Reflection echo signal processing method and system based on double-domain denoising model, and medium

Abstract

The invention discloses a reflection echo signal processing method and system based on a double-domain denoising model and a medium. Relates to the technical field of oil surface depth detection of oil wells. A double-domain denoising model is completed by training sound wave reflection echo signals, and the complementation of time domain features and frequency domain features can be met by processing noise in the noise removal process; in the training process of the double-domain denoising model, the proportion of time domain denoising and frequency domain denoising of the reflection echo in the double-domain denoising model is determined by measuring the time domain feature and the frequency domain feature of the input reflection echo signal; based on time domain and frequency domain features of an input reflection echo signal, cascade weights of double domains in a double-domain denoising model are dynamically changed, and a method for measuring time domain features and frequency domain features of the input reflection echo signal based on an energy ratio before and after time domain feature extraction and an energy ratio before and after frequency domain feature extraction is provided. Therefore, the trained model can better adapt to reflection echo signals with various different characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of oil well oil level depth detection, and in particular to a reflection echo signal processing method, system and medium based on a dual-domain denoising model. Background Art

[0002] The detection of oil well oil surface depth has always been a key issue in oil production. The oil surface depth can directly reflect the supply capacity of the underground oil layer, determine the sinking depth of the oil pump, the oil layer pressure, analyze the abnormal reasons for energy attenuation, etc., so as to reasonably arrange the oil production strategy and maximize the oil well production rate. Therefore, the detection of oil well oil surface depth is extremely important in oilfield development. At present, the most commonly used method for measuring the oil surface depth of oil wells is the acoustic logging method. In the existing technology, the reception of the logging acoustic wave reflection echo is usually achieved through distributed optical fiber acoustic sensing (DAS). However, due to the complex downhole environment, the collected sound reflection echo is sandwiched with various environmental noises and mechanical noises, resulting in large calculation errors. Therefore, the use of acoustic denoising technology to denoise the reflected echo is very important to improve the accuracy of oil surface depth detection. However, the denoising methods used for reflected echoes in the prior art are mainly embodied in linear transformation, wavelet transformation, spectral subtraction, blind source separation, and sparse decomposition. These traditional denoising methods either simply remove noise from the reflected echo from the time domain or simply remove noise from the reflected echo from the frequency domain. Both methods fail to comprehensively and reasonably consider the time domain characteristics and frequency domain characteristics of the noise, resulting in an insignificant noise removal effect. They are relatively suitable for constant simple noise, but not suitable for complex random noise in the well. As a result, the oil level depth calculation based on the reflected echo has low accuracy and cannot accurately reflect the true oil level depth, and the measurement is of little significance. Summary of the invention

[0003] Aiming at the technical problem that the denoising means used for reflected echo in the prior art are relatively suitable for constant simple noise but not for complex random noise in the well, the oil level depth calculation based on reflected echo has low accuracy and cannot accurately reflect the real oil level depth; the purpose of the present invention is to provide a reflected echo signal processing method, system and medium based on a dual-domain denoising model, by performing dual noise removal in the time domain and frequency domain on the acoustic wave reflected echo signal, and training a dual-domain denoising model, so that the processing of noise in the noise removal process can meet the complementarity of time domain characteristics and frequency domain characteristics, the noise removal effect is more obvious, and the dual-domain parallel processing makes the denoising speed faster.

[0004] The present invention is achieved through the following technical solutions:

[0005] This solution provides a reflection echo signal processing method based on a dual-domain denoising model, including:

[0006] The time domain feature extraction of the reflected echo S(t) in the training set is performed to obtain the time domain feature X1(t) of the useful signal, and the short-time Fourier transform of the reflected echo S(t) is performed to obtain the spectrum signal S m (f);

[0007] For the spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t);

[0008] Calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t), the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2;

[0009] Align the time domain feature X1(t) with the time domain feature X2(t) according to the time series information to obtain the aligned time domain feature X1(t) ′ And the time domain feature X2(t) ′ ;

[0010] Based on the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model;

[0011] The target deep well DAS reflection echo is obtained and input into the dual-domain denoising model for denoising.

