Earthquake monitoring and early warning method and system based on earthquake precursor

Abstract

The invention discloses an earthquake monitoring and early warning method and system based on an earthquake precursor, and relates to the technical field of earthquake forecasting, and the method comprises the steps: collecting the core data and auxiliary state data of the earthquake precursor, marking a sample label, and forming a training data set; establishing a core parameter screening model, and training the core parameter screening model by using the training data set to obtain core parameters; constructing a typical working condition library, calculating an Euclidean distance between a real-time data point and a historical working condition sample point, selecting first K neighbor points, and returning a working condition category with the highest frequency as a current operation working condition; corresponding anomaly prediction models are established according to the different operation conditions, and anomaly types under the different current operation conditions are predicted according to the anomaly prediction models; an EWI index is calculated through the core parameters, and when the EWI index is larger than an index threshold value, earthquake early warning is triggered; otherwise, continuously monitoring. According to the invention, real-time early warning of the earthquake precursor is realized under the condition that the equipment is abnormal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of earthquake prediction, more particularly to an earthquake monitoring and early warning method and system based on earthquake precursors. BACKGROUND

[0002] As a sudden, destructive and difficult to predict natural disaster, the core of earthquake prevention and disaster reduction lies in earthquake monitoring, and geophysical observation is the key support for earthquake prediction. With the development of technology, geophysical observation has formed a multi-means, large-scale, intelligent networked monitoring system, but the current daily operation and monitoring work of the earthquake center station still faces many pain points, and the existing monitoring system has obvious short boards in function, compatibility, localization adaptation, etc. Promoting the research and development of a new type of monitoring system has become an industry demand.

[0003] The current earthquake center station relies on manual timing operation for the management of precursor equipment. The staff needs to check, download data and confirm the connection state of all precursor equipment in the jurisdiction at least 4 times a day. When the equipment cannot be connected, the maintenance personnel are fed back for processing. Even so, due to fixed inspection time intervals, 2-3 hours of equipment interruption may not be discovered; and the lack of remote monitoring tools for operation and maintenance, most of the time, the video guidance of the station caretaker needs to be contacted, and the unmanned station needs to be checked by the operation and maintenance personnel. Some faults that can be recovered by simply restarting may also cause several hours of data interruption, seriously affecting the continuity and integrity of the earthquake monitoring data.

[0004] From the perspective of monitoring equipment coverage, current geophysical observation has involved multiple types of equipment such as crustal deformation, electromagnetic, underground fluid, and meteorological correlation, and some stations have configured auxiliary equipment such as intelligent power supply and video monitoring. It is necessary to uniformly manage multi-dimensional equipment data and status, and the traditional dispersed inspection mode has been difficult to adapt to the large-scale monitoring demand.

[0005] Therefore, how to conduct real-time early warning of earthquake precursors while taking into account equipment abnormal conditions is a problem that needs to be solved by those skilled in the art. SUMMARY

[0006] Therefore, the present application provides an earthquake monitoring and early warning method and system based on earthquake precursors, which can conduct real-time early warning of earthquake precursors while taking into account equipment abnormal conditions.

[0007] In order to achieve the above purpose, the present application adopts the following technical solutions: An earthquake monitoring and early warning method based on earthquake precursors, comprising: Step 1: data acquisition and sample construction, acquiring earthquake precursor core data and auxiliary state data, labeling sample labels, and forming a training data set; Step two: a core parameter screening model is established based on an XgBoost method, the core parameter screening model is trained based on the training data set, and the core parameter is obtained; Step three: condition recognition is performed based on KNN, a typical condition library is constructed, Euclidean distances between real-time data points and historical condition sample points are calculated, K nearest neighbor points are selected, and a condition category with the highest frequency is returned as a current running condition; Step four: abnormality prediction based on an improved BP neural network, an abnormality prediction model corresponding to different running conditions is established, and the abnormality prediction model is used to predict an abnormality type under different current running conditions; Step five: early warning judgment based on an EWI index, the EWI index is calculated based on the core parameter, and when the EWI index is greater than an index threshold, an earthquake early warning is triggered; otherwise, continuous monitoring is performed.

[0008] Preferably, the precursor core data includes crustal deformation data, electromagnetic data, underground fluid data, and meteorological correlation data; and the auxiliary state data includes equipment operation data, environmental interference data, and network data.

[0009] Preferably, the core parameter screening model established based on the XgBoost method specifically includes calculating SplitTime of node division of each data in the training data set in the XgBoost tree model and sum ∑Gain of information gain of the data as a splitting node in the entire tree group, and then calculating FeatureImportance of the data: ; The parameter corresponding to the importance exceeding a preset importance threshold is taken as the core parameter.

