A prediction model training method, a seismic prediction method, a device, and an electronic device

Abstract

The present disclosure provides a prediction model training method, a seismic prediction method, a device and an electronic device, and relates to the technical field of earthquake monitoring and earthquake prediction. The training method comprises: synthesizing a plurality of aftershock data in historical seismic waveform data of a target area with preset noise data respectively to obtain a seismic precursor data set; intercepting a plurality of continuous waveform data containing small earthquake group events from the historical seismic waveform data to obtain a random event data set; determining a noise data set by generating a white noise sequence or selecting a plurality of noise segment waveform data with a magnitude below a preset seismic magnitude from the historical seismic waveform data; and training a neural network by applying the seismic precursor data set, the random event data set and the noise data set to obtain a seismic prediction model. The present disclosure improves the identification accuracy of seismic precursor anomalies.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of earthquake monitoring and earthquake prediction, and more particularly, to a prediction model training method, an earthquake prediction method, a device and an electronic equipment. BACKGROUND

[0002] Earthquake is a medium structure mutation occurring in a local area of the earth interior, which is caused by the instantaneous rupture of underground rock structure. In the aspect of earthquake prevention and disaster reduction, short-term earthquake prediction is particularly important. The basis for realizing earthquake prediction is to understand the physical process of earthquake preparation and the changes in the physical properties and mechanical state of the crustal rock in this process. Due to the complexity of the earthquake preparation mechanism, the limitations of observation means, and the multi-solution nature of geophysical problems, the problem of earthquake prediction cannot be well solved. Precursor phenomena provide certain feasibility for earthquake prediction, but due to the complexity of precursor phenomena, many abnormal phenomena are not necessarily caused by earthquakes. Common earthquake precursor phenomena include: seismic activity anomaly, seismic wave velocity change, crustal deformation, abnormal change of underground water, change of radon content or other chemical components in underground water, change of ground stress, change of ground electricity, change of ground magnetism, gravity anomaly, animal anomaly, ground sound, ground light, ground temperature anomaly, etc. Earthquake prediction has always been a worldwide problem, especially short-term earthquake prediction. With the development of science and technology, people's observation data means are more and more, and the amount of data is also more and more, but there is still a lack of an effective and stable earthquake prediction method. The classic earthquakes that have successfully made short-term earthquake prediction are the Haicheng earthquake in 1975 and the Liu'yan earthquake in 1999. Seismic waveform records contain rich information about the earthquake source, underground structure, and underground structure change. According to the research on the successfully predicted Haicheng earthquake and Liu'yan earthquake, in the 2 to 3 days before the two major earthquakes, there appeared an abnormal earthquake pattern that small earthquakes first increased slowly and then suddenly decreased. One of the precursors of an earthquake is that small earthquakes concentrate in a short time, the number of small earthquakes first increases slowly to a certain peak value, and the large earthquake also distributes near the peak value, and then the seismic activity suddenly decreases, the number of small earthquakes decreases, the magnitude decreases, and then a major earthquake occurs. This precursor pattern may also vary in different geological backgrounds. Some precursor small earthquakes have large magnitudes, such as the Haicheng earthquake and the Liu'yan earthquake, which have very obvious foreshocks; some precursor small earthquakes have small magnitudes and are partially or completely submerged in noise, making them difficult to identify.

[0003] Deep learning technology is not only applied in knowledge-based systems, but also widely popularized in many fields such as natural language understanding, non-monotonic reasoning, machine vision, pattern recognition, and data mining. In addition, deep learning has been widely applied in geophysics, and in some cases even exceeds traditional geophysical methods, such as the deep learning method for picking up the first arrival and the traditional STA / LTA method. Therefore, there is an urgent need for a method for applying deep learning to earthquake prediction. SUMMARY

[0004] Therefore, the present disclosure provides a prediction model training method, a seismic prediction method, a device and an electronic device.

[0005] One aspect of the present disclosure provides a seismic prediction model training method, comprising:

[0006] Synthesizing each of a plurality of aftershock data in historical seismic waveform data of a target area with preset noise data to obtain a plurality of seismic precursor data samples of a fixed data length, and taking the plurality of seismic precursor data samples as a seismic precursor data set;

[0007] Cutting a plurality of continuous waveform data containing small earthquake group events from the historical seismic waveform data to obtain a plurality of random event samples of the fixed data length, and taking the plurality of random event samples as a random event data set;

[0008] Determining a plurality of noise data samples of the fixed data length by generating a white noise sequence or selecting a plurality of noise segment waveform data with a magnitude below a preset seismic magnitude from the historical seismic waveform data, and taking the plurality of noise data samples as a noise data set;

[0009] Training a neural network by applying the seismic precursor data set, the random event data set and the noise data set to obtain a seismic prediction model.

[0010] According to an embodiment of the present disclosure, the training method further comprises:

[0011] Performing a dimensionality reduction operation on each sample in the seismic precursor data set, the random event data set and the noise data set to obtain a dimensionality reduced data set;

[0012] The application of the seismic precursor data set, the random event data set and the noise data set to train the neural network to obtain the seismic prediction model comprises:

[0013] Training the neural network by applying the dimensionality reduced data set to obtain the seismic prediction model.

[0014] According to an embodiment of the present disclosure, the magnitude difference between the maximum magnitude and the minimum magnitude of each seismic precursor data sample in the seismic precursor data set is greater than a preset magnitude.

