Microseismic energy timing prediction method and system

Abstract

The present application provides a kind of microseismic energy timing prediction method and system, it is related to data processing technical field, method includes: obtaining the microseismic waveform and microseismic wave velocity of mine station;Manual wave marking positioning is carried out to microseismic waveform, and waveform onset time is obtained;Combining the position information of mine station, microseismic wave velocity and waveform onset time, the energy sequence of microseismic event is obtained;Energy sequence is pretreated;Through sliding window method, pretreated energy sequence is converted into supervised learning sample, and supervised learning sample is divided into training set, prediction set and validation set.Energy prediction model is constructed;Training set is input into energy prediction model, and energy prediction model is trained;Prediction set is input into the energy prediction model that training is completed, and energy prediction value is obtained;According to validation set, energy prediction value is verified;According to pre-set early warning rule, the risk level of energy prediction value after verification is judged, and according to risk level, early warning information is output.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for predicting microseismic energy time series. Background Technology

[0002] Deep mining faces a complex geological environment characterized by "high altitude, high humidity, high temperature, high wind speed, and high risk of disturbance," leading to frequent coal and rock dynamic disasters. Among these, rockbursts are the most serious due to their suddenness and destructive nature. Since rockbursts are often preceded by precursory microseismic activity, real-time and accurate monitoring and analysis of these microseismic events is crucial for early warning of rockbursts.

[0003] The rockburst monitoring and early warning technology system mainly includes ground stress monitoring, drill cuttings method, electromagnetic radiation method, and microseismic monitoring method. Among them, microseismic monitoring technology has become the core technology because it can capture elastic wave signals in real time in three dimensions, locate the rupture source, and calculate energy. In terms of microseismic energy prediction, existing technologies are divided into two categories: one is traditional time series analysis methods, such as the Autoregressive Integral Moving Average (ARIMA) model and wavelet analysis; the other is single neural network models, such as recurrent neural networks (RNN) and their variant Long Short-Term Memory (LSTM) networks.

[0004] However, while single neural networks such as LSTM can capture long-term temporal dependencies, they cannot effectively extract local mutation features and are susceptible to noise, resulting in poor prediction stability and weak predictive ability for mutation energy events. Overall, existing technologies struggle to simultaneously capture local mutation features and mine long-term evolutionary patterns, leading to insufficient prediction accuracy and robustness, which limits the practical effectiveness of accurate early warning for rockbursts. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a microseismic energy time series prediction method, which can solve the problems of existing technologies where single LSTM and other neural networks, although capable of capturing long-term temporal dependencies, cannot effectively extract local mutation features and are susceptible to noise, resulting in poor prediction stability and weak prediction ability for abrupt energy events. Overall, existing technologies struggle to simultaneously capture local mutation features and mine long-term evolution patterns, resulting in insufficient prediction accuracy and robustness, thus limiting the practical effectiveness of accurate early warning of rockbursts.

[0006] A first aspect of this invention provides a microseismic energy time-series prediction method, comprising:

[0007] S1: Obtain the microseismic waveform and microseismic velocity of the mine station.

[0008] S2: Manually mark the micro-vibration waveform to determine the waveform's initiation time.

[0009] S3: By combining the location information of the mine station, the microseismic wave velocity, and the initiation time of the waveform, the energy sequence of the microseismic event is obtained.

[0010] S4: Preprocess the energy sequence.

[0011] S5: Using the sliding window method, the preprocessed energy sequence is converted into supervised learning samples, and the supervised learning samples are divided into training set, prediction set and validation set according to a preset ratio.

[0012] S6: Construct an energy prediction model.

[0013] S7: Input the training set into the energy prediction model to train the energy prediction model.

[0014] S8: Input the prediction set into the trained energy prediction model to obtain the energy prediction value.

[0015] S9: Validate the energy predictions based on the validation set.

[0016] S10: Based on the preset early warning rules, determine the risk level of the verified energy prediction value, and output early warning information according to the risk level.