[0010] Preferably, the selection of the current running condition includes calculating Euclidean distances between real-time running data points to be recognized and historical condition sample points: ; wherein, is a real-time data point, is a historical sample point, m is a core parameter number, , are the kth key parameter values of the real-time and historical data respectively; the historical sample points are sorted in ascending order of Euclidean distances; the first K historical sample points are selected as the nearest neighbor points; the occurrence frequencies of condition categories to which the nearest neighbor points belong are counted, and a condition category with the highest frequency is returned as a running condition of the current equipment.

[0011] Preferably, the abnormality prediction model has a three-layer structure of an input layer, a hidden layer, and an output layer: The input layer: the number of nodes = the number of key parameters m, and the input vector is j=1,2,...,m; Hidden layer: The number of nodes is determined through cross-validation. The input to the hidden layer is then calculated. With output : ; ; in, The connection weights from the j-th node in the input layer to the i-th node in the hidden layer; The threshold value for the i-th node in the hidden layer; It is an sigmoid transfer function; Output layer: Number of nodes = 1, output value Corresponding to the exception type, 0 = normal, 1 = true exception, 2 = false exception, calculate the output of the output layer: ; in: The connection weights from the i-th node in the hidden layer to the k-th node in the output layer; The threshold value for the k-th node of the output layer; It is an sigmoid transfer function, and the output value is mapped to the interval [0,1]; p is the number of hidden layer nodes.

[0012] Preferably, the method also includes calculating the prediction error rate for each anomaly prediction model, including: Error per test sample: ; Average test error: ; Error rate: ; in, For the p-th sample, the actual anomaly type is... For the number of test samples, if The anomaly prediction model training is complete; otherwise, adjust the weights. , and threshold , Repeated training.

[0013] Preferably, calculating the EWI index using the core parameters specifically includes extracting pre-earthquake sensitive key markers from the anomaly prediction model weights, calculating the EWI index, and ultimately determining whether an earthquake early warning is triggered. Record the connection weight percentage of each core parameter in the hidden layer. : ; in, The absolute value of the weight. The sum of the absolute values ​​of all input weights in the hidden layer; select The top n% of key parameters are used as key markers of pre-earthquake sensitivity. Calculate the standardized values ​​of pre-earthquake sensitive key markers : ; in: Let be the j-th sample value of the t-th pre-earthquake sensitive key marker; Let be the sample mean of the t-th pre-earthquake sensitive key marker; Let be the sample standard deviation of the t-th pre-earthquake sensitive key marker; Calculate the standardized marker mean : ; Sample skewness calculation: quantifies data distribution skewness and eliminates the interference of extreme values ​​on early warning. ; Correlation calculation: Internal correlation coefficient Correlation among key pre-earthquake markers: ; in, This is the j-th sample value of another key pre-earthquake sensitive marker; External correlation coefficient Correlation between pre-earthquake sensitive key markers and non-key markers: ; in, The j-th sample value of a non-critical marker; EWI index calculation: .

[0014] An earthquake monitoring and early warning system based on earthquake precursors includes: The data collection unit is responsible for data acquisition and sample construction. It collects core earthquake precursor data and auxiliary status data, labels the samples, and forms a training dataset. The parameter filtering unit establishes a core parameter filtering model based on the XgBoost method, trains the core parameter filtering model with the training dataset, and obtains the core parameters. The working condition discrimination unit identifies working conditions based on KNN, constructs a typical working condition library, calculates the Euclidean distance between real-time data points and historical working condition sample points, selects the top K nearest neighbors, and returns the working condition category with the highest frequency as the current operating condition. The type prediction unit, based on the anomaly prediction of the improved BP neural network, establishes corresponding anomaly prediction models for different operating conditions, and predicts the anomaly type under different current operating conditions according to the anomaly prediction models. The alarm unit, based on the EWI index, calculates the EWI index using the core parameters. When the EWI index is greater than the index threshold, an earthquake early warning is triggered; otherwise, monitoring continues.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an earthquake monitoring and early warning method and system based on earthquake precursors. It uses XgBoost to screen sensitive core parameters before an earthquake, and combines KNN working condition identification and working condition-specific BP neural network to accurately distinguish between normal, true anomalies and false anomalies. Then, it uses EWI index to quantify the correlation between anomaly intensity and pre-earthquake, triggering early warning only for high-value true anomalies, which greatly reduces the risk of false alarms and missed alarms. It automatically identifies false anomalies caused by equipment or environment, and directly pushes targeted operation and maintenance prompts, shortening operation and maintenance response time and reducing data interruption time. It replaces traditional decentralized manual inspection and reduces the daily monitoring workload of a single station. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 The method steps provided by this invention are illustrated. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention discloses an earthquake monitoring and early warning method based on earthquake precursors, such as... Figure 1 As shown, it includes: Step 1: Data Acquisition and Sample Construction. Collect core earthquake precursor data and auxiliary status data, label the samples, and form a training dataset. Step 2: Build a core parameter selection model based on the XgBoost method, train the core parameter selection model with the training dataset, and obtain the core parameters; Step 3: Based on KNN, identify operating conditions, build a typical operating condition library, calculate the Euclidean distance between real-time data points and historical operating condition sample points, select the top K nearest neighbors, and return the operating condition category with the highest frequency as the current operating condition. Step 4: Based on the improved BP neural network for anomaly prediction, establish corresponding anomaly prediction models for different operating conditions, and predict the anomaly types under different current operating conditions according to the anomaly prediction models. Step 5: Early warning determination based on EWI index. The EWI index is calculated using core parameters. When the EWI index is greater than the index threshold, an earthquake early warning is triggered; otherwise, monitoring continues.