[0015] Another aspect of the present disclosure provides a deep learning-based seismic prediction method, comprising:

[0016] inputting the seismic waveform data of a target region at a current moment into a seismic prediction model to obtain a prediction result; the prediction result is a probability that the seismic waveform data at the current moment belongs to a seismic precursor, wherein the seismic prediction model is obtained by using the seismic prediction model training method.

[0017] According to an embodiment of the present disclosure, the prediction method further comprises:

[0018] performing a preprocessing operation on the seismic waveform data at the current moment to obtain target seismic waveform data;

[0019] The inputting the seismic waveform data of a target region at a current moment into a seismic prediction model to obtain a prediction result comprises:

[0020] inputting the target seismic waveform data into the seismic prediction model to obtain the prediction result.

[0021] According to an embodiment of the present disclosure, the performing a preprocessing operation on the seismic waveform data at the current moment to obtain target seismic waveform data specifically comprises:

[0022] performing an instrument response removal operation on the seismic waveform data at the current moment;

[0023] performing noise removal on the seismic waveform data after the instrument response removal;

[0024] performing dimensionality reduction on the seismic waveform data after the noise removal to obtain target seismic waveform data.

[0025] Another aspect of the present disclosure provides a seismic prediction model training device applying the seismic prediction model training method, comprising:

[0026] a seismic precursor data set determination module configured to synthesize a plurality of aftershock data in historical seismic waveform data of a target region with preset noise data respectively to obtain a plurality of seismic precursor data samples of a fixed data length, and take the plurality of seismic precursor data samples as a seismic precursor data set;

[0027] a random event data set determination module configured to intercept a plurality of continuous waveform data containing small earthquake group events from the historical seismic waveform data to obtain a plurality of random event samples of the fixed data length, and take the plurality of random event samples as a random event data set;

[0028] a noise data set determination module configured to determine a plurality of noise data samples of the fixed data length by generating a white noise sequence or selecting a plurality of noise segment waveform data with a magnitude below a preset earthquake magnitude from the historical seismic waveform data, and take the plurality of noise data samples as a noise data set;

[0029] a training module configured to train a neural network by applying the earthquake precursor data set, the random event data set, and the noise data set to obtain an earthquake prediction model.

[0030] Another aspect of the present disclosure provides an earthquake prediction device based on deep learning, which applies the earthquake prediction method based on deep learning described above, and includes:

[0031] a prediction module configured to input seismic waveform data of a target region at a current time into the earthquake prediction model to obtain a prediction result, the prediction result being a probability that the seismic waveform data at the current time belongs to an earthquake precursor.

[0032] Another aspect of the present disclosure provides an electronic device, which includes:

[0033] one or more processors;

[0034] a memory configured to store one or more programs,

[0035] wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method described above.

[0036] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method described above.

[0037] Another aspect of the present disclosure provides a computer program product including computer-executable instructions, which, when executed, implement the method described above.

[0038] According to embodiments of the present disclosure, by applying a plurality of aftershock data in historical seismic waveform data to be respectively synthesized with preset noise data to obtain an earthquake precursor data set, a plurality of continuous waveform data containing small earthquake group events are intercepted from the historical seismic waveform data to obtain a random event data set, a white noise sequence is generated or a plurality of noise segment waveform data with a magnitude below a preset earthquake magnitude are selected from the historical seismic waveform data to obtain a noise data set, a training set for neural network model training is constructed, and an earthquake prediction model obtained by training a neural network model by applying the training set can identify an earthquake anomaly pattern in background noise, so that at least part of the technical problems that some precursory small earthquakes have a small magnitude and are partially or completely submerged in noise and difficult to identify are overcome, and a technical effect of improving the identification accuracy of earthquake precursor anomalies is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0039] The above and other objects, features and advantages of the present disclosure will become more apparent from the following description when taken in conjunction with the accompanying drawings, in which:

[0040] Figure 1 An exemplary system architecture to which the prediction model training method and / or the deep learning-based earthquake prediction method according to embodiments of the present disclosure can be applied is schematically shown;

[0041] Figure 2 A flowchart of the earthquake prediction model training method according to embodiments of the present disclosure is schematically shown;

[0042] Figure 3 A flowchart of the deep learning-based earthquake prediction method according to embodiments of the present disclosure is schematically shown;

[0043] Figure 4 A flowchart of the pre-processing method according to embodiments of the present disclosure is schematically shown;

[0044] Figure 5 A block diagram of the earthquake prediction model training apparatus according to embodiments of the present disclosure is schematically shown;

[0045] Figure 6 A block diagram of the deep learning-based earthquake prediction apparatus according to embodiments of the present disclosure is schematically shown;

[0046] Figure 7A A schematic diagram of the east-west vibration component in three-component seismogram data according to embodiments of the present disclosure is schematically shown;

[0047] Figure 7B A schematic diagram of the north-south vibration component in three-component seismogram data according to embodiments of the present disclosure is schematically shown;

[0048] Figure 7C A schematic diagram of the vertical vibration component in three-component seismogram data according to embodiments of the present disclosure is schematically shown;

[0049] Figure 8A An earthquake magnitude-time graph in which an earthquake precursor is apparent according to embodiments of the present disclosure is schematically shown;

[0050] Figure 8B An earthquake magnitude-time graph in which an earthquake precursor is overwhelmed by noise according to embodiments of the present disclosure is schematically shown;

[0051] Figure 8C An earthquake magnitude-time graph in which there is no earthquake precursor or the precursor is of another pattern according to embodiments of the present disclosure is schematically shown;

[0052] Figure 9A schematic block diagram of a seismic prediction system based on deep learning to identify large earthquake anomaly patterns in background noise according to an embodiment of the present disclosure is shown schematically;

