Micro-seismic quantitative prediction method based on theoretical model constraint and data driving fusion

Abstract

The invention discloses a micro-seismic quantitative prediction method based on theoretical model constraint and data driving fusion. The method comprises the following steps: 1, acquiring original micro-seismic data of a coal mine working face to form an initial sample data set; 2, preprocessing data and dividing the data into a training set and a test set; 3, constructing a long and short term memory network LSTM model; 4, sampling historical micro-seismic data as a micro-seismic event quantitative prediction sample; 5, constructing a multi-objective optimization model containing seven sub-objective functions including a micro-seismic time fractal dimension, a micro-seismic space fractal dimension, a micro-seismic energy fractal dimension, micro-seismic predicted total energy, micro-seismic reconstructed mining-induced stress, a micro-seismic activity process and a micro-seismic energy-frequency power law b value; and 6, solving an optimal sampling sample as a predicted value. According to the method, quantitative prediction of the micro-seismic event based on theoretical model constraint and data driving fusion is achieved through the steps, the prediction effect is good, and the prediction result can effectively guide on-site production activity process regulation and control, implementation of pressure relief measures and disaster early warning.

Description

Technical Field

[0001] The invention relates to a microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion. Background Art

[0002] In the prevention of rock burst and destructive mine earthquake disasters, disaster prediction is the premise for their prevention and control. How to accurately and quantitatively predict rock burst disasters and then achieve accurate disaster prevention and control is an urgent problem to be solved in mines with rock burst in my country. Existing studies have shown that the generation of rock burst is inevitably accompanied by coal and rock mass fractures and microseismic responses, and is closely related to the size of the fracture. The specific on-site manifestation is the time, space and energy of microseismic occurrence obtained by the microseismic monitoring system. Traditional prediction methods use these strong spatiotemporal information for qualitative prediction, such as comprehensive index method, multi-factor coupling method and frequency-energy evolution law prediction, to make zoning and graded predictions of rock burst hazards.

[0003] As my country's coal mining enters deeper areas, the gradually increasing impact risk has put forward higher requirements for rock burst prediction. The accuracy of qualitative prediction can no longer meet the needs of the engineering site, and the realization of quantitative prediction of microseismic is of great significance to rock burst prediction. For example, the patent with application number 202211011826.9 discloses a quantitative prediction method for the three elements of time, space and intensity of microseismic events in rock burst mines. This method quantitatively predicts the occurrence time, X, Y, Z coordinates of the earthquake location and the energy size of microseismic events through deep learning regression, solving the current problem of lack of mapping relationship between field monitoring parameters and time, space and intensity of disaster events, which is difficult to quantify. However, in the actual production process of coal mines, microseismic activities are restricted and affected by many factors, such as the influence of early stress and geological structure environment and later prevention and mining conditions in the production process of coal mines. The quantitative prediction process of microseismic events described in this patent does not fully consider these factors, and the prediction effect needs to be further improved. Summary of the invention

[0004] In view of the above problems, the present invention provides a microseismic quantitative prediction method based on the fusion of theoretical model constraints and data-driven, comprehensively considers the three elements of microseismic time, space and impact pressure and their correlation, carries out quantitative prediction of microseismic events based on the fusion of theoretical model constraints and data-driven, and uses the new concept of sampling historical data as samples for microseismic quantitative prediction, constructs a multi-objective optimization model based on seven sub-objective functions: microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, microseismic predicted total energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value, imposes three elements of time and space constraints on the sampled microseismic events, and realizes quantitative prediction of microseismic events.

[0005] In order to achieve the above technical objectives and the above technical effects, the present invention is implemented through the following technical solutions:

[0006] A microseismic quantitative prediction method based on theoretical model constraints and data-driven fusion includes the following steps:

[0007] Step 1: Obtain the original microseismic data and daily footage data of coal seam mining during the production period of the coal mine working face, and statistically obtain relevant derivative variables as sample features to form an initial sample data set;

[0008] Step 2: preprocess the initial sample data set and divide the preprocessed data set into a training set and a test set;

[0009] Step 3: Construct a long short-term memory network LSTM model, use the time window sliding method based on microseismic autocorrelation to determine the lag time length of a single sample, and set it as the initial memory number of days, set the number of LSTM layers, fully connected layers and neurons in the model, use mean square error MSE and mean absolute error MAE as loss functions, adjust the model structure through training set loss and test set loss, and obtain the optimal model to predict the total frequency N of microseisms in the next day total and the total energy M total ;

[0010] Step 4: Sampling historical microseismic data as samples for quantitative prediction of microseismic events, including the three elements of microseismic time, space and energy;

[0011] Step 5: Construct a multi-objective optimization model with seven sub-objective functions, namely, microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic predicted energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value, to realize the constraints on the sampled microseismic event samples in the three dimensions of time, space and intensity;

[0012] Step 6: Find the optimal sampling sample as the predicted value.

[0013] Preferably, in step 1, the original microseismic data include the occurrence time and energy of the microseismic events, and the related derived variables include the daily total frequency of microseismic events, the daily total energy, the daily average energy, the daily maximum energy, the daily minimum energy, the daily total energy magnitude, the daily average energy magnitude, the daily maximum energy magnitude, and the daily minimum energy magnitude;

[0014] The daily total frequency, daily total energy, daily average energy, daily maximum energy, and daily minimum energy of microseismic events compared to the previous day;

[0015] The average of the previous three days’ total daily frequency of microseismic events, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, and daily minimum energy magnitude.