[0020] In this step, the multi-parameter-based anomaly type classification outputs a qualitative result of normal, true anomaly, and false anomaly, while the judgment result in step five is a quantitative conclusion on whether the anomaly is sufficient to trigger an early warning.

[0021] In one specific embodiment, if step four is determined to be a false anomaly, an operation and maintenance prompt will be triggered directly (the device is offline and a remote restart is recommended). If step four determines it to be a true anomaly, and the EWI index in step five is less than or equal to the index threshold, then it is marked as a level of concern and continuously monitored (no immediate warning is required to avoid false alarms). If step four determines that the anomaly is genuine, and the EWI index in step five is greater than the index threshold, then an early warning level is triggered and pushed to the earthquake central station (fault warning and remote management).

[0022] In one specific embodiment, the precursor core data includes crustal deformation data, electromagnetic data, underground fluid data, and meteorological correlation data; the auxiliary status data includes equipment operation data, environmental interference data, and network data.

[0023] In one specific embodiment, establishing a core parameter selection model based on the XgBoost method specifically includes calculating the number of times each data point in the training data set is split into nodes in the XgBoost tree model (SplitTime) and the sum of the information gains (∑Gain) of this data as split nodes in the entire tree group, and then calculating the data's importance (FeatureImportance). ; The top 20% of features in terms of FeatureImportance were selected as core parameters.

[0024] Wherein, the sum of information gains ∑Gain: the cumulative information gain that improves the classification accuracy of anomaly type when a certain feature is used as a split node in all decision trees (the higher the ∑Gain, the greater its contribution to the identification of pre-earthquake anomalies). SplitTime: The total number of times a feature is selected as a splitting node in all decision trees (reflecting the core importance of the feature in the classification logic).

[0025] In one specific embodiment, the selection of the current operating condition includes calculating the Euclidean distance between the real-time operating data point to be identified and the historical operating condition sample points: ; in, For real-time data points, Here, m represents the number of historical sample points, and m represents the number of core parameters. , The values ​​are the k-th key parameter values ​​for real-time and historical data, respectively; historical sample points are sorted in ascending order of Euclidean distance; the top K historical sample points are selected as nearest neighbors; the frequency of occurrence of the operating condition category of the nearest neighbors is counted, and the operating condition category with the highest frequency is returned as the current operating condition of the equipment.

[0026] In one specific embodiment, the anomaly prediction model has a three-layer structure: an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is m (the number of key parameters), the number of nodes in the hidden layer is 2m+1, and the number of nodes in the output layer is 1. The model uses tansig+logsig as the transfer function, trainlm as the training function, and learngdm as the learning function. Input layer: Number of nodes = number of key parameters m, input vector is j=1,2,...,m; Hidden layer: The number of nodes is determined through cross-validation. The input to the hidden layer is then calculated. With output : ; ; in, The connection weights from the j-th node in the input layer to the i-th node in the hidden layer; The threshold value for the i-th node in the hidden layer; Using an sigmoid transfer function enhances nonlinear fitting capabilities. ; Output layer: Number of nodes = 1, output value Corresponding to the exception type, 0 = normal, 1 = true exception, 2 = false exception, calculate the output of the output layer: ; in: The connection weights from the i-th node in the hidden layer to the k-th node in the output layer; The threshold value for the k-th node of the output layer; It is an sigmoid transfer function. The output value is mapped to the [0,1] interval for easy classification; p is the number of hidden layer nodes.