[0053] Figure 10A A distribution diagram of the number of earthquakes versus time according to an embodiment of the present disclosure is shown schematically;

[0054] Figure 10B A diagram showing that the distribution of the number of earthquakes versus magnitude according to an embodiment of the present disclosure satisfies the Gutenberg-Richter law is shown schematically;

[0055] Figure 10C A diagram of a small event signal according to an embodiment of the present disclosure is shown schematically;

[0056] Figure 10D A diagram of noise for synthesizing precursors according to an embodiment of the present disclosure is shown schematically;

[0057] Figure 10E A diagram of a synthesized precursor according to an embodiment of the present disclosure is shown schematically;

[0058] Figure 11A A diagram of eigenvalue curves of seismic precursor data samples after principal component analysis according to an embodiment of the present disclosure is shown schematically;

[0059] Figure 11B A diagram of compressed data samples after dimensionality reduction of noise data samples according to an embodiment of the present disclosure is shown schematically;

[0060] Figure 11C A diagram of compressed data samples after dimensionality reduction of random event data samples according to an embodiment of the present disclosure is shown schematically;

[0061] Figure 11D A diagram of compressed data samples after dimensionality reduction of seismic precursor data samples according to an embodiment of the present disclosure is shown schematically;

[0062] Figure 12 A diagram of the structure of a convolutional neural network according to an embodiment of the present disclosure is shown schematically;

[0063] Figure 13A A diagram of a validation set loss function curve in a model training process according to an embodiment of the present disclosure is shown schematically;

[0064] Figure 13B A diagram of a confusion matrix of a test set according to an embodiment of the present disclosure is shown schematically;

[0065] Figure 14 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is shown schematically. DETAILED DESCRIPTION

[0066] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It is to be understood, however, that the description is merely exemplary and is intended to provide a thorough understanding of the present disclosure. In the following detailed description of embodiments of the present disclosure, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it would be apparent to those skilled in the art that the present disclosure can be practiced without these specific details. In other instances, well-known structures and functions have not been described in detail in order to avoid obscuring aspects of the present disclosure.

[0067] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. As used herein, the term "includes" and tautological expressions thereof, such as "including," "includes," "include," "contains," "containing," and so on, mean the term "comprises," unless otherwise noted.

[0068] All terms used herein, including technical and scientific terms, have the same meanings as those generally understood by those skilled in the art, unless otherwise defined herein. It should be noted that the terms used herein should be interpreted as having meanings consistent with the context of the present description, and should not be interpreted in an idealized or overly formal manner.

[0069] In the case of using expressions similar to "at least one of A, B, and C, etc.", it is generally to be interpreted as including one or more of the same unless otherwise noted.

[0070] In embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of data (for example, including but not limited to user personal information) involved are in accordance with the provisions of relevant laws and regulations, are used for legal purposes, and do not violate public order and good customs. In particular, necessary measures are taken to prevent illegal access to user personal information data, and to maintain user personal information security, network security and national security.

[0071] In embodiments of the present disclosure, the authorization or consent of the user is obtained before the user's personal information is acquired or collected.

[0072] Embodiments of the present disclosure provide a prediction model training method, a seismic prediction method, an apparatus, and an electronic device.

[0073] Figure 1An exemplary system architecture 100 to which the prediction model training method and / or the deep learning based earthquake prediction method according to embodiments of the present disclosure can be applied is schematically shown. It is to be noted that Figure 1 The shown is only an example of the system architecture to which embodiments of the present disclosure can be applied, to help those skilled in the art understand the technical content of the present disclosure, but does not mean that embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0074] As Figure 1 The system architecture 100 according to this embodiment can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104 and a server 105, as shown. The network 104 is a medium to provide communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. The network 104 can include various connection types, such as wired and / or wireless communication links, etc.

[0075] The user can use the first terminal device 101, the second terminal device 102, the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients and / or social platform software, etc. (only as examples).

[0076] The first terminal device 101, the second terminal device 102, the third terminal device 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smartphones, tablet computers, laptop computers and desktop computers, etc.

[0077] The server 105 can be a server providing various services, such as a background management server supporting the website browsed by the user using the first terminal device 101, the second terminal device 102, the third terminal device 103 (only as an example). The background management server can analyze and process the received user requests and other data, and feed back the processing results (such as web pages, information or data generated or obtained according to the user requests, etc.) to the terminal device.

[0078] It should be noted that the prediction model training method and / or the deep learning based earthquake prediction method provided by the embodiments of the present disclosure can generally be executed by the server 105. Accordingly, the prediction model training system and / or the deep learning based earthquake prediction system provided by the embodiments of the present disclosure can generally be arranged in the server 105. The prediction model training method and / or the deep learning based earthquake prediction method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the prediction model training system and / or the deep learning based earthquake prediction system provided by the embodiments of the present disclosure can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Alternatively, the prediction model training method and / or the deep learning based earthquake prediction method provided by the embodiments of the present disclosure can also be executed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or by other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the prediction model training system and / or the deep learning based earthquake prediction system provided by the embodiments of the present disclosure can also be arranged in the first terminal device 101, the second terminal device 102 or the third terminal device 103, or in other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0079] For example, the historical seismic waveform data and / or the seismic waveform data at the current moment can originally be stored in any one of the first terminal device 101, the second terminal device 102 or the third terminal device 103 (for example, the first terminal device 101, but not limited thereto), or stored on an external storage device and imported into the first terminal device 101. Then, the first terminal device 101 can execute the prediction model training method and / or the deep learning based earthquake prediction method provided by the embodiments of the present disclosure locally, or send the historical seismic waveform data and / or the seismic waveform data at the current moment to other terminal devices, servers or server clusters, and execute the prediction model training method and / or the deep learning based earthquake prediction method provided by the embodiments of the present disclosure by other terminal devices, servers or server clusters receiving the historical seismic waveform data and / or the seismic waveform data at the current moment.