[0016] Preferably, in step 2, the frequency and energy data are converted according to the following formula to eliminate outliers:

[0017] M E =lg(E S +10), M N =log2(N Frequency +1)

[0018] Where M E is the energy magnitude, E S For the corresponding energy related data, M N is the total frequency level, N Frequency is the total daily frequency data;

[0019] The processed data is normalized by using the maximum and minimum value normalization method for each column. The calculation method is:

[0020]

[0021] In the formula, is the normalized value, x max is the maximum value of the sample data, x min is the minimum value of the sample data, and x is the original value of the sample;

[0022] 80% of the preprocessed dataset is divided into a training set and 20% is divided into a test set.

[0023] Preferably, the time window sliding method based on microseismic autocorrelation includes:

[0024] The autocorrelation function ACF is used to calculate the correlation between microseismic data and its lagged data. The calculation method is as follows:

[0025]

[0026] In the formula, C ΔI represents the microseismic data x at time i i and the microseismic data x at time i+ΔI i+ΔI covariance, ΔI is the lag in the time dimension, N is the total number of data that need to calculate the autocorrelation, and μ is the mean of all data;

[0027] Autocorrelation ACF of two sets of microseismic data with time lag ΔI ΔI The calculation method is:

[0028]

[0029] In the formula, ACF ΔI is the autocorrelation of two sets of microseismic data with a time difference of ΔI, C0 is the autocovariance when ΔI = 0, that is, the initial term x in the time window iThe autocovariance of

[0030] ACF ΔI The value of is between -1 and 1. ΔI Below the upper confidence interval C I Before, the microseismic data series are correlated, when ACF ΔI Reach C I At the upper limit, ΔI is the maximum correlation period of microseismic data ΔI max , when the time window lag is greater than ΔI max When ΔI max As the initial lag time length of a single sample in the LSTM model;

[0031] C I The calculation method is:

[0032]

[0033] The calculation method of mean square error MSE and mean absolute error MAE is:

[0034]

[0035] In the formula, N is the number of samples, Y i is the true value, is the predicted value.

[0036] Preferably, in step 4, the microseismic prediction simulation test is verified using a reliability function derived based on Weibull probability distribution:

[0037] The probability density function of the three elements of time, space and energy

[0038] In the formula, x is the value of the three elements, x≥0; β is the scale parameter, β>0; α is the shape parameter, α>0;

[0039] Cumulative distribution function

[0040] The reliability function R(Δt) represents the probability of normal operation of the system within the time Δt. It is the complement function of the cumulative distribution function and is specifically expressed as:

[0041] R(Δt)=1-F(Δt;β,α)

[0042] Right now

[0043] The absolute value of the difference between the three elements of the microseismic event variables is used as an independent variable to measure the failure rate of the entire system. The reliability function between microseismic events is as follows:

[0044]

[0045] Where ΔT, ΔS, and ΔlgE are the time difference, distance difference on the plane, and energy level difference between microseismic events, respectively; R(ΔT), R(ΔS), and R(ΔlgE) are the failure rates of microseismic events in terms of time difference, distance difference on the plane, and energy level difference, respectively;

[0046] Construct the reliability objective function:

[0047] maxf(i,j)=ω1·R(ΔT)+ω2·R(ΔS)+ω3·R(ΔlgE)

[0048] Where maxf(i,j) is the reliability between the i-th microseismic event of the real microseismic data and the j-th microseismic data in the historical microseismic event data set. The larger the value, the greater the reliability between the two microseismic events. ω1, ω2, and ω3 are the weights of the reliability in terms of time difference, distance difference in plane position, and energy level difference, respectively.

[0049] Preferably, the models of microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic predicted energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value are as follows:

[0050]

[0051] In the formula, They are the capacity dimensions of the microseismic data of the recent seven days in terms of time fractal, space fractal and microseismic fractal dimensions; The capacity dimension of the microseismic events in the time fractal, space fractal and microseismic fractal capacity dimensions is respectively added to the microseismic data of the recent seven days for sampling prediction; minf(D T ), minf(D S ), minf(D E ) are the errors between the microseismic data of the recent seven days and the sampled microseismic data in three dimensions; minf(Energy) is the total energy of the microseismic data of the next day based on the prediction M total With sampling N total The total energy error of microseismic data, Energy i is the energy of the i-th microseismic event in the sampled microseismic data set, is the total energy of the sampled microseismic event data set; minf(N MS ) is the constraint of the microseismic time factor on the microseismic activity process frequency warning value quantified based on the correlation analysis of production activities, is the frequency warning value of the microseismic activity process of the i-th microseismic event in the sampled microseismic data set, is the sum of the frequency warning values ​​of the microseismic activity process of the sampled microseismic event set, is the mean value of the frequency warning value of the microseismic activity process of the sampled microseismic event set; minf(AI) is the constraint of the impact risk assessment value quantified based on the mining stress reconstruction theory on the microseismic spatial elements, AI i is the risk assessment value of the i-th microseismic event in the sampled microseismic data set, is the sum of the risk assessment values ​​of the sampled microseismic event set, is the mean of the risk assessment values ​​of the sampled microseismic data set; b0 is the b value of the microseismic data of the recent seven days for power law fitting of microseismic energy-frequency; b1 is the b value of the microseismic data of the recent seven days added to the sampled predicted microseismic events for power law fitting of microseismic energy-frequency; minf(b) is the quantitative error between the microseismic data of the recent seven days and the b value of the microseismic data added to the sampled microseismic data for power law fitting of microseismic energy-frequency.