[0027] In one specific embodiment, the method further includes calculating the prediction error rate for each anomaly prediction model, including: Error per test sample: ; Average test error: ; Error rate: ; in, For the p-th sample, the actual anomaly type is... For the number of test samples, if The anomaly prediction model training is complete; otherwise, adjust the weights. , and threshold , Repeated training.

[0028] In one specific embodiment, calculating the EWI index using core parameters specifically includes extracting pre-earthquake sensitive key markers from the weights of the anomaly prediction model, calculating the EWI index, and ultimately determining whether an earthquake early warning is triggered. Record the connection weight percentage of each core parameter in the hidden layer. : ; in, The absolute value of the weight. The sum of the absolute values ​​of all input weights in the hidden layer; select The top n% of key parameters are used as key markers of pre-earthquake sensitivity. Calculate the standardized values ​​of pre-earthquake sensitive key markers : ; in: Let be the j-th sample value of the t-th pre-earthquake sensitive key marker; Let be the sample mean of the t-th pre-earthquake sensitive key marker; Let be the sample standard deviation of the t-th pre-earthquake sensitive key marker; Calculate the standardized marker mean : ; Sample skewness calculation: quantifies data distribution skewness and eliminates the interference of extreme values ​​on early warning. ; Correlation calculation: Internal correlation coefficient Correlation among key pre-earthquake markers: ; in, This is the j-th sample value of another key pre-earthquake sensitive marker; External correlation coefficient Correlation between pre-earthquake sensitive key markers and non-key markers: ; in, The j-th sample value of a non-critical marker; EWI index calculation: .

[0029] An earthquake monitoring and early warning system based on earthquake precursors includes: The data collection unit is responsible for data acquisition and sample construction. It collects core earthquake precursor data and auxiliary status data, labels the samples, and forms a training dataset. The parameter filtering unit establishes a core parameter filtering model based on the XgBoost method, trains the core parameter filtering model with the training dataset, and obtains the core parameters. The working condition discrimination unit identifies working conditions based on KNN, constructs a typical working condition library, calculates the Euclidean distance between real-time data points and historical working condition sample points, selects the top K nearest neighbors, and returns the working condition category with the highest frequency as the current operating condition. The type prediction unit, based on the improved BP neural network for anomaly prediction, establishes corresponding anomaly prediction models for different operating conditions, and predicts the anomaly type under different current operating conditions based on the anomaly prediction models. The alarm unit determines early warning based on the EWI index. It calculates the EWI index using core parameters. When the EWI index is greater than the index threshold, an earthquake early warning is triggered; otherwise, it continues to monitor.

[0030] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0031] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for earthquake monitoring and early warning based on earthquake precursors, characterized in that, include: Step 1: Data Acquisition and Sample Construction. Collect core earthquake precursor data and auxiliary status data, label the samples, and form a training dataset. Step 2: Establish a core parameter selection model based on the XgBoost method, train the core parameter selection model with the training dataset, and obtain the core parameters; Step 3: Based on KNN, identify operating conditions, build a typical operating condition library, calculate the Euclidean distance between real-time data points and historical operating condition sample points, select the top K nearest neighbors, and return the operating condition category with the highest frequency as the current operating condition. Step 4: Based on the improved BP neural network for anomaly prediction, establish corresponding anomaly prediction models for different operating conditions, and predict the anomaly types under different current operating conditions according to the anomaly prediction models. Step 5: Early warning determination based on EWI index. Calculate the EWI index using the core parameters. When the EWI index is greater than the index threshold, trigger an earthquake early warning. Conversely, continuous monitoring is required.

2. The earthquake monitoring and early warning method based on earthquake precursors according to claim 1, characterized in that, The precursory core data includes crustal deformation data, electromagnetic data, underground fluid data, and meteorological correlation data; the auxiliary status data includes equipment operation data, environmental interference data, and network data.

3. The earthquake monitoring and early warning method based on earthquake precursors according to claim 1, characterized in that, The core parameter selection model based on the XgBoost method specifically includes calculating the number of times each data point in the training data set is split into nodes in the XgBoost tree model (SplitTime) and the sum of the information gains (∑Gain) of the data as split nodes in the entire tree group, and then calculating the data's importance (FeatureImportance). ； The parameters corresponding to importance levels exceeding the preset importance threshold are used as core parameters.