[0080] It should be understood that Figure 1 The number of terminal devices, networks and servers in the above description is only illustrative. According to the needs of implementation, there can be any number of terminal devices, networks and servers.

[0081] Figure 2 A flowchart of a method for training a seismic prediction model is shown schematically according to an embodiment of the present disclosure.

[0082] As shown in FIG. 2, the training method 200 includes operations S201-S204. Figure 2

[0083] In operation S201, a plurality of aftershock data in historical seismic waveform data of a target region are respectively synthesized with preset noise data to obtain a plurality of seismic precursor data samples of a fixed data length, and the plurality of seismic precursor data samples are taken as a seismic precursor data set.

[0084] In operation S202, a plurality of continuous waveform data containing small earthquake group events are intercepted from the historical seismic waveform data to obtain a plurality of random event samples of a fixed data length, and the plurality of random event samples are taken as a random event data set.

[0085] In operation S203, a plurality of noise data samples of a fixed data length are determined by generating a white noise sequence or selecting a plurality of noise segment waveform data with a magnitude below a preset seismic magnitude from the historical seismic waveform data, and the plurality of noise data samples are taken as a noise data set.

[0086] In operation S204, the seismic precursor data set, the random event data set, and the noise data set are applied to train a neural network to obtain a seismic prediction model.

[0087] A region where seismic precursor prediction is desired can be taken as a target region, and historical seismic waveform data of the target region can be applied to construct a seismic precursor data set, a random event data set, and a noise data set. The seismic precursor data set, the random event data set, and the noise data set are applied to train a neural network, thereby obtaining a seismic prediction model for seismic precursor prediction of the target region. For example, when seismic precursor prediction is performed on another region, historical seismic waveform data of the other region can be applied to construct a seismic precursor data set, a random event data set, and a noise data set of the other region, and a neural network is trained, thereby obtaining a seismic prediction model for seismic precursor prediction of the other region.

[0088] ​In training the neural network, the training dataset containing the earthquake precursor dataset, the random event dataset and the noise dataset can be divided into a training set, a validation set and a test set according to a preset ratio, and the preset ratio can be 8:1:1, for example. Then, the test set can be input into the trained neural network, and the accuracy and / or recall rate of the trained neural network can be calculated according to the output of the trained neural network. When the accuracy and / or recall rate of the trained neural network meets the preset requirement, the trained neural network meeting the preset requirement is taken as the earthquake prediction model. For example, when the accuracy is greater than 90%, the trained neural network is considered to be able to serve as the earthquake prediction model.

[0089] Before training the neural network, the earthquake precursor dataset, the random event dataset and the noise dataset can be preprocessed, for example, each sample in the earthquake precursor dataset, the random event dataset and the noise dataset is subjected to a dimension reduction operation to obtain a dimension-reduced dataset. Then, the dimension-reduced dataset is applied to train the neural network. Since the dimension-reduced dataset has a reduced data dimension compared with the earthquake precursor dataset, the random event dataset and the noise dataset, the calculation speed of the neural network training is increased, and the features contained in the dimension-reduced dataset are more prominent, thereby improving the prediction accuracy of the trained neural network for the earthquake precursor.

[0090] In the dimension reduction operation of the earthquake precursor dataset, the random event dataset and the noise dataset, a principal component analysis method can be used, so that the dimension-reduced dataset can contain more features. The data in the earthquake precursor dataset, the random event dataset and the noise dataset can be in various data formats, such as single-station single-component, single-station three-component, multi-station single-component and multi-station three-component.

[0091] Other ways can also be used to improve the prediction accuracy of the neural network model, for example, the magnitude difference between the maximum magnitude and the minimum magnitude of each earthquake precursor data sample in the earthquake precursor dataset is greater than a preset magnitude, or the relationship between the earthquake precursor data sample and time can be described by a Chi-squared distribution. Taking the preset magnitude as an example, the preset magnitude can be selected according to the actual situation, and can be selected as 2, the magnitude range is 0 to 3, -1 to 2, and the span of 3 is divided, or the preset magnitude is set as 3. Such a setting makes the data size in the earthquake precursor data sample change obviously, which is beneficial to the neural network to better learn the characteristics of the earthquake precursor in the training process, thereby improving the prediction accuracy of the trained neural network for the earthquake precursor.

[0092] In the process of obtaining the random event data set, small earthquake group events can be selected, and data with frequent small earthquakes but no large earthquake in a period of time thereafter can be selected as the random event sample. In the process of determining the noise data set, the noise data sample in the seismic waveform data in the study area can be selected as noise or white noise, or a combination of noise and white noise. For example, the noise data sample in the seismic waveform data is selected to have a noise level of ML1 or below. The wide random event data set, noise data set, and earthquake precursor data set can generalize the ability of the earthquake prediction model to identify earthquake precursors.

[0093] The value of the fixed data length of the earthquake precursor data sample, the random event sample, and the noise data sample can be set according to actual needs, for example, the fixed data length is one day, or a few hours, or a few days.

[0094] The embodiments of the present disclosure can use the neural network to search for the probability of learning earthquake precursors after obtaining input seismic data, predict the probability of large earthquake occurrence in the next few days, and can be applied to natural earthquake prediction, and can also be used for induced earthquake prediction, etc. Combined with existing geostress, hydrological information, etc., it can have certain guiding significance for deciding whether to issue a pre-earthquake warning.