[0017] A second aspect of this invention provides a microseismic energy timing prediction system, comprising: a processor and a memory;

[0018] The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the microseismic energy timing prediction method as described in the first aspect.

[0019] A third aspect of the present invention provides a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the microseismic energy timing prediction method as described in the first aspect.

[0020] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0021] In this embodiment of the invention, the energy prediction model compensates for the lack of local key information capture by a single LSTM, is less susceptible to noise, and has good prediction stability and strong prediction ability for abrupt energy events. The energy prediction model uses a convolutional neural network to extract local features, a long short-term memory network to capture long-term temporal dependencies, and a fully connected network for fine regression, so as to take into account both the capture of local abrupt features and the mining of long-term evolutionary patterns. The prediction accuracy and robustness are high, which improves the actual effect of accurate early warning of rockburst. Attached Figure Description

[0022] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0023] Figure 1 This is a flowchart illustrating a microseismic energy time-series prediction method provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the structure of a microseismic energy timing prediction system provided in an embodiment of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions 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, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] The microseismic energy timing prediction method provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0027] Reference manual attached Figure 1 The diagram shows a flowchart of a microseismic energy timing prediction method provided by an embodiment of the present invention.

[0028] This invention provides a method for predicting the timing of microseismic energy, which may include the following steps:

[0029] S1: Obtain the microseismic waveform and microseismic velocity of the mine station.

[0030] S2: Manually mark the micro-vibration waveform to determine the waveform's initiation time.

[0031] In this embodiment of the invention, the core advantage of manual wave-marking positioning is that it can accurately identify the true initiation time of micro-seismic waveforms under complex noise environments, avoid misjudgment of weak signals or interference waveforms by automatic wave-marking, provide an accurate time reference for subsequent energy calculation, ensure the accuracy of micro-seismic event energy calculation, and solidify the key premise of the entire prediction and early warning system.

[0032] S3: By combining the location information of the mine station, the microseismic wave velocity, and the initiation time of the waveform, the energy sequence of the microseismic event is obtained.

[0033] Specifically, the energy value of the microseismic event is calculated by using the original microseismic waveforms collected from mine stations, the initiation time of the manually marked waveforms, the wave velocity, and the station location information, and applying the empirical formula for the attenuation law of the peak vibration velocity of seismic wave particles.

[0034] Specifically, the empirical formula for the attenuation law of peak velocity of seismic wave particles is as follows:

[0035]

[0036] Among them, E c F represents the energy value of a microseismic event. c This represents the radiation coefficient of seismic waves in rock structures. The average radiation coefficient of seismic waves in rock structures is represented by ρ, where ρ represents rock density, R represents the distance between the sensor and the seismic source, and c represents wave velocity. J c π represents the energy flux of seismic waves, and π represents the mathematical constant pi.

[0037] In this embodiment of the invention, by deeply integrating three key parameters—station location, microseismic wave velocity, and initiation time—and following the physical principles of seismic wave energy calculation, a time-continuous microseismic energy sequence is accurately generated. This provides real and physically meaningful core data for subsequent preprocessing, sample construction, and model training, thus solidifying the quantitative analysis foundation of the entire prediction and early warning system.

[0038] S4: Preprocess the energy sequence.

[0039] In one possible implementation, S4 specifically includes sub-steps S401 to S403:

[0040] S401: By using median filtering, noise in the energy sequence is removed to obtain a denoised sequence.

[0041] Median filtering is a nonlinear filtering method that uses a fixed-length sliding window to traverse the microseismic energy sequence and replaces the original data at the center of the window with the median of all data within the window. This method can efficiently remove impulse noise and random interference while preserving key precursor features such as local mutations to the greatest extent possible. It avoids destroying the core laws of energy change during the filtering process and provides clean data support with intact key information for subsequent data preprocessing and model training.

[0042] Specifically, the formula for median filtering is as follows:

[0043]

[0044] in, This represents the filtered energy sequence of the i-th energy sequence. Let k represent the original energy sequence of the i-th energy sequence, and k represent the median filter length. This indicates the floor function, where i represents the index of the energy sequence, N represents the total number of energy sequences, and median() represents the median function.

