Microseismic event positioning method and device based on neural network model

Abstract

The invention provides a microseism event positioning method and device based on a neural network model, and the method comprises the steps: obtaining a microseism event through a microseism velocity model of a target microseism region; determining the arrival time of a simulated seismic wave of each detector when a microseismic event assumed at a simulated seismic source position in the target microseismic area reaches each detector in a ground microseismic observation system of the target microseismic area; training an initial neural network model according to the arrival time of the plurality of simulated seismic waves and the position of the simulated seismic source to obtain a first neural network model; and under the condition that each detector outputs the first real seismic wave arrival time of the real microseism event, determining a first seismic source position of the real microseism event according to the plurality of first real seismic wave arrival time and the first neural network model. Therefore, the neural network model is trained through the simulated seismic source position and the simulated seismic wave arrival time of the target microseismic area, and accurate positioning of the seismic source position is realized based on the trained neural network model.

Description

Technical Field

[0001] This application relates to the technical field of microseismic monitoring, and particularly to a microseismic event location method and device based on a neural network model. Background Art

[0002] In microseismic monitoring, accurately obtaining the source location is the basis for any subsequent analysis work. Currently, the way to obtain the source location is as follows: Based on a large amount of labeled data or existing seismic catalogs, and using a convolutional neural network (CNN) or other machine learning algorithms to train an initial network model, and then obtaining the source location through the trained network model. However, in the microseismic monitoring scenario, it is often impossible to obtain a sufficient microseismic catalog before the monitoring starts. Summary of the Invention

[0003] This application proposes a microseismic event location method and device based on a neural network model.

[0004] An embodiment of this application on one hand proposes a microseismic event location method based on a neural network model. The method includes: obtaining a microseismic velocity model and a ground microseismic observation system of a target microseismic area, where the ground microseismic observation system includes a plurality of geophones; performing grid processing on the underground space model of the target microseismic area to obtain simulated source locations; according to the microseismic velocity model, determining the simulated arrival times of seismic waves of each geophone when a hypothetical microseismic event at the simulated source location reaches each geophone; training an initial neural network model according to the plurality of simulated arrival times of seismic waves and the simulated source locations to obtain a first neural network model; in the case where it is determined that each geophone outputs a first true arrival time of a real microseismic event to be processed, determining a first source location of the real microseismic event according to the plurality of first true arrival times of seismic waves and the first neural network model.

[0005] In another embodiment of the present application, a microseismic event positioning device based on a neural network model is proposed. The device includes: a first acquisition module, configured to acquire a microseismic velocity model and a ground microseismic observation system of a target microseismic area, where the ground microseismic observation system includes a plurality of geophones; a processing module, configured to perform grid processing on an underground space model of the target microseismic area to obtain a simulated source position; a first determination module, configured to determine, according to the microseismic velocity model, the simulated arrival times of seismic waves of each geophone when a hypothetical microseismic event at the simulated source position reaches each geophone; a model training module, configured to train an initial neural network model according to the plurality of simulated arrival times of seismic waves and the simulated source position to obtain a first neural network model; a second determination module, configured to, when it is determined that each geophone outputs a first true arrival time of a real microseismic event to be processed, determine a first source position of the real microseismic event according to the plurality of first true arrival times of seismic waves and the first neural network model.

[0006] In another embodiment of the present application, an electronic device is proposed, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the microseismic event positioning method based on the neural network model in the embodiment of the present application is implemented.

[0007] In another embodiment of the present application, a computer-readable storage medium is proposed, on which a computer program is stored. When the computer program is executed by a processor, the microseismic event positioning method based on the neural network model in the embodiment of the present application is implemented.

[0008] In another embodiment of the present application, a computer program product is proposed, including a computer program. When the computer program is executed by a processor, the microseismic event positioning method based on the neural network model in the embodiment of the present application is implemented.

[0009] The technical solutions provided in the embodiments of the present application may include the following beneficial effects:

[0010] Based on the microseismic velocity model of the target microseismic region, when determining the arrival times of the simulated seismic waves at each geophone in the surface microseismic observation system of the target microseismic region for the hypothetical microseismic events at the simulated source positions in the target microseismic region, and training the initial neural network model according to the multiple simulated seismic wave arrival times and the simulated source positions to obtain the first neural network model; when it is determined that each geophone outputs the first true seismic wave arrival time of the real microseismic event to be processed, determine the first source position of the real microseismic event according to the multiple first true seismic wave arrival times and the first neural network model. Thus, through the simulated source positions and simulated seismic wave arrival times in the target microseismic region, the neural network model is trained, so that the trained first neural network model can accurately determine the source position of the real microseismic event based on the real seismic wave arrival times of the real microseismic events output by each geophone, realizing the accurate positioning of the source position of the microseismic event. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings are used to better understand the solution of the present application and do not constitute a limitation to the present application. Among them:

[0012] Figure 1 is a schematic flowchart of a microseismic event positioning method based on a neural network model according to an embodiment of the present application;

[0013] Figure 2 is a schematic flowchart of a microseismic event positioning method based on a neural network model according to another embodiment of the present application;

[0014] Figure 3 is a schematic flowchart of a microseismic event positioning method based on a neural network model according to another embodiment of the present application;

[0015] Figure 4 is a schematic flowchart of a microseismic event positioning method based on a neural network model according to another embodiment of the present application;

[0016] Figure 5 is a schematic structural diagram of a microseismic event positioning device based on a neural network model according to an embodiment of the present application;

[0017] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the drawings, in which the same or similar reference numerals indicate the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application and should not be construed as a limitation to the present application.

