CN-121995434-A - Coal seam hydraulic fracturing microseismic signal identification method and system based on power spectral density

CN121995434ACN 121995434 ACN121995434 ACN 121995434ACN-121995434-A

Abstract

The invention discloses a method and a system for identifying a hydraulic fracturing microseismic signal of a coal seam based on power spectrum density, wherein the method and the system are used for preprocessing acquired signals to obtain the microseismic signal, the microseismic signal is divided into a plurality of sub-bands, each sub-band is used as an independent branch, continuous wavelet convolution is adopted in each independent branch to extract multi-scale time-frequency characteristics, a space-spectrum-time characteristic block is constructed, the multi-scale time-frequency characteristics are input into the characteristic block, a channel attention mechanism is introduced to carry out self-adaptive weighting on key information to obtain the multi-scale time-frequency characteristics, then the power spectrum density difference is introduced into a microseismic signal identification process, the power spectrum density statistical characteristics and the multi-scale time-frequency characteristics are fused to form a joint characteristic vector, and finally an intelligent identification model of a double-branch parallel mixed structure is utilized to analyze the joint characteristic vector, so that automatic identification and classification of microseismic events and noise signals are realized. The invention obviously improves the accuracy and stability of the identification of the microseism event under the conditions of strong noise and non-stability.

Inventors

QIAN YANAN
LIU TING
ZHAI CHENG
LI QUANGUI
XU HEXIANG
Wen hongda
XU JIZHAO
YU XU
WANG JIWEI

Assignees

中国矿业大学

Dates

Publication Date: 20260508
Application Date: 20260225

Claims (8)

1. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density is characterized by comprising the following steps of: step one, preprocessing the collected continuous microseismic monitoring waveforms to obtain processed microseismic signals; Step two, obtaining an effective frequency range according to the processed microseismic signals, dividing the effective frequency range into a plurality of sub-frequency bands in the range, and inputting each sub-frequency band as an independent branch; Step three, constructing a space-spectrum-time feature block, inputting the multi-scale time-frequency features extracted in the step two into the feature block, extracting space and time combined features through convolution operation, introducing a channel attention mechanism to carry out self-adaptive weighting on key information, and respectively obtaining a low-dimensional global feature and a high-dimensional local feature; Step four, calculating power spectral density of the microseismic signals processed in the step one, extracting power spectral density statistical features, and fusing the power spectral density statistical features with the low-dimensional global features and the high-dimensional local features obtained in the step three to form a joint feature vector; And fifthly, constructing an intelligent recognition model of the double-branch parallel hybrid architecture, inputting the combined feature vector obtained in the fourth step into the recognition model, and outputting a recognition result by the model to realize automatic recognition and classification of the microseism event and the noise signal.
2. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the preprocessing in the first step comprises the steps of trending treatment, band-pass filtering and amplitude normalization in sequence so as to eliminate background drift and low-frequency interference and reserve a main energy distribution frequency band.
3. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the continuous wavelet convolution in the second step is realized by adopting a Morlet wavelet function.
4. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the space-spectrum-time characteristic block in the third step consists of a two-layer convolution and channel attention mechanism, and is specifically as follows: I. The first layer convolution is to operate on the time-frequency image by utilizing two-dimensional convolution kernel, so that the space and time characteristics are extracted simultaneously: Wherein, the Representing an input time-frequency characteristic tensor; Representing a first layer two-dimensional convolution kernel weight parameter; Wherein: 10 represents the size of the convolution kernel in the time dimension; representing a two-dimensional convolution operation; Representing a convolutional layer bias parameter; (-) represents a nonlinear activation function; representing a first layer convolution output feature map; II. Channel attention mechanism, obtaining channel weights by global average pooling And obtaining importance coefficients through nonlinear mapping, wherein the calculation mode is as follows: Wherein the method comprises the steps of For the purpose of global averaging pooling, In order to provide a ReLU, Is Sigmoid; Representing a second layer two-dimensional convolution kernel weight parameter; the final reinforced features are: III, second-layer convolution, namely further extracting local time features on the features after channel enhancement: And then, the key information is promoted through a channel attention mechanism, so that the following steps are obtained: Wherein, the As a low-dimensional global feature, Is a high-dimensional local feature.
5. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the power spectrum density statistical characteristics in the fourth step comprise power spectrum shannon entropy, dominant frequency, average frequency, energy and duration.
6. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the power spectrum density in the fourth step is calculated by adopting a Welch method or a multi-window spectrum estimation method.
7. The method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density according to claim 1, wherein the dual-branch parallel hybrid architecture of the intelligent identification model in the fifth step comprises branches for deep time-frequency characteristic learning and branches for power spectrum density statistical characteristic analysis.
8. A system for using the identification method of any one of claims 1 to 7, comprising: the data acquisition module is used for acquiring continuous microseism monitoring waveform data generated in the hydraulic fracturing process of the coal seam; the signal preprocessing module is used for preprocessing the continuous microseismic monitoring waveform data and eliminating the influence of background drift and environmental noise; the frequency division and time-frequency characteristic extraction module is used for carrying out multi-branch frequency division on the preprocessed microseismic signals and extracting multi-scale time-frequency characteristics through continuous wavelet convolution; the space-spectrum-time feature extraction module is used for carrying out space, spectrum and time joint modeling on the multi-scale time-frequency features, enhancing feature information with discrimination through an attention mechanism and obtaining low-dimensional global features and high-dimensional local features; the power spectral density characteristic extraction module is used for calculating the power spectral density of the signal and extracting the power spectral density statistical characteristic; The feature fusion module is used for fusing the low-dimensional global features, the high-dimensional local features and the power spectral density statistical features to construct a joint feature vector; And the signal identification and classification module takes the joint feature vector as input, and realizes automatic identification and classification of the microseism event and the noise signal through a double-branch parallel hybrid architecture.