[0095] Figure 3 The flowchart of the deep learning-based earthquake prediction method according to the embodiments of the present disclosure is schematically shown.

[0096] As shown in Figure 3 The prediction method 300 includes operation S301:

[0097] In operation S301, the seismic waveform data of the target region at the current time is input into the earthquake prediction model to obtain a prediction result; the prediction result is the probability that the seismic waveform data at the current time belongs to an earthquake precursor, wherein the earthquake prediction model is trained by using the earthquake prediction model training method of the embodiments of the present disclosure.

[0098] Optionally, the current time seismic waveform data is preprocessed to obtain target seismic waveform data, so that when the current time seismic waveform data is predicted for earthquake precursors, the target seismic waveform data is input into the earthquake prediction model to obtain a prediction result.

[0099] Figure 4 The flowchart of the preprocessing method according to the embodiments of the present disclosure is schematically shown.

[0100] As shown in Figure 4 According to the embodiments of the present disclosure, the preprocessing method 210 for preprocessing the current time seismic waveform data to obtain target seismic waveform data includes operations S211-S213:

[0101] In operation S211, an instrument response removal operation is performed on the seismic waveform data at the current moment.

[0102] In operation S212, the seismic waveform data after the instrument response is removed is denoised.

[0103] In operation S213, the denoised seismic waveform data is dimensionality reduced to obtain the target seismic waveform data.

[0104] There are many preprocessing methods. One or more of the following can be chosen for the current seismic waveform data: instrument response removal, denoising, and dimensionality reduction. Alternatively, denoising can be performed first, followed by instrument response removal, and finally dimensionality reduction. Instrument response removal overcomes the bias in prediction results caused by differences in the acquisition instruments used in the current seismic waveform data and the training dataset of the seismic prediction model. Denoising overcomes the inaccuracy in prediction results caused by noise in the current seismic waveform data. Dimensionality reduction reduces the computation time of the seismic prediction model and improves the efficiency of its output.

[0105] Figure 5 A block diagram of an earthquake prediction model training apparatus according to an embodiment of the present disclosure is shown schematically.

[0106] like Figure 5 As shown, the training device 400 includes an earthquake precursor dataset determination module 401, a random event dataset determination module 402, a noise dataset determination module 403, and a training module 404.

[0107] The earthquake precursor dataset determination module 401 is used to synthesize multiple aftershock data from the historical earthquake waveform data of the target area with preset noise data to obtain multiple earthquake precursor data samples with a fixed data length, and to use the multiple earthquake precursor data samples as the earthquake precursor dataset.

[0108] The random event dataset determination module 402 is used to extract multiple continuous waveform data containing small earthquake swarm events from historical earthquake waveform data, obtain multiple random event samples with a fixed data length, and use the multiple random event samples as a random event dataset.

[0109] The noise dataset determination module 403 is used to determine multiple noise data samples of fixed data length by generating a white noise sequence or selecting multiple noise segment waveform data with magnitudes below a preset earthquake magnitude from historical earthquake waveform data, and to use the multiple noise data samples as a noise dataset.

[0110] The training module 404 is configured to train the neural network by using the earthquake precursor data set, the random event data set, and the noise data set to obtain the earthquake prediction model.

[0111] Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure, or at least part of the functions of any one or more of the modules, sub-modules, units, sub-units, can be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be split into multiple modules for implementation. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be implemented at least in part as a hardware circuit, for example, a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner of hardware or firmware by integrating or packaging the circuit, or in any one of software, hardware, and firmware, or in a proper combination of any one or more of them. Alternatively, one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be implemented at least in part as computer program modules, which can perform corresponding functions when the computer program modules are run.

[0112] For example, any one or more of the earthquake precursor data set determination module 401, the random event data set determination module 402, the noise data set determination module 403, and the training module 404 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of the modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units, and implemented in one module / unit / sub-unit. According to the embodiments of the present disclosure, at least one of the earthquake precursor data set determination module 401, the random event data set determination module 402, the noise data set determination module 403, and the training module 404 can be implemented at least in part as a hardware circuit, for example, a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner of hardware or firmware by integrating or packaging the circuit, or in any one of software, hardware, and firmware, or in a proper combination of any one or more of them. Alternatively, at least one of the earthquake precursor data set determination module 401, the random event data set determination module 402, the noise data set determination module 403, and the training module 404 can be implemented at least in part as computer program modules, which can perform corresponding functions when the computer program modules are run.

[0113] Figure 6 A block diagram of a deep learning based earthquake prediction apparatus is shown.

[0114] As shown in Figure 6 The prediction apparatus 500 includes a prediction module 501.

[0115] The prediction module 501 is configured to input the current time seismic waveform data of the target region into an earthquake prediction model to obtain a prediction result, where the prediction result is a probability that the current time seismic waveform data belongs to an earthquake precursor.

[0116] Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure, or at least part of the functions of any one or more of the modules, sub-modules, units, sub-units can be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be split into multiple modules for implementation. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of hardware or firmware that can be integrated or packaged with a circuit, or implemented in any one of software, hardware and firmware or in a proper combination of any of them. Alternatively, one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present disclosure can be at least partially implemented as computer program modules that can perform corresponding functions when executed.