[0019] The microseismic event location method, device, electronic device, and storage medium based on a neural network model according to an embodiment of the present application will be described below with reference to the accompanying drawings.

[0020] Figure 1 It is a schematic flowchart of a microseismic event location method based on a neural network model according to an embodiment of the present application. It should be noted that the microseismic event location method based on a neural network model provided in this embodiment is executed by a microseismic event location device based on a neural network model. The microseismic event location device based on a neural network model in this embodiment can be implemented in software and / or hardware. Among them, the microseismic event location device based on a neural network model in this example can be an electronic device, or can be configured in an electronic device.

[0021] Among them, the electronic device in this example embodiment can include a terminal device, a server, etc. Among them, the terminal device can be a PC (Personal Computer), a mobile device, a tablet computer, etc. This embodiment does not make specific limitations on this.

[0022] As Figure 1 shown, the microseismic event location method based on a neural network model may include:

[0023] Step 101: Obtain a microseismic velocity model and a surface microseismic observation system of a target microseismic area, where the surface microseismic observation system includes a plurality of geophones.

[0024] Among them, the target microseismic area can be any microseismic area, and this embodiment does not make specific limitations on the target microseismic area.

[0025] In this embodiment, a microseismic velocity model of the target microseismic area can be established according to existing well logging data, active source seismic data, or acoustic logging results of the target microseismic area.

[0026] Among them, the geophones in this embodiment can be three-component geophones.

[0027] It should be noted that the surface microseismic observation system of the target microseismic area is established in advance.

[0028] Step 102: Perform grid processing on the underground space model of the target microseismic area to obtain simulated source positions.

[0029] Among them, the underground space model of the target microseismic area is established based on the real underground space data of the target microseismic area.

[0030] In this embodiment, the underground space model of the target microseismic region can be meshed to obtain grid points, and the grid points can be used as the simulated seismic source positions.

[0031] It should be noted that the simulated seismic source position in this embodiment can also be referred to as the synthetic seismic source position.

[0032] It should be noted that there can be multiple simulated seismic source positions in this embodiment.

[0033] Step 103: According to the microseismic velocity model, determine the simulated seismic wave arrival times of each geophone when the assumed microseismic events at the simulated seismic source positions reach each geophone.

[0034] In this embodiment, based on the Fast Sweeping Method, according to the microseismic velocity model, determine the simulated seismic wave arrival times of each geophone when the assumed microseismic events at the simulated seismic source positions reach each geophone.

[0035] It should be noted that the simulated seismic wave arrival time in this embodiment can also be referred to as the synthetic seismic wave arrival time.

[0036] The simulated seismic wave arrival time in this embodiment can include the simulated P-wave arrival time and / or the simulated S-wave arrival time. That is to say, in some embodiments, the above simulated seismic wave band can include the simulated P-wave arrival time, or can include the simulated S-wave arrival time, or can include the simulated P-wave arrival time and the simulated S-wave arrival time.

[0037] Step 104: Train the initial neural network model according to multiple simulated seismic wave arrival times and simulated seismic source positions to obtain the first neural network model.

[0038] In this embodiment, in order to eliminate the influence of the unknown starting time and improve the accuracy of the trained first neural network model, a possible implementation of training the initial neural network model according to multiple simulated seismic wave arrival times and simulated seismic source positions to obtain the first neural network model can be: perform an averaging process on multiple simulated seismic wave arrival times to obtain the first average arrival time; determine the first arrival time deviation of the geophone according to the simulated seismic wave arrival time of the geophone and the first average arrival time; input the first arrival time deviation of each geophone into the initial neural network model to obtain the first predicted seismic source position; train the initial neural network model according to the first predicted seismic source position and the simulated seismic source position to obtain the first neural network model.

[0039] In this embodiment, the simulated seismic wave arrival time can be subtracted from the first average arrival time to obtain the first arrival time deviation of the geophone.

[0040] For example, there are N geophones, where N is an integer greater than 1. For the i-th geophone, assume that the arrival time of the analog seismic wave recorded by the i-th geophone is represented by t i Correspondingly, the first arrival time deviation Δt i of the i-th geophone is obtained, and the formula is as follows:

[0041]

[0042] In an embodiment of the present application, according to the first predicted source position and the simulated source position, an initial neural network model is trained. A possible implementation of obtaining the first neural network model is as follows: According to the loss function of the initial neural network model, the first predicted source position, and the simulated source position, the loss function value is determined; according to the loss function value, the initial neural network model is trained to obtain the first neural network model.