[0012] Working principle of this scheme: The reflected echo signal processing method based on the dual-domain denoising model proposed in this scheme performs dual noise removal in the time domain and frequency domain on the sound wave reflected echo signal collected by DAS, and completes the training of the dual-domain denoising model, so that the processing of noise in the noise removal process can meet the complementarity of time domain characteristics and frequency domain characteristics, the noise removal effect is more obvious, and the dual-domain parallel processing makes the denoising speed faster; and, in the training process of the dual-domain denoising model, the proportion of time domain denoising and frequency domain denoising of the sound wave reflected echo in the dual-domain denoising model is determined based on the energy ratio before and after time domain feature extraction and the energy ratio before and after frequency domain feature extraction, so that the trained dual-domain denoising model can better adapt to reflection echo signals with various different characteristics.

[0013] A further optimization scheme is that the calculation method of the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t) includes:

[0014]

[0015] Among them, P[X1(t)] is the energy of the time domain feature X1(t), and P[S(t)] is the energy of the reflected echo S(t).

[0016] A further optimization scheme is that the frequency domain feature X m2 (f) and the spectrum signal S m The calculation method of the energy ratio r2 of (f) includes:

[0017]

[0018] Among them, P[X m2 (f)] is the frequency domain feature X m2 (f) energy, P[S m (f)] is the spectrum signal S m (f) Energy.

[0019] A further optimization scheme is that the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; including methods:

[0020] The time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 are added element by element to obtain the output result of the dual-domain denoising model

[0021] A further optimization scheme is that the loss function in the dual-domain denoising model training process includes:

[0022]

[0023] Where ω1, ω2 and ω3 represent weight coefficients, ‖*‖ 1 represents 1-norm calculation, ‖*‖ 2 represents the 2-norm calculation, Represents the output of the dual-domain denoising model The sequence obtained by sampling n times; N-1 represents the output result The total number of sampling times.

[0024] A further optimization scheme is that the energy includes signal strength.

[0025] The present solution also provides a reflection echo signal processing system based on a dual-domain denoising model, wherein the computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described above; the system comprises:

[0026] The first processing module is used to extract the time domain features of the reflection echo S(t) in the training set to obtain the time domain features X1(t) of the useful signal, and to perform short-time Fourier transform on the reflection echo S(t) to obtain the spectrum signal S m (f);

[0027] The second processing module is used to process the spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t);

[0028] The weight calculation module is used to calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t), the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2;

[0029] The feature alignment module is used to align the time domain feature X1(t) and the time domain feature X2(t) according to the time sequence information to obtain the aligned time domain feature X1(t). ′ And the time domain feature X2(t) ′ ;

[0030] Output module, used to generate the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model;

[0031] The control module is used to obtain the target deep well DAS reflection echo, and input the deep well DAS reflection echo into the dual-domain denoising model for denoising processing.

[0032] A further optimization scheme is that the loss function in the dual-domain denoising model training process is determined as: L = ω1*Xt-X1(t)1+ω2*Sm(f)-Xm2(f)2+ω3*n=0N-1[Xn+1-Xn2]; where ω1, ω2 and ω3 are weight coefficients, ‖*‖ 1 is the 1-norm calculation, ‖*‖ 2is the 2-norm calculation, The output result of the dual-domain denoising model The sequence obtained by sampling N times.

[0033] A further optimization solution is that the energy ratio includes a signal intensity ratio.

[0034] The present solution also provides a computer-readable medium having a computer program stored thereon, and the computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described above.