[0045] It should be noted that the core advantage of median filtering for denoising is that it can efficiently remove impulse noise and random interference from microseismic energy sequences, while preserving key precursor features such as local mutations to the greatest extent possible. This provides clean data for subsequent interpolation, normalization, and model training, significantly improving feature extraction accuracy and system prediction reliability.

[0046] S402: By using linear interpolation, missing data in the denoised sequence is supplemented to obtain a sequence without missing data.

[0047] Among them, linear interpolation refers to the classic data completion method that uses known valid data points adjacent to the missing data, assumes that the data between the two points has a linear trend, and calculates the missing value through a linear equation.

[0048] Specifically, the formula for linear interpolation is as follows:

[0049]

[0050]

[0051] Among them, y i This represents the true target value of the i-th energy sequence. This represents the target value of the i-th energy sequence after missing value processing, where NaN represents the standard identifier for missing values ​​in the data domain.

[0052] It should be noted that the core advantage of using linear interpolation is that it is computationally efficient and can best fit the temporal evolution trend of the original energy sequence, filling in missing data to form a continuous and complete sequence, avoiding data breakage that could disrupt the energy change pattern, and providing coherent and reliable data support for subsequent normalization, sample construction and model training, ensuring the smooth progress of the entire process and the accuracy of the results.

[0053] S403: The sequence without missing parts is processed by Min-Max normalization to obtain the preprocessed energy sequence.

[0054] Min-Max normalization is a classic data standardization method that eliminates differences in the units and numerical ranges of different data by linearly mapping the original data to the [0,1] interval.

[0055] Specifically, the formula for Min-Max normalization is:

[0056]

[0057] in, Let e ​​represent the i-th normalized energy sequence. min e represents the minimum value in the filtered energy sequence. max This represents the maximum value in the filtered energy sequence.

[0058] It should be noted that the core advantage of using Min-Max normalization is that it can quickly eliminate the dimensional differences of energy sequences and map the data to a unified interval [0,1]. This preserves the relative temporal changes of the original energy while avoiding extreme values ​​from dominating model training, significantly improving the convergence speed and feature extraction efficiency of the CNN-LSTM-FC ensemble model, and providing standardized and highly adaptable input data for accurate prediction.

[0059] In this embodiment of the invention, by integrating median filtering, linear interpolation and Min-Max normalization to form a complete preprocessing process, the noise interference, data missing and dimensional differences of microseismic energy sequences are systematically solved, achieving data purification, continuity completion and scale unification, providing high-quality standardized input for subsequent sample construction and model training, and laying a solid foundation for prediction accuracy and system stability from the data source.

[0060] S5: Using the sliding window method, the preprocessed energy sequence is converted into supervised learning samples, and the supervised learning samples are divided into training set, prediction set and validation set according to a preset ratio.

[0061] The preset ratio is 8:1:1.

[0062] In this embodiment of the invention, a sliding window method is used to transform the preprocessed energy sequence into supervised learning samples corresponding to the input and output, accurately preserving the temporal evolution correlation of microseismic energy and adapting to the temporal input characteristics of the CNN-LSTM-FC model. Then, the training set, validation set, and prediction set are divided according to a preset ratio, achieving a scientific division of labor for model training, parameter optimization, and generalization capability verification, ensuring the effectiveness of model training and the reliability of prediction results at the sample level.

[0063] S6: Construct an energy prediction model.

[0064] Specifically, energy prediction models include convolutional neural networks, long short-term memory networks, and fully connected networks.