[0052] Preferably, the microseismic energy-frequency power law b value calculation formula is:

[0053] lgN(≥lgE)=a-blgE

[0054] Where, lgE is the microseismic energy level; N (≥lgE) is the number of microseismic events with energy levels greater than or equal to lgE; a and b are constants, and the least squares method is used to calculate the values ​​of a and b:

[0055]

[0056] Where m is the total number of energy level bins; lgE i is the energy level of the i-th gear; N i is the actual number of microseisms at the i-th energy level, and a and b are the estimated parameters obtained by the least squares fitting method.

[0057] Preferably, in step 6, a multi-objective particle swarm algorithm MOPSO is used to find the optimal sampling sample as the prediction value.

[0058] Preferably, the multi-objective particle swarm algorithm MOPSO comprises the following steps:

[0059] 1) Generate an initial particle swarm, where each particle contains a position and velocity vector, as well as the corresponding objective function value. At the same time, initialize an external archive to store non-dominated solutions;

[0060] 2) For each iteration, perform the following steps:

[0061] ① Update particle velocity according to inertia weight, individual optimal solution pBest and global optimal solution gBest:

[0062] v id (t+1)=ω·v id(t)+c1·rand1·(pBest id -x id (t))+c2·rand2·(gBest id -x id (t))

[0063] In the formula, v id (t) is the velocity of the particle at iteration t, x id (t) is the position of the particle, pBest id is the individual optimal position of the particle, gBest id is the global optimal position, ω is the inertia weight, c1 and c2 are learning factors, rand1 and rand2 are random factors;

[0064] ② Adjust the position of the particle according to the updated speed:

[0065] x id (t+1)=x id (t)+v id (t+1)

[0066] ③ Perform non-dominated sorting on the individuals in the population to find the current non-dominated solution set;

[0067] ④ Compare the current non-dominated solution set with the solutions in the external archive and update the external archive;

[0068] ⑤ Select the leader particle gBest from the external archive based on crowding distance and probability;

[0069] 3) When the maximum number of iterations or other termination conditions are reached, the algorithm ends, otherwise repeat step 2);

[0070] 4) Output the non-dominated solution set in the external archive as the optimal solution set of the multi-objective function.

[0071] The beneficial effects of the present invention are:

[0072] (I) The present invention establishes a microseismic total energy and total frequency prediction model that takes into account the regulation of production activity procedures; and constructs a multi-objective optimization microseismic quantitative prediction model containing seven sub-objective functions: microseismic time fractal dimension, space fractal dimension, energy fractal dimension, predicted total energy, reconstructed mining stress, activity procedure, and energy-frequency power law b value. Through the above steps, the present invention realizes the quantitative prediction of microseismic events based on the fusion of theoretical model constraints and data drive, with good prediction effect. The prediction results can effectively guide the regulation of on-site production activity procedures, the implementation of pressure relief measures, and disaster warning.

[0073] (ii) By calculating the fractal dimensions of microseismic time, space, and energy, it is verified that microseismic events have self-similarity in time, space, and energy. Microseismic events similar to real events in the three dimensions of time, space, and energy can be found from historical data. Therefore, the present invention proposes a new concept of sampling historical data as samples for quantitative prediction of microseismic events, and constructs a theoretical constraint model based on production activity correlation analysis, mining stress reconstruction, and energy-frequency power law distribution, thus realizing the constraints on the three dimensions of time, space, and intensity for sampled microseismic events.

[0074] (III) The present invention introduces empirical knowledge, numerical simulation and other theoretical drivers to obtain information such as early stress and geological structure environment and later prevention and mining conditions, and uses microseismic data to drive the reconstruction of mining stress, vibration wave CT detection stress and other information. Mathematical models such as Weibull and normal probability density distribution are used to reconstruct mining stress indicators, and the average of the corresponding indicator values ​​of the predicted samples is taken. The larger the value, the closer it is to the actual microseismic spatial position. The engineering examples have proved that the prediction effect of the present invention is good. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a flow chart of a microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion of the present invention;

[0076] Figure 2 This is a typical ACF plot;

[0077] Figure 3 It is a schematic diagram of the optimal model structure suitable for the prediction of the total energy of microseismic earthquakes the next day in an embodiment of the present invention;

[0078] Figure 4 It is a schematic diagram of the optimal model structure suitable for predicting the total frequency of microseismic events the next day in an embodiment of the present invention;

[0079] Figure 5 It is a schematic diagram of performing prediction simulation prediction by using the obtained reliability function in an embodiment of the present invention;

[0080] Figure 6 is a schematic diagram of comparison between real microseismic data and predicted microseismic data in terms of time variables, energy variables and plane position variables in an embodiment of the present invention;

[0081] Figure 7 1 is a schematic diagram of the effect of prediction simulation when the weights of the three variables ω1:ω2:ω3=2 / 7:4 / 7:1 / 7 in an embodiment of the present invention;

[0082] Figure 8 It is a schematic diagram showing the overall optimal solution set results of three constraint samplings by selecting three constraint conditions of energy, risk assessment value and microseismic activity process frequency warning value in an embodiment of the present invention;

[0083] Fig. 9 It is a schematic diagram of position comparison between three sampling prediction data sets and real microseismic data in an embodiment of the present invention. DETAILED DESCRIPTION

[0084] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0085] like Figure 1 As shown in FIG. 1 , a quantitative microseismic prediction method based on the fusion of theoretical model constraints and data drive includes the following steps:

[0086] Step 1: Obtain the original microseismic data and daily coal seam mining footage data during the production of the coal mine working face, and statistically obtain the relevant derivative variables as sample features to form an initial sample data set, which will be used to construct a long short-term memory network (LSTM) deep learning model in step 3 to predict the total frequency and total energy of microseismic events in the next day.