[0117] For example, the prediction module 501 can be implemented in one module / unit / sub-unit, or one module / unit / sub-unit can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of the modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units, and implemented in one module / unit / sub-unit. According to the embodiments of the present disclosure, the prediction module 501 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of hardware or firmware that can be integrated or packaged with a circuit, or implemented in any one of software, hardware and firmware or in a proper combination of any of them. Alternatively, the prediction module 501 can be at least partially implemented as computer program modules that can perform corresponding functions when executed.

[0118] It should be noted that the earthquake prediction model training device part in the embodiments of the present disclosure corresponds to the earthquake prediction model training method part in the embodiments of the present disclosure, and the description of the earthquake prediction model training device part is specifically referred to the earthquake prediction model training method part. In addition, the deep learning based earthquake prediction device part in the embodiments of the present disclosure corresponds to the deep learning based earthquake prediction method part in the embodiments of the present disclosure, and the description of the deep learning based earthquake prediction device part is specifically referred to the deep learning based earthquake prediction method part, which will not be repeated here.

[0119] Figure 7A An east-west vibration component schematic diagram in three-component seismogram data according to an embodiment of the present disclosure is schematically shown; Figure 7B An east-west vibration component schematic diagram in three-component seismogram data according to an embodiment of the present disclosure is schematically shown; Figure 7C An east-west vibration component schematic diagram in three-component seismogram data according to an embodiment of the present disclosure is schematically shown; Figure 7A 、 Figure 7B and Figure 7C In the above, the horizontal axis is relative time (second), and the vertical axis is the velocity of the medium vibration at the position of the seismograph. The station position is 31.48°N, 103.60°E, and the waveform length is 2 minutes. In Figure 7A , BHE is Broadband High-Gain E-component waveform, which refers to a wideband high-gain east component waveform. Figure 7B , BHN is Broadband High-Gain N-component waveform, which refers to a wideband high-gain north component waveform. Figure 7C , BHZ is Broadband High-Gain Z-component waveform, which refers to a wideband high-gain Z component waveform.

[0120] Figure 8A An earthquake magnitude-time diagram with obvious precursors according to an embodiment of the present disclosure is schematically shown; Figure 8B An earthquake magnitude-time diagram with precursors submerged in noise according to an embodiment of the present disclosure is schematically shown; Figure 8C An earthquake magnitude-time diagram without earthquake precursors or with other patterns of precursors according to an embodiment of the present disclosure is schematically shown. There are three states of earthquake precursors: obvious earthquake precursors, earthquake precursors submerged in noise, and no earthquake precursors or other unidentified earthquake precursors. As shown in Figure 8A , the earthquake magnitude signal is obviously higher than the noise, which belongs to the case of obvious earthquake precursors. Figure 8BAs shown, some of the precursory signals of the earthquakes are weak due to different geological backgrounds, and the effective precursors are submerged in the noise, i.e., the magnitude signal is lower than the noise signal; for example Figure 8C As shown, there is no earthquake precursor or there is an unidentified earthquake precursor.

[0121] Figure 9 The schematic diagram of the earthquake prediction system based on deep learning for identifying abnormal patterns of large earthquakes in background noise according to the earthquake prediction model training device 400 and the deep learning-based earthquake prediction device 500 is shown schematically according to the embodiments of the present disclosure; the system 900 includes a training part and a working part. The training part includes: a seismic map data set 910 for constructing an earthquake precursor data set, a random event data set, and a noise data set; a data compression device 930 for preprocessing the seismic map data set 910, reducing the data volume while retaining the main information of the data; a neural network model 940 for training the reduced dimension data set output by the data compression device 930, and the trained network model can be distributed in a remote server or distributed storage. The working part includes: an input device 920 which can be a data interface, for example, can be connected to the output device or communication interface of a single or multiple seismic monitoring stations to receive the seismic map records of the generated seismic events from the single or multiple seismic monitoring stations. The input device 120 can also include a user interface such as a keyboard, a touch screen, etc., and a user or operator can input system settings, predetermined conditions, etc. through the input device 120 to control and manage the operation of the system. Any appropriate structure or form can be used to implement the input device 120. The input device 120 is used to input the seismic waveform data of the station obtained in real time from the seismic station network data stream, which can contain earthquake precursors, random events, or noise, etc.; the seismic waveform data is input into the data compression device 930 to obtain target seismic waveform data, and the target seismic waveform data is input into the trained neural network model 940 to obtain the probability that the target seismic waveform data belongs to the earthquake precursor.

[0122] The system 900 can also include a decision device (not shown), which can comprehensively evaluate the possibility of a destructive large earthquake occurring in the near future by comparing other seismic monitoring parameters such as stress changes, velocity changes, etc.

[0123] Figure 10A The distribution of the number of earthquakes and time according to the embodiments of the present disclosure is shown schematically; Figure 10B The distribution of the number of earthquakes and magnitude according to the embodiments of the present disclosure satisfies the Gutenberg-Richter law, as shown schematically; Figure 10C The small event signal according to the embodiments of the present disclosure is shown schematically; Figure 10D The noise for synthesizing precursors according to the embodiments of the present disclosure is shown schematically;Figure 10E A schematic diagram of a synthetic precursor is shown according to an embodiment of the present disclosure. Figure 10A 、 Figure 10B 、 Figure 10C 、 Figure 10D and Figure 10E A process of synthesizing a seismic precursor dataset is shown, and the constructed seismic precursor dataset, random event dataset and noise dataset meet the Gutenberg-Richter law of magnitude distribution and the Chi-squared law of earthquake number distribution. The length of the synthetic data can be 1-15 days, and one data sample can contain three-component data of multiple stations. Among them Figure 10A The number of earthquakes meets the Chi-squared distribution of time, Figure 10B The number of earthquakes and the magnitude meet the Gutenberg-Richter law distribution. The description of the unidentified seismic precursor is not limited to the Chi-squared distribution. Given the time range, magnitude range, noise level and other parameters of the synthesized seismic precursor data, a seismic precursor dataset can be obtained. The noise as shown in Figure 10D and the screened seismic waveform data as shown in Figure 10C are superimposed according to the synthetic seismic precursor catalog, and a seismic precursor signal as shown in Figure 10E is obtained. By transforming the parameters and data values of the seismic precursor signal, a seismic precursor dataset is obtained.