[0043] Among them, the initial neural network model in this embodiment can be a Convolutional Neural Network (CNN) model or a model of other network types, and this embodiment does not specifically limit the initial neural network model.

[0044] In this embodiment, the first predicted source position and the simulated source position can be input into the loss function of the initial neural network model to obtain the loss function value of the loss function.

[0045] In some embodiments, the loss function is preset according to actual needs.

[0046] As an example, the loss function can be the L2-norm mean square error J(θ). Among them, the formula of J(θ) is as follows:

[0047]

[0048] Among them, θ is all trainable parameters ({w i}, b j ) etc. in the initial neural network model; L represents the total number of simulated source positions; is the coordinate vector of the l-th predicted source position output by the initial neural network model; is the coordinate vector of the l-th simulated source position; ||·|| 2 represents the Euclidean distance (2-norm).

[0049] In this embodiment, the model parameters of the initial neural network model can be adjusted according to the loss function value until the loss function value meets the preset conditions or reaches the maximum number of iteration rounds, and then the training stops. The neural network model corresponding to the stop of training is used as the trained first neural network model.

[0050] Among them, the preset condition is the condition for the end of model training. The preset condition can be configured accordingly according to actual needs. For example, the loss function value satisfying the preset condition can be that the loss function value is less than a preset value, or the change of the loss function value approaches stability, that is, the difference between the loss function values corresponding to two adjacent or multiple trainings is less than a set value, that is, the loss function value basically no longer changes.

[0051] Among them, it can be understood that in the process of training the initial neural network model by simulating the seismic source position and the first arrival time deviation, the model parameters of the initial neural network model are continuously adjusted according to the loss function value of each training. For example, the model parameters of the initial neural network model can be adjusted in the direction of the decreasing loss function value. When the loss function value satisfies the preset condition, the trained first neural network model is obtained.

[0052] In some embodiments, in order to accelerate the training convergence of the initial neural network model, a possible implementation of inputting the first arrival time deviation of each geophone into the initial neural network model to obtain the first predicted seismic source position is as follows: normalizing the first arrival time deviation to obtain the processed first arrival time deviation; inputting the processed first arrival time deviation of each geophone into the initial neural network model to obtain the first predicted seismic source position. Thus, the convergence speed of the model can be accelerated, and the efficiency of obtaining the trained first neural network model can be improved.

[0053] As an example, a possible implementation of normalizing the first arrival time deviation to obtain the processed first arrival time deviation can be: determining the maximum value and the minimum value among the multiple first arrival time deviations, and normalizing each first arrival time deviation according to the maximum value and the minimum value to obtain the corresponding processed first arrival time deviation.

[0054] Continuing with the above example, the processed first arrival time deviation τ i of the i-th geophone is obtained as follows:

[0055]

[0056] Among them, Δt min in the formula represents the minimum value among the multiple first arrival time deviations, and Δt max represents the maximum value among the multiple first arrival time deviations.

[0057] Among them, it should be noted that τ i in the above formula represents the normalized first arrival time deviation of the i-th geophone, and the value range of this τ i is between [0, 1].

[0058] Among them, it should be noted that in the case where the above initial neural network model is a CNN model, the CNN model includes a feed-forward neural network. Among them, an example expression of the feed-forward neural network can be:

[0059] (1) Network structure

[0060] The dimension of the input layer corresponds to the number N of geophones, and the input is {τ 1 , τ 2 , τ 3 , τ 4 ,..., τ N};

[0061] The dimension of the output layer is 3, that is, (x, y, z);

[0062] Several hidden layers can be set in the middle, and the number and layers of neurons in the hidden layer are usually selected through experiments or parameter tuning.

[0063] (2) Neuron output form

[0064] In a feed-forward neural network (Feed-Forward Neural Network, abbreviated as ANN), the output of the j-th neuron can be expressed by the following formula:

[0065]

[0066] Among them, ζ j is the output of the j-th neuron; w i represents the weight corresponding to the input x i ; b j is the bias term; f(·) represents the activation function.

[0067] Among them, it can be understood that in the training process of the initial neural network model, that is, on the basis of a large number of known input (synthetic or real data)-output (real source position) pairs, adjust {w i} and b j values to make the neural network model achieve the best fit.

[0068] Step 105, when it is determined that each geophone outputs the first true arrival time of the real microseismic event to be processed, determine the first source position of the real microseismic event according to the multiple first true arrival times and the first neural network model.

[0069] In some embodiments, when it is determined that each geophone outputs the first true arrival time of the true microseismic event to be processed, a possible implementation of determining the first source location of the true microseismic event based on multiple first true arrival times and the first neural network model is as follows: perform an averaging process on the multiple first true arrival times to obtain a second average arrival time; determine the second arrival time deviation of the geophone based on the first true arrival time and the second average arrival time of the geophone; input the second arrival time deviation of each geophone into the first neural network model to obtain the first source location of the true microseismic event. Thus, through the trained first neural network model and the second arrival time deviation, the first source location of the true microseismic event is accurately determined.