Description

Coal seam hydraulic fracturing microseismic signal identification method and system based on power spectral density Technical Field The invention belongs to the technical field of microseism monitoring and signal processing, and particularly relates to a method and a system for identifying a hydraulic fracturing microseism signal of a coal seam based on power spectral density. Background The hydraulic fracturing of the coal seam is an important engineering technical means for improving the permeability of the coal seam and enhancing the gas extraction effect, and the microseismic monitoring technology is one of key methods for evaluating the hydraulic fracturing transformation effect because the microseismic monitoring technology can reflect the cracking and expanding processes of the crack in real time. The accurate identification of the microseismic event is the basis of microseismic monitoring data analysis and seismic source positioning, and the identification quality of the microseismic event directly influences the reliability of the fracture network inversion result. The existing microseismic signal identification method is mainly based on time domain characteristics, time-frequency analysis or artificial experience threshold values for judgment. However, in underground coal mine or ground well fracturing monitoring, the monitoring environment is complex, the background noise types are various, including mechanical vibration, electromagnetic interference, environmental noise and the like, and the traditional method has limited capability of describing the intrinsic spectrum characteristics of the microseismic signals under the condition of strong noise, so that misjudgment or missed judgment is easy to occur. The power spectral density is taken as an important tool for describing the distribution characteristics of signal energy in the frequency domain, and can reflect the stability and the concentration of the overall spectrum structure of the signal. The existing researches mostly use the power spectrum density for analysis of background noise of a station or evaluation of signal quality, but the researches for systematically introducing the power spectrum density difference into a micro-seismic signal identification process are still limited, and a micro-seismic signal identification method which is oriented to complex noise environment of hydraulic fracturing of a coal bed and can be applied in engineering is not yet available. Therefore, the method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density is provided, so that accuracy and stability of microseismic event identification under a complex noise background are improved, and the method has important engineering application value. Disclosure of Invention Aiming at the problems in the prior art, the invention provides a coal seam hydraulic fracturing microseismic signal identification method and system based on power spectral density, which can effectively solve the problems in the prior art. In order to achieve the purpose, the technical scheme adopted by the invention is that the method for identifying the hydraulic fracturing microseismic signals of the coal seam based on the power spectrum density comprises the following steps: step one, preprocessing the collected continuous microseismic monitoring waveforms to obtain processed microseismic signals; Step two, obtaining an effective frequency range according to the processed microseismic signals, dividing the effective frequency range into a plurality of sub-frequency bands in the range, and inputting each sub-frequency band as an independent branch to enhance the frequency resolution capability; Step three, constructing a space-spectrum-time feature block, inputting the multi-scale time-frequency features extracted in the step two into the feature block, extracting space and time combined features through convolution operation, introducing a channel attention mechanism to carry out self-adaptive weighting on key information, and respectively obtaining a low-dimensional global feature and a high-dimensional local feature; Step four, calculating power spectral density of the microseismic signals processed in the step one, extracting power spectral density statistical features, and fusing the power spectral density statistical features with the low-dimensional global features and the high-dimensional local features obtained in the step three to form a joint feature vector; And fifthly, constructing an intelligent recognition model of the double-branch parallel hybrid architecture, inputting the combined feature vector obtained in the fourth step into the recognition model, and outputting a recognition result by the model to realize automatic recognition and classification of the microseism event and the noise signal. Further, the preprocessing in the first step includes sequentially performing trending processing, band-pass filte