[0087] Preferably, in step 1, the original microseismic data includes the time and energy of the microseismic event, wherein in step 1, not only the total daily frequency and total energy of microseismic events are used as input parameters of the model, but also the daily footage data of coal seam mining and related derivative variables are introduced, including further processing of frequency and energy data and mean data considering geological factors. Related derivative variables are obtained based on the statistics of the original microseismic data, including:

[0088] Daily total frequency of microseismic events, daily total energy, daily average energy, daily maximum energy, daily minimum energy, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, daily minimum energy magnitude;

[0089] The daily total frequency, daily total energy, daily average energy, daily maximum energy, and daily minimum energy of microseismic events compared to the previous day;

[0090] The average of the previous three days’ total daily frequency of microseismic events, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, and daily minimum energy magnitude.

[0091] Step 2: Preprocess the initial sample data set and divide the preprocessed data set into a training set and a test set.

[0092] When predicting the total frequency and total energy, since the minimum and maximum values ​​are often quite different, and the statistical frequency and energy values ​​may contain 0 values, which affects the prediction accuracy, the frequency and energy data are converted to eliminate outliers from the original data.

[0093] Preferably, in step 2, the frequency and energy data are converted according to the following formula to eliminate outliers:

[0094] M E =lg(E S +10), M N =log2(N Frequency +1)

[0095] Where M E is the energy magnitude, E S is the corresponding energy-related data (such as the daily total energy, daily average energy, daily maximum energy, and daily minimum energy of microseismic events), M N is the total frequency level, N Frequency is the total daily frequency data;

[0096] The processed data is normalized by using the maximum and minimum value normalization method for each column. The calculation method is:

[0097]

[0098] In the formula, is the normalized value, x max is the maximum value of the sample data, x min is the minimum value of the sample data, and x is the original value of the sample;

[0099] 80% of the preprocessed dataset can be divided into a training set and 20% into a test set.

[0100] Step 3: Construct a long short-term memory network LSTM model, use the time window sliding method based on microseismic autocorrelation to determine the lag time length of a single sample, set it as the initial memory days, and set the model LSTM layer, fully connected layer number and number of neurons. The LSTM layer is a long short-term memory layer, which controls the information flow through the input gate, forget gate and output gate. It can flexibly control the transmission and forgetting of information when processing time series data. The fully connected layer converts the output features of the LSTM layer into the final prediction results. The mean square error MSE and mean absolute error MAE are used as loss functions, and the model structure is adjusted through the training set loss and test set loss to obtain the optimal model to predict the total frequency N of microseisms in the next day. total and the total energy M total .

[0101] Preferably, the time window sliding method based on microseismic autocorrelation includes:

[0102] The autocorrelation function ACF is used to calculate the correlation between microseismic data and its lagged data. The calculation method is as follows:

[0103]

[0104] In the formula, C ΔI represents the microseismic data x at time i i and the microseismic data x at time i+ΔI i+ΔI covariance, ΔI is the lag in the time dimension, N is the total number of data that need to calculate the autocorrelation, and μ is the mean of all data;

[0105] Autocorrelation ACF of two sets of microseismic data with time lag ΔI ΔI The calculation method is:

[0106]

[0107] In the formula, ACF ΔI is the autocorrelation of two sets of microseismic data with a time difference of ΔI, C0 is the autocovariance when ΔI = 0, that is, the initial term x in the time window i The autocovariance of

[0108] ACF ΔI The value of is between -1 and 1, and it tends to decrease with the increase of ΔI. Figure 2 is a typical ACF graph. When ACF ΔI Below the upper confidence interval C I Before, the microseismic data series are correlated, when ACF ΔI Reach C I At the upper limit, ΔI is the maximum correlation period of microseismic data ΔI max , when the time window lag is greater than ΔI max When ΔI max As the initial lag time length of a single sample in the LSTM model;

[0109] C I The calculation method is:

[0110]

[0111] The calculation method of mean square error MSE and mean absolute error MAE is:

[0112]

[0113] In the formula, N is the number of samples, Y i is the true value, is the predicted value.

[0114] Step 4: Sample historical microseismic data as samples for quantitative prediction of microseismic events, including the three elements of microseismic time, space and energy. The time element is the time of occurrence without date, such as 12:34:52, the space element is the relative coordinate relative to the center of the current working face mining footage, not the absolute coordinate, and the energy element is the energy level, that is, the logarithm of the absolute value of the microseismic energy.