[0070] In some embodiments, when obtaining the second arrival time deviation, the second arrival time deviation can be normalized to obtain the processed second arrival time deviation, and the processed second arrival time deviation is input into the first neural network model to obtain the first source location of the true microseismic event. Thus, through the first neural network model, the first source location of the true microseismic event is accurately determined.

[0071] In some embodiments, the maximum value and the minimum value of the multiple second arrival time deviations can be determined, and for each second arrival time deviation, the second arrival time deviation is normalized based on the maximum value and the minimum value to obtain the processed second arrival time deviation. Among them, the processed second arrival time deviation can also be referred to as the second arrival time deviation after normalization processing.

[0072] The microseismic event positioning method based on a neural network model provided by the embodiments of the present application determines the simulated arrival times of the microseismic waves at each geophone in the ground microseismic observation system of the target microseismic area when the assumed microseismic event at the simulated source location in the target microseismic area reaches through the microseismic velocity model of the target microseismic area, and trains the initial neural network model based on the multiple simulated arrival times and the simulated source location to obtain the first neural network model; when it is determined that each geophone outputs the first true arrival time of the true microseismic event to be processed, the first source location of the true microseismic event is determined based on the multiple first true arrival times and the first neural network model. Thus, through the simulated source location and the simulated arrival times of the microseismic waves in the target microseismic area, the neural network model is trained, so that the trained first neural network model can accurately determine the source location of the true microseismic event based on the true arrival times of the true microseismic events output by each geophone, realizing the accurate positioning of the source location of the microseismic event.

[0073] Figure 2It is a schematic flowchart of a microseismic event location method based on a neural network model according to another embodiment of the present application. It should be noted that this embodiment is a further refinement or optimization of the foregoing embodiment.

[0074] Step 201: Obtain a microseismic velocity model of a target microseismic area and a surface microseismic observation system, where the surface microseismic observation system includes a plurality of geophones.

[0075] Step 202: Perform grid processing on the underground space model of the target microseismic area to obtain simulated source positions.

[0076] Step 203: According to the microseismic velocity model, determine the simulated arrival times of the seismic waves at each geophone when a hypothetical microseismic event at the simulated source position reaches each geophone.

[0077] Step 204: Train an initial neural network model based on the multiple simulated arrival times and the simulated source positions to obtain a first neural network model.

[0078] It should be noted that for the specific descriptions of steps 201 to 204, reference can be made to the relevant descriptions in other embodiments, which will not be elaborated here.

[0079] Step 205: Obtain verification data, where the verification data includes: the true arrival times of the seismic waves output by each geophone for historical microseismic events, and the true source positions of the historical microseismic events.

[0080] It should be noted that the historical microseismic events in this embodiment are historical microseismic events that occurred in the target microseismic area.

[0081] It should be noted that the number of historical microseismic events can be multiple. Correspondingly, the verification data can include multiple true source positions.

[0082] Step 206: Determine a second predicted source position based on the true arrival times of the seismic waves and the first neural network model.

[0083] In some embodiments, an average processing can be performed on the multiple true arrival times of the seismic waves to obtain a third average arrival time; for each geophone, according to the true arrival time of the seismic wave of this geophone and the third average arrival time, determine the third arrival time deviation of this geophone, and input the third arrival time deviations of each geophone into the first neural network model to obtain the second predicted source position of the historical microseismic event.

[0084] Step 207: Determine the root mean square error between the true source position and the second predicted source position.

[0085] In this embodiment, the calculation formula for the root mean square error is as follows:

[0086]

[0087] where Y i ' represents the i-th second predicted source position output by the first neural network model; Yi represents the i-th true source position, where n represents the total number of true source positions included in the validation data.

[0088] Step 208: Determine that the root mean square error is within a preset error range.

[0089] The preset error range is an error range preset according to actual requirements.

[0090] It should be noted that when the root mean square error is within the preset error range, it indicates that the first neural network model meets the accuracy requirements. At this time, the first neural network model can be put online, that is, the true arrival time of the seismic wave of the real microseismic event to be processed monitored on site can be processed through the first neural network model to obtain the source position of the real microseismic event to be processed.

[0091] It should be noted that in the case where the root mean square error is not within the preset error range, the first neural network model can be trained until the root mean square error determined based on the neural network model is within the preset error range.

[0092] Step 209: When it is determined that each geophone outputs the first true arrival time of the real microseismic event to be processed, determine the first source position of the real microseismic event according to the multiple first true arrival times and the first neural network model.

[0093] In this embodiment, after training the initial neural network model through the simulated source position and the simulated arrival time of the seismic wave to obtain the first neural network model, the first neural network model is verified through the validation data, and when it is determined that the root mean square error is within the preset error range, the first true arrival time of the real microseismic event to be processed is processed through the first neural network model to accurately obtain the first source position of the real microseismic event to be processed.