[0065] Convolutional Neural Networks (CNNs) are deep learning models based on local receptive fields, weight sharing, and pooling operations. They are primarily used to process grid-structured data (such as images) or one-dimensional time-series data. Through convolution operations, they automatically extract local features (such as texture, mutation patterns, and local correlation information) from the data and enhance feature representation through dimensionality reduction and nonlinear transformations, ultimately adapting to classification or regression tasks. Long Short-Term Memory Networks (LSTMs) are improved variants of Recurrent Neural Networks (RNNs), specifically designed to address the vanishing / exploding gradient problems that traditional RNNs easily encounter when processing long-sequence data. Their core functionality involves a gating mechanism to selectively remember and forget historical information, accurately capturing long-term dependencies and dynamic evolution patterns in time-series data. Fully Connected Networks (FCs) are fundamental deep learning modules composed of multiple fully connected layers connected in series. Their core feature is that each neuron is fully connected to all neurons in the previous layer, enabling nonlinear integration and dimensionality mapping of high-dimensional input features, ultimately transforming abstract features into output results adapted to specific tasks (classification / regression).

[0066] In this embodiment of the invention, CNN is used to accurately extract local mutation features of microseismic energy, LSTM is used to deeply mine long-term temporal dependence patterns, and FC is used to complete accurate regression mapping. This achieves complementary advantages of multiple modules, effectively adapts to the nonlinear and non-stationary characteristics of microseismic energy sequences, and greatly improves prediction accuracy and robustness, providing high-performance core model support for early warning of rockburst.

[0067] S7: Input the training set into the energy prediction model to train the energy prediction model.

[0068] In this embodiment of the invention, by inputting the training set into the energy prediction model for training, the model can fully learn the local mutation characteristics and long-term evolution law of the microseismic energy sequence, continuously optimize the parameters of each layer to reduce prediction error, significantly improve the model's ability to fit the energy change law, and lay a solid performance foundation for subsequent model verification, accurate prediction and early warning of rockburst.

[0069] S8: Input the prediction set into the trained energy prediction model to obtain the energy prediction value.

[0070] In one possible implementation, S8 specifically includes sub-steps S801 to S803:

[0071] S801: Extract local temporal features and abrupt change features from the prediction set in a convolutional neural network.

[0072] In one possible implementation, S801 specifically includes sub-steps S8011 to S8013:

[0073] S8011: Extracts local temporal features of the prediction set through one-dimensional convolution operations.

[0074] The specific formula for one-dimensional convolution operation is as follows:

[0075]

[0076] in, Let j represent the j-th feature of the l-th layer, and ReLU represent the activation function. This represents the connection weight between the j-th feature of layer l and the p-th feature of layer (l-1). M represents the corresponding bias coefficient. j The input waveform is represented by ∑, which represents the summation operation, and ∈ indicates that it belongs to. This represents element-wise multiplication, where l represents the l-th layer of the convolutional neural network, and j represents the j-th feature.

[0077] It should be noted that by accurately extracting local temporal features from the prediction set through one-dimensional convolution operations, the model can efficiently adapt to the characteristics of one-dimensional time-series microseismic energy data, accurately capture local fluctuations, short-term trend changes, and temporal correlation information of the energy sequence, filter redundant interference, and provide clean and representative local feature inputs for subsequent LSTM modeling. This significantly improves the model's ability to capture the local evolution of energy, enhances the accuracy of prediction details, and improves the effectiveness of precursor identification.

[0078] S8012: Local temporal features are activated through a nonlinear activation function to obtain an activated feature sequence.

[0079] It should be noted that by activating local time-series features through nonlinear activation functions, nonlinear transformations are introduced to break the limitations of linear operations, enabling the model to accurately fit the complex nonlinear evolution of microseismic energy sequences. This significantly enhances the expression and discrimination capabilities of features, providing more discriminative feature inputs for subsequent pooling dimensionality reduction and long-term LSTM-dependent modeling, and significantly improving the model's ability to capture and model complex precursor patterns.

[0080] S8013: By using a pooling layer, the activated feature sequence is pooled to obtain mutation features.

[0081] Specifically, the formula for the pooling layer is as follows:

[0082]

[0083] in, This represents the max-pooling output value of the j-th feature map in the l-th layer at time step t. This indicates the time interval corresponding to the pooling window. denoted as the original activation value of the j-th feature map in layer l at time step t′, where t represents time step t, t′ represents the time step of the pooling window, and max represents the maximum value.