[0115] Sampling historical microseismic data as samples for quantitative prediction of microseismic events is a new concept proposed by the present invention. Preferably, in step 4, a reliability function derived based on Weibull probability distribution is used to verify the microseismic prediction simulation test:

[0116] The probability density function of the three elements of time, space and energy

[0117] In the formula, x is the value of the three elements, x≥0; β is the scale parameter, β>0; α is the shape parameter, α>0;

[0118] Cumulative distribution function

[0119] The reliability function R(Δt) represents the probability of normal operation of the system within the time Δt. It is the complement function of the cumulative distribution function and is specifically expressed as:

[0120] R(Δt)=1-F(Δt;β,α)

[0121] Right now

[0122] The absolute value of the difference between the three elements of the microseismic event variables is used as an independent variable to measure the failure rate of the entire system. The reliability function between microseismic events is as follows:

[0123]

[0124] Where ΔT, ΔS, and ΔlgE are the time difference, distance difference on the plane, and energy level difference between microseismic events, respectively; R(ΔT), R(ΔS), and R(ΔlgE) are the failure rates of microseismic events in terms of time difference, distance difference on the plane, and energy level difference, respectively;

[0125] Construct the reliability objective function:

[0126] maxf(i,j)=ω1·R(ΔT)+ω2·R(ΔS)+ω3·R(ΔlgE)

[0127] Where maxf(i,j) is the reliability between the i-th microseismic event of the real microseismic data and the j-th microseismic data in the historical microseismic event data set. The larger the value, the greater the reliability between the two microseismic events. ω1, ω2, and ω3 are the weights of the reliability in terms of time difference, distance difference in plane position, and energy level difference, respectively.

[0128] Step 5: Construct a multi-objective optimization model containing seven sub-objective functions: microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, microseismic predicted total energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value, to realize the constraints on the three dimensions of time, space and intensity of the sampled microseismic event samples.

[0129] Preferably, the models of microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic predicted energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value are as follows:

[0130]

[0131] In the formula, They are the capacity dimensions of the microseismic data of the recent seven days in terms of time fractal, space fractal and microseismic fractal dimensions; The capacity dimension of the microseismic events in time fractal, space fractal and microseismic fractal dimensions is respectively added to the microseismic data of the recent seven days for sampling prediction; minf(D T ), minf(D S ), minf(D E ) are the errors between the microseismic data of the recent seven days and the sampled microseismic data in three dimensions; minf(Energy) is the total energy of the microseismic data of the next day based on the prediction M total With sampling N total The total energy error of microseismic data, Energy i is the energy of the i-th microseismic event in the sampled microseismic data set, is the total energy of the sampled microseismic event data set; minf(N MS ) is the constraint of the microseismic activity process frequency warning value on the microseismic time factor based on the correlation analysis of production activities, N MSi is the frequency warning value of the microseismic activity process of the i-th microseismic event in the sampled microseismic data set, is the sum of the frequency warning values ​​of the microseismic activity process of the sampled microseismic event set, is the mean value of the frequency warning value of the microseismic activity process of the sampled microseismic event set; minf(AI) is the constraint of the impact risk assessment value quantified based on the mining stress reconstruction theory on the microseismic spatial elements, AI iis the risk assessment value of the i-th microseismic event in the sampled microseismic data set, is the sum of the risk assessment values ​​of the sampled microseismic event set, is the mean of the risk assessment values ​​of the sampled microseismic data set; b0 is the b value of the microseismic data of the recent seven days for power law fitting of microseismic energy-frequency; b1 is the b value of the microseismic data of the recent seven days added to the sampled predicted microseismic events for power law fitting of microseismic energy-frequency; minf(b) is the quantitative error between the microseismic data of the recent seven days and the b value of the microseismic data added to the sampled microseismic data for power law fitting of microseismic energy-frequency.

[0132] The calculation formulas for microseismic time fractal dimension, microseismic space fractal dimension and microseismic energy fractal dimension are as follows:

[0133]

[0134] In the formula, D0 is the fractal capacity dimension, ε represents the box measurement scale in the fractal dimension calculation method - the box counting method, and N(ε) represents the minimum number of boxes with a length of ε required to cover all microseismic events in time, space, and energy scales. For detailed calculations, please refer to the invention patent (CN201910210964.1, a comprehensive intelligent method for predicting impact hazard levels based on microseismic fractals).

[0135] The calculation of mining stress reconstructed by microseismicity introduces information such as the early stress and geological structure environment and the later prevention and excavation generalization driven by empirical knowledge, numerical simulation and other theories, and combines the mining stress reconstructed by damage driven by microseismic monitoring data, the velocity distribution of vibration wave CT detection and other information. For detailed calculations, please refer to the invention patent (CN202111337335.9, a rock burst risk assessment method based on theoretical and data-driven fusion).

[0136] The microseismic activity process calculation uses the time window sliding method to divide the microseismic sequence into equally spaced sliding time windows, and the cumulative method is used to calculate the microseismic frequency corresponding to each hour in the time window, and the hourly warning value is obtained based on the standard deviation coefficient. For detailed calculations, please refer to the invention patent (CN202210872839.9, a shock warning method based on microseismic activity process time series analysis).

[0137] Preferably, the microseismic energy-frequency power law b value calculation formula is:

[0138] lgN(≥lgE)=a-blgE

[0139] Where, lgE is the microseismic energy level; N (≥lgE) is the number of microseismic events with energy levels greater than or equal to lgE; a and b are constants, and the least squares method is used to calculate the values ​​of a and b:

[0140]

[0141] Where m is the total number of energy level bins; lgE i is the energy level of the i-th gear; N i is the actual number of microseisms at the i-th energy level, and a and b are estimated parameters obtained by least squares fitting. The a value reflects the magnitude of microseismic activity: the larger the value, the stronger the microseismic activity. The b value reflects the magnitude of microseismic intensity: the larger the value, the greater the proportion of low-energy microseismic events.

[0142] Step 6: Find the optimal sampling sample as the predicted value.