[0094] Based on any of the above embodiments, in order to accurately determine the first occurrence time of the real microseismic event, after determining the first occurrence time, the first occurrence time of the real microseismic event can also be determined according to the multiple first true arrival times and the first source position.

[0095] For example, the coordinates of the first source position are (x 0 , y 0 , z0 ) Based on multiple first true arrival times of seismic waves and the first source location, the least squares method can be used to fit and obtain the first origin time t of the true microseismic event 0 , where the formula for obtaining the first origin time t 0 is as follows:

[0096]

[0097] where it should be noted that t i represents the first true arrival time of the seismic wave output by the i-th geophone; is the theoretical travel time calculated under the known velocity model and the known coordinates (x 0 , y 0 , z 0 ).

[0098] where it should be noted that by minimizing F(t 0 ), the optimal first origin time t 0 can be obtained.

[0099] Figure 3 is a schematic flow chart of a microseismic event location method based on a neural network model according to another embodiment of the present application. It should be noted that this embodiment is a further refinement or optimization of the foregoing embodiment.

[0100] As Figure 3 shown, the microseismic event location method based on the neural network model may include:

[0101] Step 301: Obtain the microseismic velocity model of the target microseismic area and the ground microseismic observation system, where the ground microseismic observation system includes multiple geophones.

[0102] Step 302: Perform grid processing on the underground space model of the target microseismic area to obtain the simulated source locations.

[0103] Step 303: According to the microseismic velocity model, determine the simulated arrival times of the simulated seismic waves at each geophone when the assumed microseismic event at the simulated source location reaches each geophone.

[0104] Step 304: Train the initial neural network model based on multiple simulated arrival times of seismic waves and the simulated source locations to obtain the first neural network model.

[0105] It should be noted that for the specific descriptions of steps 301 to 304, reference can be made to the relevant descriptions in other embodiments, which will not be elaborated here.

[0106] Step 305, when it is determined that the first detector among multiple detectors can output the arrival time of the second true seismic wave, while the second detector cannot output the arrival time of the true seismic wave, obtain a second neural network model, where the model parameters of the second neural network model are the same as those of the first neural network model.

[0107] It should be noted that in this embodiment, the detector among multiple detectors that can output the arrival time of the true seismic wave is referred to as the first detector.

[0108] The detector among multiple detectors that cannot output the arrival time of the true seismic wave is referred to as the second detector.

[0109] In some embodiments, to determine that the first neural network model meets the accuracy requirements, correspondingly, before step 305, verification data can also be obtained, where the verification data includes: the arrival time of the true seismic wave output by each detector for historical microseismic events, and the true source location of the historical microseismic events; determine the second predicted source location according to the arrival time of the true seismic wave and the first neural network model; determine the root mean square error between the true source location and the second predicted source location; determine that the root mean square error is within a preset error range.

[0110] Step 306, fine-tune the second neural network model according to the simulated seismic wave arrival time and the simulated source location corresponding to the first detector to obtain a third neural network model.

[0111] In some embodiments, the simulated seismic wave arrival time corresponding to the first detector can be normalized to obtain the processed simulated seismic wave arrival time, and the processed simulated seismic wave arrival times of each first detector are input into the second neural network model to obtain the third predicted source location. The second neural network model is trained according to the third predicted source location and the simulated source location to obtain a third neural network model.

[0112] Step 307, determine the second source location of the true microseismic event according to the arrival time of the second true seismic wave and the third neural network model.

[0113] In this embodiment, the arrival time of the second true seismic wave can be normalized to obtain the processed arrival time of the second true seismic wave, and the processed arrival times of the second true seismic wave of each first detector are input into the third neural network model to obtain the second source location of the true microseismic event.

[0114] Based on the above description, it can be seen that in this embodiment, when it is determined that the first geophone among multiple geophones can output the arrival time of the second true seismic wave, while the second geophone cannot output the arrival time of the true seismic wave, the second neural network model is fine-tuned based on the simulated seismic wave arrival time and the simulated source position corresponding to the first geophone to obtain the third neural network model, and the third neural network model is used to process the arrival time of the second true seismic wave output by the first geophone to obtain the second source position. Thus, rapid response in the case of missing pick-up geophones is achieved, and the positioning accuracy is maintained to the greatest extent.

[0115] Based on the above embodiments, in order to accurately determine the second earthquake occurrence time of the true microseismic event, after determining the second earthquake occurrence time, the second earthquake occurrence time of the true microseismic event can also be determined according to the arrival time of the second true seismic wave and the second source position.

[0116] It should be noted that the process of determining the second earthquake occurrence time is similar to the process of determining the first earthquake occurrence time described above, and will not be elaborated here.

[0117] To clearly understand the present application, the method disclosed in this embodiment will be described exemplarily below in combination with Figure 4 the following.

[0118] Figure 4 FIG. is a schematic flowchart of a microseismic event positioning method based on a neural network model according to another embodiment of the present application.