[0035] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0036] The reflected echo signal processing method, system and medium based on the dual-domain denoising model provided by the present invention can perform denoising processing in parallel through the time-frequency dual-domains, and train the dual-domain denoising model by performing dual noise removal in the time domain and the frequency domain on the input reflected echo signal, so that the processing of noise in the noise removal process can meet the complementarity of the time domain characteristics and the frequency domain characteristics, the noise removal effect is more obvious, and the dual-domain parallel processing makes the denoising speed faster; and in the dual-domain denoising model training process, the time domain denoising and the frequency domain denoising are not simply cascaded, but the input is measured by The time domain characteristics and frequency domain characteristics of the input reflected echo signal are determined, the proportion of time domain denoising and frequency domain denoising of the reflected echo in the dual-domain denoising model is determined, the cascade weights of the dual domains in the dual-domain denoising model are dynamically changed based on the time domain and frequency domain characteristics of the input reflected echo signal, and based on this, a method is proposed to measure the time domain characteristics and frequency domain characteristics of the input reflected echo signal based on the energy ratio before and after time domain feature extraction and the energy ratio before and after frequency domain feature extraction, so that the trained dual-domain denoising model can better adapt to reflected echo signals with various different characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0038] Figure 1 It is a flowchart of a reflection echo signal processing method based on a dual-domain denoising model;

[0039] Figure 2 It is a schematic diagram of a reflection echo signal processing system based on a dual-domain denoising model; Figure 3Schematic diagram of the working principle of the reflection echo signal processing system based on the dual-domain denoising model. DETAILED DESCRIPTION

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0041] The denoising method used for reflected echo in the prior art is relatively suitable for constant simple noise, but not suitable for complex random noise in the well, which makes the oil level depth calculation based on reflected echo low in accuracy and cannot accurately reflect the real oil level depth. In view of this, the present invention provides the following embodiments to solve the above technical problems.

[0042] Embodiment 1: This embodiment provides a reflection echo signal processing method based on a dual-domain denoising model, such as Figure 1 As shown, including:

[0043] S1, extract the time domain features of the reflection echo S(t) in the training set to obtain the time domain features X1(t) of the useful signal, and perform short-time Fourier transform on the reflection echo S(t) to obtain the spectrum signal S m (f);

[0044] Wherein, S(t)=X(t)+N(t), X(t) is the useful signal in the reflected echo, and N(t) is the noise signal in the reflected echo;

[0045] Preferably, in the training set, each reflected echo is a combination of a pure acoustic signal and various noises at different signal-to-noise ratios;

[0046] Where S m (f) = X m (f)+N m (f), X m (f) is the spectrum of the useful signal in the reflected echo, N m (f) is the spectrum of the noise signal in the reflected echo;

[0047] S2, spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t);

[0048] S3, calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t):

[0049]

[0050] Among them, P[X1(t)] is the energy of the time domain feature X1(t), and P[S(t)] is the energy of the reflected echo S(t).

[0051] Calculate the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2:

[0052]

[0053] Among them, P[X m2 (f)] is the frequency domain feature X m2 (f) energy, P[S m (f)] is the spectrum signal S m (f) energy; the energy includes signal strength;

[0054] Preferably, the input and output energy can be measured by a power meter, an oscilloscope or other scientific instrument in the prior art, or the energy intensity of the input and output signals can be characterized based on their intensity;

[0055] Specifically, the ratio of the output signal energy of the feature extraction module to the input signal energy can well measure the time domain characteristics and frequency domain characteristics of the reflected echo signal, determine the proportion of the time domain denoising and frequency domain denoising of the reflected echo in the dual-domain denoising model, and dynamically change the dual-domain cascade weights in the dual-domain denoising model based on the time domain and frequency domain characteristics of the input reflected echo signal, so that the trained dual-domain denoising model can adapt to the denoising of more reflected echo signals with different time and frequency domain characteristics;

[0056] S4, align the time domain feature X1(t) and the time domain feature X2(t) according to the time sequence information to obtain the aligned time domain feature X1(t) ′ And the time domain feature X2(t) ′ ;

[0057] Since the feature distribution of the reflected echo signal in the time domain and frequency domain is different, and the method of feature extraction and the amount of features that can be extracted in the time domain and frequency domain are also different, it is difficult to directly add the features extracted in the dual domains element by element. Adding the elements one by one under the premise that the dual domain features are not aligned will lead to feature misalignment, which will further lead to output signal misalignment. Not only will it fail to achieve better denoising results, but it will also introduce new mutual interference. Therefore, in this step, the time domain features obtained in the dual domains are aligned according to the timing information of the signal. Avoiding the existence of feature misalignment is an indispensable step in the cascade output of dual domain denoising.