[0084] It should be noted that pooling the activated feature sequences through pooling layers significantly reduces feature dimensionality while preserving core mutation features, thereby reducing redundant information and computational overhead and improving model efficiency. It also enhances feature robustness, filters out minor fluctuations, and allows subsequent LSTM to focus more on key precursors of mutations, significantly improving the model's ability to capture microseismic energy mutation events and ensuring the accuracy and efficiency of early warning.

[0085] In this embodiment of the invention, convolutional neural networks are used to accurately extract local temporal features and abrupt change features of concentrated microseismic energy, efficiently capture key precursor signals before rockbursts occur, make up for the shortcomings of single models in capturing local anomaly information, provide pure and critical feature inputs for subsequent long-term time-dependent modeling, significantly improve the model's ability to identify and predict abrupt energy events, and enhance the foresight and reliability of early warning.

[0086] S802: In long short-term memory networks, long-term temporal dependence characteristics and dynamic evolution laws are obtained based on local temporal characteristics and mutation characteristics.

[0087] In one possible implementation, S802 specifically includes sub-steps S8021 to S8025:

[0088] S8021: By using the forgetting gate, the degree of retention of historical memory information is controlled to obtain the state of memory retention.

[0089] Specifically, the formula for the forgetting gate is:

[0090]

[0091] in, f represents the output of the forget gate at time step t. t W represents the input to the forget gate at time step t. f U represents the input weight matrix of the forget gate. f h represents the weight matrix for the self-connection from the hidden layer to the hidden layer via the forget gate. t-1 b represents the hidden layer state at time step t-1. f σ represents the bias term of the forget gate, and σ represents the Sigmoid activation function.

[0092] It should be noted that by precisely controlling the degree of retention of historical memory information through the forget gate, redundant and invalid information in the microseismic energy sequence is intelligently removed, key long-term temporal dependence features are accurately retained, adapting to the characteristics of long sequence data, effectively avoiding interference from invalid information, significantly improving the model's ability to capture the long-term evolution law of energy, and enhancing the stability and accuracy of prediction.

[0093] S8022: By using the input gate, the influence of memory information on local temporal features and mutation features is controlled to obtain candidate memory states.

[0094] Specifically, the formula for the input gate is:

[0095]

[0096] Among them, i t W represents the output of the input gate at time step t. i U represents the input weight matrix of the input gate. i b represents the weight matrix for the self-connection from the input gate hidden layer to the hidden layer. i This represents the bias term of the input gate.

[0097] Specifically, the formula for candidate memory states is as follows:

[0098]

[0099] in, Let W represent the candidate cell state at time step t, tanh represent the hyperbolic tangent activation function, and W... c The input feature weight matrix, U, represents the candidate cell state. c The hidden state weight matrix, b, represents the candidate cell state. c Bias terms representing the candidate cell state.

[0100] It should be noted that by precisely controlling the influence of memory information on local temporal features and mutation features through the input gate, effective new features are intelligently selected and integrated into the memory unit, achieving efficient fusion of old and new features, strengthening the capture and integration of key precursor information, providing high-quality candidate features for subsequent memory state updates and long-term dependency modeling, and significantly improving the model's ability to fit complex energy change patterns and its predictive reliability.

[0101] S8023: Update the memory cell state based on the retained memory state and the candidate memory state.

[0102] Specifically, the formula for calculating the updated memory cell state is as follows:

[0103]

[0104] Among them, c t Represents the memory unit updated at time step t, ⊙ represents element-wise multiplication, and c t-1 i represents the memory cell updated at time step t-1. t This represents the output value of the input gate at time step t.

[0105] It should be noted that by fusing retained memory states and candidate memory states to achieve dynamic updates of memory units, the model accurately integrates historical effective information with newly extracted key precursor features, continuously optimizes the memory units' ability to represent the long-term dependence and abrupt change patterns of microseismic energy, provides memory support that is more in line with the real laws for subsequent time series output and accurate prediction, and significantly improves the reliability of the model for long-term modeling and early warning of complex energy evolution.