4. The earthquake monitoring and early warning method based on earthquake precursors according to claim 1, characterized in that, The selection of the current operating condition includes calculating the Euclidean distance between the real-time operating data point to be identified and the historical operating condition sample points: ； in, For real-time data points, Here, m represents the number of historical sample points, and m represents the number of core parameters. , The values ​​are the k-th key parameter values ​​for real-time and historical data, respectively; historical sample points are sorted in ascending order of Euclidean distance; the top K historical sample points are selected as nearest neighbors; the frequency of occurrence of the operating condition category of the nearest neighbors is counted, and the operating condition category with the highest frequency is returned as the current operating condition of the equipment.

5. The earthquake monitoring and early warning method based on earthquake precursors according to claim 1, characterized in that, The anomaly prediction model has a three-layer structure: input layer, hidden layer, and output layer. Input layer: Number of nodes = number of key parameters m, input vector is j=1,2,...,m; Hidden layer: The number of nodes is determined through cross-validation. The input to the hidden layer is calculated. With output : ； ； in, The connection weights from the j-th node in the input layer to the i-th node in the hidden layer; The threshold value for the i-th node in the hidden layer; It is an sigmoid transfer function; Output layer: Number of nodes = 1, output value Corresponding to the exception type, 0 = normal, 1 = true exception, 2 = false exception, calculate the output of the output layer: ； in: The connection weights from the i-th node in the hidden layer to the k-th node in the output layer; The threshold value for the k-th node of the output layer; It is an sigmoid transfer function, and the output value is mapped to the interval [0,1]; p is the number of hidden layer nodes.

6. The earthquake monitoring and early warning method based on earthquake precursors according to claim 5, characterized in that, It also includes calculating the prediction error rate for each anomaly prediction model, including: Error per test sample: ; Average test error: ; Error rate: ; in, Let p be the actual anomaly type of the p-th sample. For the number of test samples, if The anomaly prediction model training is complete; otherwise, adjust the weights. , and threshold , Repeated training.

7. The earthquake monitoring and early warning method based on earthquake precursors according to claim 5, characterized in that, The calculation of the EWI index using the core parameters specifically involves extracting pre-earthquake sensitive key markers from the weights of the anomaly prediction model, calculating the EWI index, and ultimately determining whether an earthquake early warning has been triggered. Record the connection weight percentage of each core parameter in the hidden layer. : ; in, The absolute value of the weight. The sum of the absolute values ​​of all input weights in the hidden layer; select The top n% of key parameters are used as key markers of pre-earthquake sensitivity. Calculate the standardized values ​​of key pre-earthquake sensitive markers : ; in: Let be the j-th sample value of the t-th pre-earthquake sensitive key marker; Let be the sample mean of the t-th pre-earthquake sensitive key marker; Let be the sample standard deviation of the t-th pre-earthquake sensitive key marker; Calculate the standardized marker mean : ; Sample skewness calculation: quantifies data distribution skewness and eliminates the interference of extreme values ​​on early warning. ； Correlation calculation: Internal correlation coefficient Correlation among key pre-earthquake markers: ； in, This is the j-th sample value of another key pre-earthquake sensitive marker; External correlation coefficient Correlation between pre-earthquake sensitive key markers and non-key markers: ； in, The j-th sample value of a non-critical marker; EWI index calculation: .

8. An earthquake monitoring and early warning system based on earthquake precursors, employing the earthquake monitoring and early warning method based on earthquake precursors as described in any one of claims 1-7, characterized in that, Its features include: The data collection unit is responsible for data acquisition and sample construction. It collects core earthquake precursor data and auxiliary status data, labels the samples, and forms a training dataset. The parameter filtering unit establishes a core parameter filtering model based on the XgBoost method, trains the core parameter filtering model with the training dataset, and obtains the core parameters. The working condition discrimination unit identifies working conditions based on KNN, constructs a typical working condition library, calculates the Euclidean distance between real-time data points and historical working condition sample points, selects the top K nearest neighbors, and returns the working condition category with the highest frequency as the current operating condition. The type prediction unit, based on the anomaly prediction of the improved BP neural network, establishes corresponding anomaly prediction models for different operating conditions, and predicts the anomaly type under different current operating conditions according to the anomaly prediction models. The alarm unit, based on the EWI index, calculates the EWI index using the core parameters. When the EWI index is greater than the index threshold, an earthquake early warning is triggered; otherwise, monitoring continues.

Images