[0058] S6, based on the time domain feature X1(t)′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model; specifically including the method:

[0059] The time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 are added element by element to obtain the output result of the dual-domain denoising model

[0060]

[0061] The loss function in the dual-domain denoising model training process includes:

[0062]

[0063] Where ω1, ω2 and ω3 represent weight coefficients, ‖*‖ 1 represents 1-norm calculation, ‖*‖ 2 represents the 2-norm calculation, Represents the output of the dual-domain denoising model The sequence obtained by sampling n times; N-1 represents the output result The total number of samplings. The parameters of the dual-domain denoising model are continuously adjusted through the optimization algorithm to minimize the loss or to complete the training of the dual-domain denoising model when the predetermined number of training times is reached; the method for determining the loss function can also adopt other methods known in the prior art, which are not specifically limited here.

[0064] S6, obtaining the target deep well DAS reflection echo, and inputting the deep well DAS reflection echo into the dual-domain denoising model for denoising processing.

[0065] Specifically, the specific determination process of the oil surface depth may include the following process: the acoustic wave reflection echo signal collected from the deep well by DAS includes a high-frequency first reflection echo signal and a low-frequency second reflection echo signal. After receiving the acoustic wave reflection echo signal, the first reflection echo signal and the second reflection echo signal are separated by a set filter, and then the first reflection echo signal and the second reflection echo signal are input into the above-mentioned dual-domain denoising model for noise removal. M peak points are continuously selected from the denoised second reflection echo, and the average time T between two adjacent peak points is calculated. Then, the average speed of the acoustic wave in the oil well v=2L / T can be calculated, where L is the length of a single section of oil pipe in the oil well; the maximum amplitude average value of multiple segments is determined from the denoised first reflection echo as A1, and the maximum amplitude average value of multiple segments is determined from the denoised second reflection echo as A2, thereby determining the oil surface position A=(A1+A2) / 2, given the sampling frequency f, the time when the oil surface appears T'=A / f can be determined, and the above The formula can determine the oil level depth H=v*T' / 2 in the oil well; thus, it can be determined that in order to calculate the depth of the oil level in the deep well, it is necessary to rely on the acoustic reflection echo signal of the acoustic detection, and thus the denoising of the acoustic reflection echo is very important for the detection accuracy of the oil level depth in the deep well. The dual-domain denoising model proposed in the present invention can take into account the time domain characteristics and frequency domain characteristics of the acoustic reflection echo signal, and realize the denoising in parallel in the dual domains, so as to realize the complementarity of the time domain characteristics and frequency domain characteristics of the acoustic reflection echo signal, and proposes to use the dual-domain The energy ratio before and after feature extraction is used to measure the proportion of the time domain characteristics and frequency domain characteristics of the acoustic wave reflection echo signal, thereby determining the cascade weights of the time domain denoising and frequency domain denoising of the reflection echo in the dual-domain denoising model. The cascade weights of the dual domains in the dual-domain denoising model are dynamically changed based on the time domain and frequency domain characteristics of the input reflection echo signal, so that the trained dual-domain denoising model can better adapt to reflection echo signals with various different characteristics, better improve the noise removal effect, and thus assist in improving the accuracy of oil level depth detection.