[0106] S8024: Obtain the hidden state through the output gate based on the updated memory unit.

[0107] Specifically, the formula for the output gate is as follows:

[0108]

[0109] Among them, o t W represents the output value of the output gate at time step t. o U represents the input weight matrix of the output gate. o b represents the weight matrix for the self-connection between hidden layers from the output gate. o This represents the bias term of the output gate.

[0110] Specifically, the formula for the hidden state is as follows:

[0111]

[0112] Among them, h t This represents the hidden layer state at time step t.

[0113] It should be noted that by accurately filtering the effective information in the updated memory unit through the output gate, a hidden state with high representational power is generated, which enhances the characterization of the complex temporal pattern of microseismic energy, provides high-quality feature input for subsequent fully connected layer prediction, and significantly improves the accuracy of model output and the reliability of early warning.

[0114] S8025: Through multi-timestep long short-term memory network operations, long-term temporal dependence characteristics and dynamic evolution laws are obtained based on the hidden states.

[0115] It should be noted that by using multi-time-step short-time memory network operations, the long-term temporal dependence characteristics and dynamic evolution laws of microseismic energy are deeply mined based on the hidden state, accurately depicting the long-term changes and dynamic evolution of energy, breaking through the limitations of traditional models in capturing long-term temporal dependence, significantly improving the model's ability to model complex energy dynamic laws, and providing deep-seated support for accurate prediction and early warning.

[0116] In this embodiment of the invention, by deeply integrating local temporal features and abrupt change features through a long short-term memory network, the long-term temporal dependence and complex dynamic evolution of microseismic energy sequences are accurately mined. This overcomes the limitations of traditional models in characterizing long-term dynamic logic, strengthens the ability to represent deep energy change patterns, provides core legal support for subsequent accurate prediction and early warning of rockbursts, and significantly improves the reliability of the model's analysis and early warning of complex energy dynamics.

[0117] S803: In a fully connected network, energy prediction values ​​are obtained based on long-term time-dependent characteristics and dynamic evolution laws.

[0118] Specifically, in a fully connected network, a fully connected mapping is performed, and the formula for the fully connected mapping is as follows:

[0119]

[0120] in, The values ​​represent predicted energy, where W1 represents the first weighting matrix and W2 represents the second weighting matrix. Let represent the nonlinear activation function, z represent the input feature vector of the fully connected layer, b1 represent the first bias term of the fully connected layer, and b2 represent the second bias term of the fully connected layer.

[0121] It should be noted that by using a fully connected network to efficiently map long-term time-dependent features and dynamic evolution laws into energy prediction values, and accurately integrate deep time-series and dynamic law information, the direct conversion from features to prediction results is achieved, outputting highly accurate energy prediction values. This provides an intuitive and reliable quantitative basis for early warning of rockbursts, significantly improving the practicality of the warning and its decision-making reference value.

[0122] In this embodiment of the invention, the prediction set is input into the trained energy prediction model. Through the collaborative operation of CNN, LSTM and fully connected layers, the entire process from local feature extraction and long-term pattern mining to the final prediction value output is efficiently transformed. This accurately depicts the complex dynamic evolution characteristics of microseismic energy and outputs highly accurate energy prediction results, providing a reliable quantitative basis for early warning of rockbursts and significantly improving the accuracy and practicality of the warning.

[0123] S9: Validate the energy predictions based on the validation set.

[0124] In one possible implementation, S9 specifically includes sub-steps S901 to S904:

[0125] S901: Calculate the prediction error index for the validation set.

[0126] It should be noted that by calculating the prediction error index of the validation set, the performance of the energy prediction model can be quantitatively evaluated, accurately reflecting the model's generalization ability and prediction reliability on unknown data, intuitively presenting the degree of prediction deviation, providing objective data basis for model parameter optimization and structural adjustment, effectively ensuring the accuracy of the final energy prediction results, and laying a solid evaluation foundation for the credibility of rockburst early warning.