[0119] As Figure 4 shown, the microseismic event positioning method based on a neural network model may include:

[0120] Step 401: Perform grid processing on the underground space model of the target microseismic region to obtain the simulated source position, and according to the microseismic velocity model of the target microseismic region, determine the simulated seismic wave arrival times of each geophone when the assumed microseismic event at the simulated source position reaches each geophone in the ground microseismic observation system of the target microseismic region.

[0121] In this embodiment, the microseismic velocity model of the target microseismic region can be established according to the existing logging data, active source seismic data or acoustic logging results of the target microseismic region.

[0122] The geophone in this embodiment can be a three-component geophone.

[0123] It should be noted that the ground microseismic observation system of the target microseismic region is established in advance.

[0124] In this embodiment, the underground space model of the target microseismic area can be meshed to obtain grid points, and the grid points can be used as the simulated seismic source positions.

[0125] It should be noted that the simulated seismic source position in this embodiment can also be referred to as the synthetic seismic source position.

[0126] Step 402: Train the initial neural network model according to the simulated seismic source position and the arrival time of the simulated seismic wave to obtain the trained first neural network model.

[0127] It should be noted that for the specific implementation manner of step 402, reference can be made to the relevant descriptions in other embodiments, which will not be elaborated here.

[0128] Step 403: In actual monitoring, if the first detector among multiple detectors is available while the second detector cannot pick up or picks up wrongly, the second neural network model can be fine-tuned according to the arrival time of the simulated seismic wave output by the first detector and the simulated seismic source position to obtain the third neural network model, where the model parameters of the second neural network model and the first neural network model are the same.

[0129] It should be noted that in some embodiments, an alternative solution for step 403 can be: In actual monitoring, if the first detector among multiple detectors is available while the second detector cannot pick up or picks up wrongly, the second neural network model can be fine-tuned by combining the historical arrival time of the seismic wave output by the first detector and the true seismic source position determined based on the historical arrival time of the seismic wave.

[0130] In this embodiment, when training the second neural network model, the hidden layer in the second neural network model can be frozen or partially frozen, and the input layer of the second neural network model can be adjusted so that the input layer can only receive the arrival time of the seismic wave of the first detector.

[0131] Step 404: Process the verification data through the third neural network model. When it is determined that the root mean square error of the verification data is within the preset error range, process the arrival time of the seismic wave output by the first detector for the microseismic event to be processed through the third neural network model to obtain the seismic source position of the microseismic event to be processed.

[0132] Step 405: Determine the occurrence time of the microseismic event to be processed according to the seismic source position of the microseismic event to be processed.

[0133] In this embodiment, rapid response in the case of missing pick-up detectors is achieved, the positioning accuracy is maintained to the greatest extent, and combined with the determined seismic source position, the occurrence time of the microseismic event to be processed is accurately obtained, realizing the precise spatio-temporal positioning of the microseismic event.

[0134] Corresponding to the microseismic event location method based on a neural network model provided in the above several embodiments, an embodiment of the present application further provides a microseismic event location device based on a neural network model. Since the microseismic event location device based on a neural network model provided in the embodiments of the present application corresponds to the microseismic event location method based on a neural network model provided in the above several embodiments, the implementation manners of the microseismic event location method based on a neural network model are also applicable to the microseismic event location device based on a neural network model in this embodiment and will not be described in detail in this embodiment.

[0135] Figure 5 It is a schematic structural diagram of a microseismic event location device based on a neural network model according to an embodiment of the present application.

[0136] As Figure 5 shown, the microseismic event location device 500 based on a neural network model includes: a first acquisition module 501, a processing module 502, a first determination module 503, a model training module 504, and a second determination module 505, where:

[0137] The first acquisition module 501 is configured to acquire a microseismic velocity model and a ground microseismic observation system of a target microseismic area, where the ground microseismic observation system includes a plurality of geophones.

[0138] The processing module 502 is configured to perform grid processing on the underground space model of the target microseismic area to obtain a simulated source location.

[0139] The first determination module 503 is configured to determine the simulated arrival times of seismic waves of each geophone when a hypothetical microseismic event at the simulated source location reaches each geophone according to the microseismic velocity model.

[0140] The model training module 504 is configured to train an initial neural network model according to a plurality of simulated arrival times of seismic waves and the simulated source location to obtain a first neural network model.

[0141] The second determination module 505 is configured to determine a first source location of the real microseismic event according to a plurality of first real arrival times of seismic waves and the first neural network model when it is determined that each geophone outputs a first real arrival time of a real microseismic event to be processed.

[0142] In one embodiment of the present application, the model training module 504 is specifically configured to: perform an averaging process on multiple simulated seismic wave arrival times to obtain a first average arrival time; determine a first arrival time deviation of the geophone according to the simulated seismic wave arrival time and the first average arrival time of the geophone; input the first arrival time deviations of each geophone into an initial neural network model to obtain a first predicted source location; and train the initial neural network model according to the first predicted source location and the simulated source location to obtain a first neural network model. In one embodiment of the present application, training the initial neural network model according to the first predicted source location and the simulated source location to obtain a first neural network model includes: determining a loss function value according to the loss function of the initial neural network model, the first predicted source location, and the simulated source location; and training the initial neural network model according to the loss function value to obtain a first neural network model.