[0143] Preferably, in step 6, a multi-objective particle swarm algorithm MOPSO is used to find the optimal sampling sample as the prediction value, and the multi-objective particle swarm algorithm MOPSO includes the following steps:

[0144] 1) Generate an initial particle swarm, where each particle contains a position and velocity vector, as well as the corresponding objective function value. At the same time, initialize an external archive to store non-dominated solutions;

[0145] 2) For each iteration, perform the following steps:

[0146] ① Update particle velocity according to inertia weight, individual optimal solution pBest and global optimal solution gBest:

[0147] v id (t+1)=ω·v id (t)+c1·rand1·(pBest id -x id (t))+c2·rand2·(gBest id -x id (t))

[0148] In the formula, v id (t) is the velocity of the particle at iteration t, x id (t) is the position of the particle, pBest id is the individual optimal position of the particle, gBest id is the global optimal position, ω is the inertia weight, c1 and c2 are learning factors, rand1 and rand2 are random factors;

[0149] ② Adjust the position of the particle according to the updated speed:

[0150] x id (t+1)=x id (t)+v id (t+1)

[0151] ③ Perform non-dominated sorting on the individuals in the population to find the current non-dominated solution set;

[0152] ④ Compare the current non-dominated solution set with the solutions in the external archive and update the external archive;

[0153] ⑤ Select the leader particle gBest from the external archive based on crowding distance and probability;

[0154] 3) When the maximum number of iterations or other termination conditions are reached, the algorithm ends, otherwise repeat step 2);

[0155] 4) Output the non-dominated solution set in the external archive as the optimal solution set of the multi-objective function.

[0156] The present invention is further explained below in conjunction with an embodiment. In this example, the microseismic monitoring system data of the 7302 working face of the Zhaolou Coal Mine from February 2, 2021 to October 31, 2022 is selected as the original data for LSTM model prediction, and a multi-objective optimization model is constructed to realize quantitative prediction of microseismic.

[0157] According to the idea of ​​the present invention, the specific steps are as follows:

[0158] (1) Based on the original microseismic data, the daily total frequency, daily total energy, daily coal mining advance, daily average energy, daily maximum energy, daily minimum energy, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, daily minimum energy magnitude, the increment of frequency and energy series data compared with the previous day, and the three-day average of frequency and energy magnitude series data are obtained to form the initial sample set.

[0159] (2) Preprocess the initial sample data set, eliminate outliers from the original data, eliminate zero values ​​in frequency and energy values, normalize the data after outlier processing according to the maximum and minimum normalization method, and divide 80% of the preprocessed data set into a training set and 20% into a test set.

[0160] (3) Determine the maximum correlation period ΔI of microseismic data using the time window sliding method based on microseismic autocorrelation max = 16 days, set the initial memory days of a single LSTM sample to 16 days, build an LSTM model, use MSE and MAE as model evaluation indicators, train the model for daily total frequency prediction and daily total energy prediction, and adjust the structure. The optimal model structure for the next day’s total energy prediction is [memory days: 16, LSTM: 2, Dense: 6, Units: 256], and the optimal model structure for the next day’s total frequency prediction is [memory days: 20, LSTM: 6, Dense: 4, Units: 1024]. The model structure is as follows: Figure 3 and Figure 4 As shown, the angle brackets indicate the vector dimensions corresponding to the parameters in each layer.

[0161] (4) Sampling historical microseismic data as samples for quantitative prediction of microseismic events includes three elements: time, space, and energy. The time element is the time of occurrence without date, such as 12:34:52. The space element is the relative coordinate relative to the center of the current working face, not the absolute coordinate. The energy element is the energy level, that is, the logarithm of the absolute value of the microseismic energy.

[0162] A microseismic event that occurred in the mine on July 15, 2022 was selected for predictive simulation to verify the feasibility of the concept.

[0163] Select 1000 microseismic data before July 15, count and calculate the reliability function, such as Figure 5 As shown, the derived reliability function is used for predictive simulation.

[0164] First, simulation predictions are performed using reliability functions of individual variables, including time, space, and energy; Figure 6 As shown in the figure, based on the predictions on time variables, energy variables and plane position variables, it can be seen that the actual microseismic data events are almost completely identical to the predicted data, that is, under the single variable constraint, the microseismic events that will occur in the future can be simulated and predicted completely from the sampled events in the historical microseismic data.

[0165] Secondly, the predicted simulation results are compared using multivariable constraints: e.g. Figure 7 The figure shows the effect of prediction simulation when the weights of the three variables ω1:ω2:ω3=2 / 7:4 / 7:1 / 7. The approximate location area of ​​the microseismic can be predicted in space, and the errors in the time and energy level predictions between the real microseismic and predicted microseismic data are very small.

[0166] The prediction results show that the reliability function can effectively constrain variables for prediction, and when predicting with a single variable constraint, microseismic events that are almost consistent with real microseismic events can be predicted, that is, microseismic events similar to real microseismic events can be found in historical microseismic data in terms of time, space, and energy; when using multiple variables to constrain prediction, the weights between the three variables can be adjusted according to the demand for the variables to achieve a more accurate prediction effect, which verifies the concept of sampling from historical microseismic data as samples for quantitative microseismic prediction proposed in the present invention.

[0167] (5) A multi-objective optimization model with seven sub-objective functions, namely, microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic predicted energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b-value, was constructed to realize the constraints on the three dimensions of time, space and intensity of the sampled microseismic event samples.