[0127] S902: Verify the energy prediction value based on the prediction error index to obtain the prediction error result.

[0128] It should be noted that by verifying the energy prediction value based on the prediction error index, the accurate prediction error results are output, the accuracy and reliability of the model prediction are objectively quantified, the prediction deviation is presented intuitively, and clear data guidance is provided for model iteration and optimization. This effectively ensures the credibility of the energy prediction results and lays a solid verification foundation for the accuracy of early warning of rockburst.

[0129] S903: Calculate the root mean square error of the prediction error result.

[0130] It should be noted that by calculating the root mean square error of the prediction results, the average deviation of the model prediction is quantified in a sensitive and intuitive way, accurately reflecting the deviation between the predicted value and the true value. This provides core quantitative basis for model performance evaluation, helps to accurately locate model shortcomings, and provides key data support for subsequent model iteration optimization and improvement of early warning reliability.

[0131] S904: Evaluate the prediction performance of the energy prediction model based on the root mean square error.

[0132] It should be noted that the prediction effect of the energy prediction model is scientifically evaluated based on the root mean square error. The quantitative indicators objectively reflect the prediction accuracy and generalization ability of the model, accurately judge the performance of the model, provide a clear direction for model iteration and optimization, effectively ensure the reliability of energy prediction results, and provide key performance support for the stable operation and accurate early warning of rockburst early warning system.

[0133] In this embodiment of the invention, the energy prediction value is verified in a closed loop throughout the entire process based on the validation set. Through multi-dimensional error calculation, result verification and quantitative evaluation, a complete model performance verification system is formed to objectively measure the model's generalization ability and prediction accuracy, provide accurate data support for model iterative optimization, ensure that the prediction results are true and reliable, and build a solid core performance guarantee for early warning of rockburst.

[0134] S10: Based on the preset early warning rules, determine the risk level of the verified energy prediction value, and output early warning information according to the risk level.

[0135] Specifically, the risk levels include: normal level, attention level, warning level, and danger level.

[0136] For example, a threshold can be set based on the predicted residuals from the training set and historical energy quantiles:

[0137] When the verified energy prediction value is less than the first set threshold, the risk level of the verified energy prediction value is determined to be normal.

[0138] When the verified energy prediction value is greater than or equal to the first set threshold and less than the second set threshold, the risk level of the verified energy prediction value is determined to be a concern level.

[0139] When the verified energy prediction value is greater than or equal to the second set threshold and less than the third set threshold, the risk level of the verified energy prediction value is determined to be the warning level.

[0140] When the verified energy prediction value is greater than or equal to the third set threshold, the risk level of the verified energy prediction value is determined to be dangerous.

[0141] When consecutive preset time steps meet the same level, an early warning message is output.

[0142] It should be noted that those skilled in the art can set the first set threshold, the second set threshold, the third set threshold, and the preset time step according to actual needs, and the present invention does not limit these settings.

[0143] In this embodiment of the invention, the risk level of the verified energy prediction value is determined according to the preset early warning rules, the abstract prediction result is transformed into an intuitive risk classification conclusion, and then targeted early warning information is output according to the level, so as to realize the standardization and intelligence of early warning, which not only improves the accuracy and timeliness of early warning response, but also provides a clear decision basis for on-site risk prevention and control, effectively enhancing the practicality and operability of early warning of rockburst.

[0144] The microseismic energy time-series prediction method provided in this application can be executed by a microseismic energy time-series prediction device. This application uses the microseismic energy time-series prediction device executing the method as an example to illustrate the microseismic energy time-series prediction device provided in this application.

[0145] Reference manual attached Figure 2 The diagram shows a structural schematic of a microseismic energy timing prediction device provided in an embodiment of the present invention.

[0146] This invention provides a microseismic energy timing prediction system 20, including: a processor 201 and a memory 202;

[0147] The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the microseismic energy timing prediction method described above and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0148] It should be understood that the processor 201 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0149] It should also be understood that the memory 202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM).