[0066] Embodiment 2: This embodiment provides a reflection echo signal processing system 200 based on a dual-domain denoising model, such as Figure 2 and Figure 3 As shown, the computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described in Example 1; the system includes:

[0067] The first processing module includes a time domain feature extraction module and a short-time Fourier transform module. The time domain feature extraction module is used to extract the time domain features of the reflection echo S(t) in the training set to obtain the time domain features X1(t) of the useful signal. The short-time Fourier transform module is used to perform short-time Fourier transform on the reflection echo S(t) to obtain the spectrum signal S m (f);

[0068] The second processing module includes a frequency domain feature extraction module and an inverse short-time Fourier transform module. The frequency domain feature extraction module is used to extract the spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), the inverse short-time Fourier transform module is used to transform the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t);

[0069] The weight calculation module includes a time domain weight calculation module and a frequency domain weight calculation module. The time domain weight calculation module is used to calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t). The frequency domain weight calculation module is used to calculate the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2;

[0070] Preferably, the input and output energy can be measured by a power meter, an oscilloscope or other scientific instrument in the prior art, or the energy intensity of the input and output signals can be characterized based on their intensity;

[0071] Specifically, the ratio of the output signal energy of the feature extraction module to the input signal energy can well measure the time domain characteristics and frequency domain characteristics of the reflected echo signal, determine the proportion of the time domain denoising and frequency domain denoising of the reflected echo in the dual-domain denoising model, and dynamically change the dual-domain cascade weights in the dual-domain denoising model based on the time domain and frequency domain characteristics of the input reflected echo signal, so that the trained dual-domain denoising model can adapt to the denoising of more reflected echo signals with different time and frequency domain characteristics;

[0072] The feature alignment module is used to align the time domain feature X1(t) and the time domain feature X2(t) according to the time sequence information to obtain the aligned time domain feature X1(t). ′ And the time domain feature X2(t) ′ ;

[0073] Output module, used to generate the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model;

[0074] The control module is used to obtain the target deep well DAS reflection echo, and input the deep well DAS reflection echo into the dual-domain denoising model for denoising processing.

[0075] Preferably, in the process of completing the training of the dual-domain denoising model, the loss function can be defined according to the difference between the denoised output of the acoustic wave reflection signal of the dual-domain of the model and the real acoustic wave signal and combined with the continuity of the acoustic wave sequence. The loss function is determined as: where ω1, ω2 and ω3 are weight coefficients, ‖*‖ 1 is the 1-norm calculation, ‖*‖ 2 is the 2-norm calculation, The output result of the dual-domain denoising model The sequence obtained by sampling N times.

[0076] The energy ratio includes a signal strength ratio.

[0077] The model parameters are continuously adjusted through the optimization algorithm to minimize the loss or determine that the training of the dual-domain denoising model is completed when the predetermined number of training times is reached; the method for determining the loss function can also adopt other methods known in the prior art, which are not specifically limited here.

[0078] Embodiment 3: This embodiment provides a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described in Embodiment 1; specifically, the following steps are performed:

[0079] S1, extract the time domain features of the reflection echo S(t) in the training set to obtain the time domain features X1(t) of the useful signal, and perform short-time Fourier transform on the reflection echo S(t) to obtain the spectrum signal S m (f);

[0080] S2, spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t);

[0081] S3, calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t), the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2;

[0082] S4, align the time domain feature X1(t) and the time domain feature X2(t) according to the time sequence information to obtain the aligned time domain feature X1(t) ′ And the time domain feature X2(t)′ ;

[0083] S5, based on the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model;

[0084] S6, obtaining the target deep well DAS reflection echo, and inputting the deep well DAS reflection echo into the dual-domain denoising model for denoising processing.

[0085] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A reflection echo signal processing method based on a dual-domain denoising model, characterized in that: include: The time domain feature extraction of the reflected echo S(t) in the training set is performed to obtain the time domain feature X1(t) of the useful signal, and the short-time Fourier transform of the reflected echo S(t) is performed to obtain the spectrum signal S m (f); For the spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t); Calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t), the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2; Align the time domain feature X1(t) with the time domain feature X2(t) according to the time series information to obtain the aligned time domain feature X1(t) ′ And the time domain feature X2(t) ′ ; Based on the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model; The target deep well DAS reflection echo is obtained and input into the dual-domain denoising model for denoising.