[0143] In one embodiment of the present application, inputting the first arrival time deviations of each geophone into an initial neural network model to obtain a first predicted source location includes: performing a normalization process on the first arrival time deviations to obtain processed first arrival time deviations; and inputting the processed first arrival time deviations of each geophone into the initial neural network model to obtain a first predicted source location.

[0144] In one embodiment of the present application, the second determination module 505 is specifically configured to: perform an averaging process on multiple first true seismic wave arrival times to obtain a second average arrival time; determine a second arrival time deviation of the geophone according to the first true seismic wave arrival time and the second average arrival time of the geophone; and input the second arrival time deviations of each geophone into the first neural network model to obtain a first source location of a true microseismic event.

[0145] In one embodiment of the present application, the apparatus may further include: a third determination module, specifically configured to determine a first occurrence time of a true microseismic event according to multiple first true seismic wave arrival times and the first source location.

[0146] In one embodiment of the present application, the apparatus may further include:

[0147] a second acquisition module, configured to acquire a second neural network model when it is determined that a first geophone among multiple geophones can output a second true seismic wave arrival time while a second geophone cannot output a true seismic wave arrival time, where the second neural network model has the same model parameters as the first neural network model;

[0148] a fine-tuning module, configured to fine-tune the second neural network model according to the simulated seismic wave arrival time and the simulated source location corresponding to the first geophone to obtain a third neural network model;

[0149] A fourth determination module, configured to determine a second source location of a real micro-seismic event according to the second arrival time of the real seismic wave and a third neural network model.

[0150] In an embodiment of the present application, the apparatus may further include: a fifth determination module, configured to determine a second occurrence time of a real micro-seismic event according to the second arrival time of the real seismic wave and the second source location.

[0151] In an embodiment of the present application, the apparatus may further include: a verification module, specifically configured to: obtain verification data, where the verification data includes: the arrival time of the real seismic wave output by each geophone for a historical micro-seismic event, and the real source location of the historical micro-seismic event; determine a second predicted source location according to the arrival time of the real seismic wave and a first neural network model; determine the root mean square error between the real source location and the second predicted source location; and determine that the root mean square error is within a preset error range.

[0152] The micro-seismic event positioning apparatus based on a neural network model provided in the embodiments of the present application determines the arrival time of the simulated seismic wave of each geophone when a hypothetical micro-seismic event at the simulated source location in the target micro-seismic region reaches each geophone in the ground micro-seismic observation system of the target micro-seismic region through the micro-seismic velocity model of the target micro-seismic region, and trains an initial neural network model according to a plurality of simulated seismic wave arrival times and simulated source locations to obtain a first neural network model; in the case where it is determined that each geophone outputs the first arrival time of the real micro-seismic event to be processed, determine the first source location of the real micro-seismic event according to a plurality of first arrival times of the real seismic wave and the first neural network model. Thus, the neural network model is trained through the simulated source location and simulated seismic wave arrival time in the target micro-seismic region, so that the trained first neural network model can accurately determine the source location of the real micro-seismic event based on the arrival time of the real seismic wave of the real micro-seismic event output by each geophone, realizing the accurate positioning of the source location of the micro-seismic event.

[0153] According to an embodiment of the present application, the present application further provides an electronic device.

[0154] Figure 6 It is a structural block diagram of an electronic device according to an embodiment of the present application.

[0155] As Figure 6 shown, the electronic device 600 includes: a memory 610, a processor 620, and computer instructions stored on the memory 610 and executable on the processor 620.

[0156] When the processor 620 executes the instructions, it implements the micro-seismic event positioning method based on a neural network model provided in the above embodiment.

[0157] Furthermore, the electronic device 600 further includes:

[0158] A communication interface 630 for communication between the memory 610 and the processor 620.

[0159] A memory 610 for storing computer instructions that can be run on the processor 620.

[0160] The memory 610 may include high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory.

[0161] A processor 620 for implementing the microseismic event location method based on the neural network model in the above embodiments when executing a program.

[0162] If the memory 610, the processor 620, and the communication interface 630 are implemented independently, the communication interface 630, the memory 610, and the processor 620 can be interconnected through a bus and complete communication with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0163] Optionally, in a specific implementation, if the memory 610, the processor 620, and the communication interface 630 are integrated on a chip, the memory 610, the processor 620, and the communication interface 630 can complete communication with each other through an internal interface.

[0164] The processor 620 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0165] Another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the microseismic event location method based on a neural network model disclosed in the embodiments of the present application is implemented.

[0166] Another embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the microseismic event location method based on a neural network model in the embodiments of the present application is implemented.

[0167] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0168] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present application, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically and clearly defined.