[0150] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0151] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0152] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0153] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0154] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0155] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0156] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0157] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0158] This invention provides a readable storage medium comprising: storing a program or instructions on the readable storage medium, wherein when the program or instructions are executed by a processor, the program or instructions implement the steps of the above-described microseismic energy timing prediction method and achieve the same technical effect. To avoid repetition, this invention will not elaborate further.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the time series of microseismic energy, characterized in that, include: S1: Obtain the microseismic waveform and microseismic wave velocity of the mine station; S2: Manually mark the micro-vibration waveform to determine the waveform's initiation time; S3: Combining the location information of the mine station, the microseismic wave velocity, and the initiation time of the waveform, the energy sequence of the microseismic event is obtained; S4: Preprocess the energy sequence; S5: Using the sliding window method, the preprocessed energy sequence is converted into supervised learning samples, and the supervised learning samples are divided into training set, prediction set and validation set according to a preset ratio; S6: Construct an energy prediction model; S7: Input the training set into the energy prediction model to train the energy prediction model; S8: Input the prediction set into the trained energy prediction model to obtain the energy prediction value; S9: Verify the energy prediction value based on the verification set; S10: Based on the preset early warning rules, determine the risk level of the verified energy prediction value, and output early warning information according to the risk level.

2. The microseismic energy time-series prediction method according to claim 1, characterized in that, S4 specifically includes: S401: By using median filtering, noise in the energy sequence is removed to obtain a denoised sequence; S402: By using linear interpolation, the missing data in the denoised sequence is supplemented to obtain a sequence without missing data; S403: The missing sequence is processed by Min-Max normalization to obtain the preprocessed energy sequence.

3. The microseismic energy time-series prediction method according to claim 1, characterized in that, The energy prediction model includes: convolutional neural networks, long short-term memory networks, and fully connected networks.

4. The microseismic energy time series prediction method according to claim 3, characterized in that, S8 specifically includes: S801: In the convolutional neural network, extract the local temporal features and abrupt change features of the prediction set; S802: In the Long Short-Term Memory network, based on the local temporal characteristics and the mutation characteristics, the long-term temporal dependence characteristics and dynamic evolution law are obtained; S803: In the fully connected network, the predicted energy value is obtained based on the long-term temporal dependence characteristics and the dynamic evolution law.

5. The microseismic energy time-series prediction method according to claim 4, characterized in that, The convolutional neural network specifically includes: one-dimensional convolution operation, non-linear activation function, and pooling layer; Specifically, S801 includes: S8011: Extract local temporal features of the prediction set through the one-dimensional convolution operation; S8012: Activate the local temporal features through the nonlinear activation function to obtain the activation feature sequence; S8013: The activation feature sequence is pooled through the pooling layer to obtain the mutation feature.

6. The microseismic energy time series prediction method according to claim 4, characterized in that, The Long Short-Term Memory Network specifically includes: a forget gate, an input gate, and an output gate; Specifically, S802 includes: S8021: By using the forgetting gate, the degree of retention of historical memory information is controlled to obtain the memory retention state; S8022: By controlling the influence of memory information on the local temporal features and the mutation features through the input gate, candidate memory states are obtained; S8023: Update the memory cell state according to the retained memory state and the candidate memory state; S8024: Obtain the hidden state through the output gate based on the updated memory unit; S8025: Through multi-timestep long short-term memory network operations, the long-term temporal dependence features and the dynamic evolution law are obtained based on the hidden state.

7. The microseismic energy time-series prediction method according to claim 1, characterized in that, S9 specifically includes: S901: Calculate the prediction error index of the validation set; S902: Verify the predicted energy value according to the prediction error index to obtain the prediction error result; S903: Calculate the root mean square error of the prediction error result; S904: Evaluate the prediction performance of the energy prediction model based on the root mean square error.

8. The microseismic energy time-series prediction method according to claim 1, characterized in that, The risk levels include: normal level, attention level, warning level, and danger level.

9. A microseismic energy time-series prediction system, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the microseismic energy timing prediction method as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the microseismic energy timing prediction method as described in any one of claims 1 to 8.

Images