[0168] (6) The MOPSO multi-objective particle swarm algorithm is used to sample and predict the occurrence of microseismic events on July 15, 2022. The total frequency and total energy of microseismic events on that day predicted by the LSTM model are used as the number of samples and the total energy size in the algorithm in historical microseismic events. The three capacity dimensions in time, space and energy capacity, the total energy size, the frequency warning value of the microseismic activity process, the risk assessment value and the b value of the GR power law fitting are used as constraints. April 16-July 14, 2022 are selected as the sampling data set, and the MOPSO algorithm is set to iterate and optimize 100 times. The algorithm calculates multiple objective functions to find the optimal solution set for all. Figure 8 The figure shows the optimal solution set of three constraint sampling conditions, namely, energy, risk assessment value and microseismic activity process frequency warning value. Fig. 9 It is the position comparison between the three sampling prediction data sets and the real microseismic data.

[0169] (1) The simulation verification results of microseismic prediction show that the microseismic events predicted by single-variable constraints are almost identical to the real microseismic events, which confirms that microseismic events similar to real events can be found from historical data in the three dimensions of time, space and energy. Multivariable constraints can achieve quantitative prediction of microseismic events by adjusting the weights of the reliability functions of the three variables.

[0170] (2) The verification results of quantitative prediction of microseismic events show that the total energy prediction loss MAE and MSE based on the LSTM deep learning model are 0.0827 and 0.0144, respectively, and the total frequency prediction loss MAE and MSE are 0.1672 and 0.0451, respectively, with good prediction effects. On this basis, the MOPSO multi-objective particle swarm algorithm is further used to quantitatively predict the three elements of microseismic time, space and energy. Among them, the time error is mainly controlled in the range of 0 to 4 h, the energy error mainly falls in the range of 0 to 1000 J, and the spatial error accounts for an average of 45% in the same location area.

[0171] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A quantitative microseismic prediction method based on theoretical model constraints and data-driven fusion, characterized in that: The steps include: Step 1: Obtain the original microseismic data and daily footage data of coal seam mining during the production period of the coal mine working face, and statistically obtain relevant derivative variables as sample features to form an initial sample data set; Step 2: preprocess the initial sample data set and divide the preprocessed data set into a training set and a test set; Step 3: Construct a long short-term memory network LSTM model, use the time window sliding method based on microseismic autocorrelation to determine the lag time length of a single sample, and set it as the initial memory number of days, set the number of LSTM layers, fully connected layers and neurons in the model, use mean square error MSE and mean absolute error MAE as loss functions, adjust the model structure through training set loss and test set loss, and obtain the optimal model to predict the total frequency N of microseisms in the next day total and the total energy M total ; Step 4: Sampling historical microseismic data as samples for quantitative prediction of microseismic events, including the three elements of microseismic time, space and energy; Step 5: Construct a multi-objective optimization model with seven sub-objective functions, namely, microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic predicted energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value, to realize the constraints on the sampled microseismic event samples in the three dimensions of time, space and intensity; Step 6: Find the optimal sampling sample as the predicted value.

2. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 1 is characterized by: In step 1, the original microseismic data include the occurrence time and energy of microseismic events, and the related derived variables include the daily total frequency of microseismic events, daily total energy, daily average energy, daily maximum energy, daily minimum energy, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, and daily minimum energy magnitude; The daily total frequency, daily total energy, daily average energy, daily maximum energy, and daily minimum energy of microseismic events compared to the previous day; The average of the previous three days’ total daily frequency of microseismic events, daily total energy magnitude, daily average energy magnitude, daily maximum energy magnitude, and daily minimum energy magnitude.

3. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 2 is characterized by: In step 2, the frequency and energy data are transformed according to the following formula to eliminate outliers: M E =lg(E S +10),M N =log2(N Frequency +1) Where M E is the energy magnitude, E S For the corresponding energy related data, M N is the total frequency level, N Frequency is the total daily frequency data; The processed data is normalized by using the maximum and minimum value normalization method for each column. The calculation method is: In the formula, is the normalized value, x max is the maximum value of the sample data, x min is the minimum value of the sample data, and x is the original value of the sample; 80% of the preprocessed dataset is divided into a training set and 20% is divided into a test set.

4. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 3 is characterized by: The time window sliding method based on microseismic autocorrelation includes: The autocorrelation function ACF is used to calculate the correlation between microseismic data and its lagged data. The calculation method is as follows: In the formula, C ΔI represents the microseismic data x at time i i and the microseismic data x at time i+ΔI i+ΔI covariance, ΔI is the lag in the time dimension, N is the total number of data that need to calculate the autocorrelation, and μ is the mean of all data; Autocorrelation ACF of two sets of microseismic data with time lag ΔI ΔI The calculation method is: In the formula, ACF ΔI is the autocorrelation of two sets of microseismic data with a time difference of ΔI, C0 is the autocovariance when ΔI = 0, that is, the initial term x in the time window i The autocovariance of ACF ΔI The value of is between -1 and 1. When ACF ΔI Below the upper confidence interval C I Before, the microseismic data series are correlated, when ACF ΔI Reach C I At the upper limit, ΔI is the maximum correlation period of microseismic data ΔI max , when the time window lag is greater than ΔI max When ΔI max As the initial lag time length of a single sample in the LSTM model; C I The calculation method is: The calculation method of mean square error MSE and mean absolute error MAE is: In the formula, N is the number of samples, Y i is the true value, is the predicted value.

5. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 4 is characterized by: In step 4, the reliability function derived from the Weibull probability distribution is used to verify the microseismic prediction simulation test: The probability density function of the three elements of time, space and energy Where x is the value of the three elements, x≥0; β is the scale parameter, β>0; α is the shape parameter, α>0; the cumulative distribution function The reliability function R(Δt) represents the probability of normal operation of the system within the time Δt. It is the complement function of the cumulative distribution function and is specifically expressed as: R(Δt)=1-F(Δt;β,α) Right now The absolute value of the difference between the three elements of the microseismic event variables is used as an independent variable to measure the failure rate of the entire system. The reliability function between microseismic events is as follows: ΔT=|T i -T j |(i≠j) ΔS=|S i -S j |(i≠j) Δlg E=|lg E i -lg E j |(i≠j) Where ΔT, ΔS, and ΔlgE are the time difference, distance difference on the plane, and energy level difference between microseismic events, respectively; R(ΔT), R(ΔS), and R(ΔlgE) are the failure rates of microseismic events in terms of time difference, distance difference on the plane, and energy level difference, respectively; Construct the reliability objective function: max f(i,j)=ω1·R(ΔT)+ω2·R(ΔS)+ω3·R(Δlg E) Where max f(i,j) is the reliability between the i-th microseismic event of the real microseismic data and the j-th microseismic data in the historical microseismic event data set. The larger the value, the greater the reliability between the two microseismic events. ω1, ω2, and ω3 are the weights of the reliability in terms of time difference, distance difference in plane position, and energy level difference, respectively.

6. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 5 is characterized by: The models of microseismic time fractal dimension, microseismic space fractal dimension, microseismic energy fractal dimension, total microseismic prediction energy, microseismic reconstruction mining stress, microseismic activity process and microseismic energy-frequency power law b value are as follows: In the formula, They are the capacity dimensions of the microseismic data of the recent seven days in terms of time fractal, space fractal and microseismic fractal dimensions; The capacity dimension of the microseismic events in the time fractal, space fractal and microseismic fractal capacity dimensions is respectively added to the microseismic data of the recent seven days for sampling prediction; min f(D T ), min f(D S ), min f(D E ) are the errors between the microseismic data of the recent seven days and the sampled microseismic data in three dimensions; min f(Energy) is the total energy of the microseismic data of the next day based on the prediction M total With sampling N total The total energy error of microseismic data, Energy i is the energy of the i-th microseismic event in the sampled microseismic data set, is the total energy of the sampled microseismic event data set; min f(N MS ) is the constraint of the microseismic time factor on the microseismic activity process frequency warning value quantified based on the correlation analysis of production activities, is the frequency warning value of the microseismic activity process of the i-th microseismic event in the sampled microseismic data set, is the sum of the frequency warning values ​​of the microseismic activity process of the sampled microseismic event set, is the mean value of the frequency warning value of the microseismic activity process of the sampled microseismic event set; min f(AI) is the constraint of the impact risk assessment value quantified based on the mining stress reconstruction theory on the microseismic spatial elements, AI i is the risk assessment value of the i-th microseismic event in the sampled microseismic data set, is the sum of the risk assessment values ​​of the sampled microseismic event set, is the mean of the risk assessment values ​​of the sampled microseismic data set; b0 is the b value of the microseismic data of the recent seven days for power law fitting of microseismic energy-frequency; b1 is the b value of the microseismic data of the recent seven days added to the sampled predicted microseismic events for power law fitting of microseismic energy-frequency; min f(b) is the quantitative error between the microseismic data of the recent seven days and the b value of the microseismic data added to the sampled microseismic data for power law fitting of microseismic energy-frequency.

7. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 6 is characterized by: The microseismic energy-frequency power law b value calculation formula is: lg N (≥lg E) = ab lg E Where lg E is the microseismic energy level; N (≥lg E) is the number of microseismic events with energy levels greater than or equal to lg E; a and b are constants, and the values ​​of a and b are calculated using the least squares method: Where m is the total number of energy level bins; lgE i is the energy level of the i-th gear; N i is the actual number of microseisms at the i-th energy level, and a and b are the estimated parameters obtained by the least squares fitting method.

8. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 7 is characterized by: In step 6, the multi-objective particle swarm algorithm MOPSO is used to find the optimal sampling sample as the prediction value.

9. The microseismic quantitative prediction method based on theoretical model constraint and data-driven fusion according to claim 8 is characterized by: The multi-objective particle swarm algorithm MOPSO includes the following steps: 1) Generate an initial particle swarm, where each particle contains a position and velocity vector, as well as the corresponding objective function value. At the same time, initialize an external archive to store non-dominated solutions; 2) For each iteration, perform the following steps: ① Update particle velocity according to inertia weight, individual optimal solution pBest and global optimal solution gBest: v id (t+1)=ω·v id (t)+c1·rand1·(pBest id -x id (t))+c2·rand2·(gBest id -x id (t)) In the formula, v id (t) is the velocity of the particle at iteration t, x id (t) is the position of the particle, pBest id is the individual optimal position of the particle, gBest id is the global optimal position, ω is the inertia weight, c1 and c2 are learning factors, rand1 and rand2 are random factors; ② Adjust the position of the particle according to the updated speed: x id (t+1)=x id (t)+v id (t+1) ③ Perform non-dominated sorting on the individuals in the population to find the current non-dominated solution set; ④ Compare the current non-dominated solution set with the solutions in the external archive and update the external archive; ⑤ Select the leader particle gBest from the external archive based on crowding distance and probability; 3) When the maximum number of iterations or other termination conditions are reached, the algorithm ends, otherwise repeat step 2); 4) Output the non-dominated solution set in the external archive as the optimal solution set of the multi-objective function.

Images