[0169] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for locating microseismic events based on a neural network model, characterized in that: The method comprises: Acquire a microseismic velocity model and a ground microseismic observation system of a target microseismic area, wherein the ground microseismic observation system includes a plurality of geophones; Performing grid processing on the underground space model of the target microseismic area to obtain a simulated earthquake source position; Determining, according to the microseismic velocity model, when the hypothetical microseismic event at the simulated source position reaches each geophone, the arrival time of the simulated seismic wave of each geophone; Training the initial neural network model according to the arrival times of the plurality of simulated seismic waves and the simulated earthquake source locations to obtain a first neural network model; When it is determined that each detector outputs the first real seismic wave arrival time of the real microseismic event to be processed, the first hypocenter position of the real microseismic event is determined based on multiple first real seismic wave arrival times and the first neural network model.

2. The method according to claim 1, characterized in that The initial neural network model is trained according to the arrival times of the plurality of simulated seismic waves and the simulated earthquake source positions to obtain a first neural network model, comprising: Averaging a plurality of the simulated seismic wave arrival times to obtain a first average arrival time; Determining a first arrival time deviation of the geophone according to the simulated seismic wave arrival time of the geophone and the first average arrival time; Inputting the first arrival time deviation of each of the geophones into an initial neural network model to obtain a first predicted earthquake source position; The initial neural network model is trained according to the first predicted earthquake source position and the simulated earthquake source position to obtain a first neural network model.

3. The method according to claim 2, characterized in that The initial neural network model is trained according to the first predicted earthquake source position and the simulated earthquake source position to obtain a first neural network model, comprising: Determining a loss function value according to the loss function of the initial neural network model, the first predicted earthquake source location and the simulated earthquake source location; The initial neural network model is trained according to the loss function value to obtain a first neural network model.

4. The method according to claim 2, characterized in that The step of inputting the first arrival time deviation of each of the geophones into an initial neural network model to obtain a first predicted earthquake source position comprises: normalizing the first arrival time deviation to obtain a processed first arrival time deviation; The processed first arrival time deviations of the respective detectors are input into the initial neural network model to obtain the first predicted earthquake source position.

5. The method according to claim 2, characterized in that The method of determining the first source position of the real microseismic event according to the first real seismic wave arrival times and the first neural network model when it is determined that each geophone outputs the first real seismic wave arrival time of the real microseismic event to be processed comprises: averaging a plurality of the first real seismic wave arrival times to obtain a second average arrival time; Determine a second arrival time deviation of the geophone according to a first true seismic wave arrival time of the geophone and a second average arrival time; The second arrival time deviation of each of the detectors is input into the first neural network model to obtain the first hypocenter position of the real microseismic event.

6. The method according to claim 5, characterized in that The method further comprises: The first occurrence moment of the real microseismic event is determined based on the arrival times of the first multiple real seismic waves and the first earthquake source positions.

7. The method according to claim 2, characterized in that The method further comprises: When it is determined that the first geophone among the plurality of geophones can output the second real seismic wave arrival time, and the second geophone cannot output the real seismic wave arrival time, obtaining a second neural network model, wherein the model parameters of the second neural network model are the same as those of the first neural network model; According to the arrival time of the simulated seismic wave corresponding to the first geophone and the simulated earthquake source position, the second neural network model is fine-tuned to obtain a third neural network model; According to the arrival time of the second real seismic wave and the third neural network model, the second hypocenter position of the real microseismic event is determined.

8. The method according to claim 7, characterized in that The method further comprises: The second occurrence time of the real microseismic event is determined according to the arrival time of the second real seismic wave and the second earthquake source position.

9. The method according to any one of claims 1 to 8, characterized in that In the case where it is determined that each geophone outputs the first real seismic wave arrival time of the real microseismic event to be processed, before determining the first source position of the real microseismic event according to the plurality of the first real seismic wave arrival times and the first neural network model, the method further comprises: Acquiring verification data, wherein the verification data includes: the arrival time of the real seismic waves output by each geophone for the historical microseismic events, and the real source position of the historical microseismic events; Determining a second predicted earthquake source position according to the real earthquake wave arrival time and the first neural network model; determining a root mean square error between the true earthquake source location and the second predicted earthquake source location; It is determined that the root mean square error is within a preset error range.

10. A microseismic event location device based on a neural network model, characterized in that: The device comprises: A first acquisition module, used to acquire a microseismic velocity model and a ground microseismic observation system of a target microseismic area, wherein the ground microseismic observation system includes a plurality of geophones; A processing module, used for performing grid processing on the underground space model of the target microseismic area to obtain a simulated earthquake source position; A first determination module is used to determine, based on the microseismic velocity model, when the assumed microseismic event at the simulated source position reaches each geophone, and when the simulated seismic wave of each geophone arrives at the respective geophone; A model training module, used to train the initial neural network model according to the arrival times of the plurality of simulated seismic waves and the simulated earthquake source locations to obtain a first neural network model; The second determination module is used to determine the first hypocenter position of the real microseismic event according to multiple first real seismic wave arrival times and the first neural network model when it is determined that each detector outputs the first real seismic wave arrival time of the real microseismic event to be processed.

11. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 9 when executing the computer program.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

13. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 9 when being executed by a processor.

Images