[0124] Figure 11A A characteristic value curve diagram of a seismic precursor data sample processed by principal component analysis is shown according to an embodiment of the present disclosure. Figure 11B A compressed data sample diagram of a noise data sample after dimensionality reduction is shown according to an embodiment of the present disclosure. Figure 11C A compressed data sample diagram of a random event data sample after dimensionality reduction is shown according to an embodiment of the present disclosure. Figure 11D A compressed data sample diagram of a seismic precursor data sample after dimensionality reduction is shown according to an embodiment of the present disclosure. As shown in Figure 11A 、 Figure 11B 、 Figure 11C and Figure 11D are schematic diagrams of different preprocessing operations on the seismic precursor dataset, random event dataset and noise dataset.

[0125] Figure 12 A structure diagram of a convolutional neural network is shown according to an embodiment of the present disclosure. As shown in Figure 12 , the convolutional neural network has three convolutional layers, four fully connected layers and pooling layers, etc., wherein the convolution kernel size is selected as 5x5.

[0126] Figure 13AA validation set loss function curve diagram in a model training process is schematically shown according to an embodiment of the present disclosure. Figure 13B A confusion matrix diagram of a test set is schematically shown according to an embodiment of the present disclosure. As Figure 13A and Figure 13B shown, the neural network constructed by training with the training set data is continuously adjusted in parameters according to the test results of the validation set, the network structure and other parameters of the model are optimized, and the neural network that has been trained is tested with the data of the test set. It can be seen from the confusion matrix that the recall rate and the accuracy of each index reach a certain level. The feasibility of the method for synthetic data is verified.

[0127] Figure 14 A block diagram of an electronic device suitable for implementing the method described above is schematically shown according to an embodiment of the present disclosure. Figure 14 The electronic device shown is only an example and should not bring any limitation to the functions and use range of the embodiments of the present disclosure.

[0128] As Figure 14 shown, the electronic device 1400 according to an embodiment of the present disclosure includes a processor 1401, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 1402 or loaded from a storage portion 1408 into a random access memory (RAM) 1403. The processor 1401 may, for example, include a general-purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chipset, and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1401 can also include an on-board memory for cache use. The processor 1401 can include a single processing unit or multiple processing units for performing different actions of the method processes according to embodiments of the present disclosure.

[0129] In the RAM 1403, various programs and data required for the operation of the electronic device 1400 are stored. The processor 1401, the ROM 1402, and the RAM 1403 are connected to each other through a bus 1404. The processor 1401 performs various operations of the method processes according to embodiments of the present disclosure by executing programs in the ROM 1402 and / or the RAM 1403. It should be noted that the programs can also be stored in one or more memories other than the ROM 1402 and the RAM 1403. The processor 1401 can also perform various operations of the method processes according to embodiments of the present disclosure by executing programs stored in the one or more memories.

[0130] According to an embodiment of the present disclosure, the electronic device 1400 can further include an input / output (I / O) interface 1405 that is also connected to the bus 1404. The electronic device 1400 can further include one or more of the following components connected to the input / output (I / O) interface 1405: an input part 1406 including, for example, a keyboard and a mouse; an output part 1407 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage part 1408 including, for example, a hard disk; and a communication part 1409 including, for example, a LAN card, a modem, and the like. The communication part 1409 performs communication processing via a network such as the Internet. A drive 1410 is also connected to the input / output (I / O) interface 1405 as necessary. A removable medium 1411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 1410 as necessary, so that a computer program read therefrom is installed into the storage part 1408 as necessary.

[0131] According to an embodiment of the present disclosure, the method flow according to the embodiment of the present disclosure can be implemented as a computer software program. For example, the embodiment of the present disclosure includes a computer program product including a computer program carried on a computer-readable storage medium, the computer program containing program codes for executing the method shown in the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network by the communication part 1409, and / or installed from the removable medium 1411. When the computer program is executed by the processor 1401, the above-described functions defined in the system implementing the embodiment of the present disclosure are performed. According to an embodiment of the present disclosure, the system, device, apparatus, module, unit, and the like described above can be implemented by computer program modules.

[0132] The present disclosure also provides a computer-readable storage medium, which can be included in the device / apparatus / system described in the above embodiments, or can exist separately without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiment of the present disclosure.

[0133] According to an embodiment of the disclosure, the computer readable storage medium can be a nonvolatile computer readable storage medium. For example, it can include, but is not limited to, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this disclosure, a computer readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in connection with an instruction execution system, apparatus, or device.

[0134] For example, according to an embodiment of the disclosure, the computer readable storage medium can include one or more memories other than the ROM 1402 and / or the RAM 1403 and / or the ROM 1402 and the RAM 1403 described above.

[0135] Embodiments of the disclosure also include a computer program product, which includes a computer program containing program codes for executing the method provided by the embodiments of the disclosure, and when the computer program product is run on an electronic device, the program codes are used to make the electronic device implement the seismic prediction model training method and / or the seismic prediction method based on deep learning provided by the embodiments of the disclosure.