2. The method for processing reflected echo signals based on a dual-domain denoising model according to claim 1, characterized in that: The calculation method of the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t) includes: Among them, P[X1(t)] is the energy of the time domain feature X1(t), and P[S(t)] is the energy of the reflected echo S(t).

3. The method for processing reflected echo signals based on a dual-domain denoising model according to claim 1, characterized in that: The frequency domain feature X m2 (f) and the spectrum signal S m The calculation method of the energy ratio r2 of (f) includes: Among them, P[X m2 (f)] is the frequency domain feature X m2 (f) energy, P[S m (f)] is the spectrum signal S m (f) Energy.

4. The method for processing reflected echo signals based on a dual-domain denoising model according to claim 1, characterized in that: The time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; including method: The time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 are added element by element to obtain the output result of the dual-domain denoising model 5. The method for processing reflected echo signals based on a dual-domain denoising model according to claim 1, characterized in that: The loss function in the dual-domain denoising model training process includes: Where ω1, ω2 and ω3 represent weight coefficients, ‖*‖ 1 represents 1-norm calculation, ‖*‖ 2 represents the 2-norm calculation, Represents the output of the dual-domain denoising model The sequence obtained by sampling n times; N-1 represents the output result The total number of sampling times.

6. The method for processing reflected echo signals based on a dual-domain denoising model according to claim 2 or 3, characterized in that: The energy includes signal strength.

7. A reflection echo signal processing system based on a dual-domain denoising model, characterized in that: The computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described in any one of claims 1 to 6; the system includes: The first processing module is used to extract the time domain features of the reflection echo S(t) in the training set to obtain the time domain features X1(t) of the useful signal, and to perform short-time Fourier transform on the reflection echo S(t) to obtain the spectrum signal S m (f); The second processing module is used to process the spectrum signal S m (f) Extract the frequency domain features to obtain the frequency domain features X of the useful signal m2 (f), and the frequency domain feature X m2 (f) Perform inverse short-time Fourier transform to obtain the corresponding time domain feature X2(t); The weight calculation module is used to calculate the energy ratio r1 of the time domain feature X1(t) and the reflected echo S(t), the frequency domain feature X m2 (f) and the spectrum signal S m (f) Energy ratio r2; The feature alignment module is used to align the time domain feature X1(t) and the time domain feature X2(t) according to the time sequence information to obtain the aligned time domain feature X1(t). ′ And the time domain feature X2(t) ′ ; Output module, used to generate the time domain feature X1(t) ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to obtain the output result of the dual-domain denoising model; and based on the time domain features X1(t) of multiple echoes in the training set ′ , time domain features X2(t) ′ , energy ratio r1 and energy ratio r2 to train the dual-domain denoising model; The control module is used to obtain the target deep well DAS reflection echo, and input the deep well DAS reflection echo into the dual-domain denoising model for denoising processing.

8. The reflected echo signal processing system based on the dual-domain denoising model according to claim 7, characterized in that: The loss function in the dual-domain denoising model training process is determined as: L = ω1*‖X(t)-X1(t)‖ 1 +ω2*‖S m (f)-Xm2(f)2+ω3*n=0N-1[Xn+1-Xn2]; where ω1, ω2 and ω3 are weight coefficients, *1 is the 1-norm calculation, ‖*‖ 2 is the 2-norm calculation, The output result of the dual-domain denoising model The sequence obtained by sampling N times.

9. The reflected echo signal processing system based on the dual-domain denoising model according to claim 7, characterized in that: The energy ratio includes a signal strength ratio.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the reflection echo signal processing method based on the dual-domain denoising model as described in any one of claims 1 to 6.

Images