnormality in the amplitude value within the time window, the seismic phase obtained by the initial seismic phase identification is used to determine whether a microearthquake has occurred.

9. The microseismic identification method based on neural network according to claim 1, characterized in that: When the amplitude value in the time window is abnormal, the seismic phase obtained by secondary seismic phase identification is used to determine whether a microearthquake has occurred.

10. The microseismic identification method based on neural network according to claim 8 or 9, characterized in that: When a micro-earthquake is identified, a spatial search method is used to scan and locate the stratum where the micro-earthquake occurred.

11. A computer-readable storage medium storing at least one computer-executable program, wherein when the at least one program is executed by the computer, the computer executes the steps of the neural network-based microseismic identification method according to any one of claims 1 to 10.

Images