[0136] When the computer program is executed by the processor 1401, the above-mentioned functions defined in the system / apparatus of the embodiments of the disclosure are executed. According to an embodiment of the disclosure, the system, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0137] In one embodiment, the computer program can rely on tangible storage media such as optical storage media, magnetic storage media, etc. In another embodiment, the computer program can also be transmitted, distributed, and downloaded in the form of a signal on a network medium, and be downloaded and installed through the communication part 1409, and / or installed from the detachable medium 1411. The program codes contained in the computer program can be transmitted by any appropriate network medium, including but not limited to wireless, wired, etc., or any suitable combination of the foregoing.

[0138] According to embodiments of the present disclosure, program code of the computer program for performing the methods provided by the embodiments of the present disclosure can be written in any combination of one or more programming languages, including a high-level procedural and / or object-oriented programming language, and / or an assembly / machine language. Programming languages include, but are not limited to, Java, C++, python, "C" language, or the like. The program code can execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, such as through the Internet using an Internet Service Provider.

[0139] The computer program code can also be loaded onto a computer or other programmable information processing system to cause a series of operational steps to be performed on the computer or other programmable information processing system to produce the operations described.

[0140] The computer program code can also be loaded onto a computer or other programmable information processing system to cause a series of operational steps to be performed on the computer or other programmable information processing system to produce the operations described. The above-described embodiments of the present disclosure are merely descriptive and are not intended to be limiting. Although the above embodiments have been described in detail, those skilled in the art will understand that various modifications can be made without departing from the scope of the present disclosure. Such modifications are also intended to fall within the scope of the present disclosure.

Claims

1. A method for training a seismic prediction model, characterized in that, The training method comprises: synthesizing a plurality of aftershock data in historical seismic waveform data of a target region with preset noise data respectively to obtain a plurality of seismic precursor data samples of a fixed data length, and taking the plurality of seismic precursor data samples as a seismic precursor data set; cutting a plurality of continuous waveform data containing small earthquake group events from the historical seismic waveform data to obtain a plurality of random event samples of the fixed data length, and taking the plurality of random event samples as a random event data set; determining a plurality of noise data samples of the fixed data length by generating a white noise sequence or selecting a plurality of noise segment waveform data with a magnitude below a preset seismic magnitude from the historical seismic waveform data, and taking the plurality of noise data samples as a noise data set; training a neural network by applying the seismic precursor data set, the random event data set and the noise data set to obtain a seismic prediction model. 2.The method of claim 1, wherein, The training method further comprises: performing dimension reduction operation on each sample in the seismic precursor data set, the random event data set and the noise data set to obtain a dimension-reduced data set; wherein the training of the neural network by applying the seismic precursor data set, the random event data set and the noise data set to obtain a seismic prediction model comprises: training the neural network by applying the dimension-reduced data set to obtain the seismic prediction model. 3.The method of claim 1, wherein, The magnitude difference between the maximum magnitude and the minimum magnitude of each seismic precursor data sample in the seismic precursor data set is greater than a preset magnitude.

4. A deep learning-based earthquake prediction method, characterized by, The seismic prediction method comprises: inputting seismic waveform data at a current time of a target region into a seismic prediction model to obtain a prediction result; the prediction result is the probability that the seismic waveform data at the current time belongs to a seismic precursor, wherein the seismic prediction model is trained by using the seismic prediction model training method according to any one of claims 1-3.

5. The deep learning based earthquake prediction method of claim 4, wherein, The prediction method further comprises: performing preprocessing operation on the seismic waveform data at the current time to obtain target seismic waveform data; wherein the inputting of the seismic waveform data at the current time of the target region into the seismic prediction model to obtain the prediction result comprises: inputting the target seismic waveform data into the seismic prediction model to obtain the prediction result.

6. The deep learning based earthquake prediction method of claim 5, wherein, The preprocessing operation on the seismic waveform data at the current time to obtain target seismic waveform data specifically comprises: performing de-instrument response operation on the seismic waveform data at the current time; performing de-noising on the seismic waveform data after de-instrument response; performing dimension reduction on the seismic waveform data after de-noising to obtain target seismic waveform data. 7.A device for training a seismic prediction model, characterized in that, The seismic prediction model training method according to any one of claims 1-3, the training device comprises: a seismic precursor data set determination module configured to synthesize a plurality of aftershock data in historical seismic waveform data of a target region with preset noise data respectively to obtain a plurality of seismic precursor data samples of a fixed data length, and take the plurality of seismic precursor data samples as a seismic precursor data set; a random event data set determination module configured to extract a plurality of continuous waveform data containing small earthquake group events from the historical seismic waveform data, obtain a plurality of random event samples of the fixed data length, and take the plurality of random event samples as a random event data set; a noise data set determination module configured to generate a white noise sequence or select a plurality of noise segment waveform data with a magnitude below a preset earthquake magnitude from the historical seismic waveform data, determine a plurality of noise data samples of the fixed data length, and take the plurality of noise data samples as a noise data set; a training module configured to train a neural network by using the earthquake precursor data set, the random event data set, and the noise data set, and obtain an earthquake prediction model. 8.A device for earthquake prediction based on deep learning, characterized in that, The earthquake prediction device comprises: a prediction module configured to input seismic waveform data of a target region at a current time into the earthquake prediction model, and obtain a prediction result; the prediction result is a probability that the seismic waveform data at the current time belongs to an earthquake precursor. 9.An electronic device, comprising: one or more processors; memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-3 and / or 4-6. 10.A computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the method of any one of claims 1-3 and / or 4-6.

Images