Magnetotelluric denoising method and system based on hybrid model

Abstract

The invention belongs to the technical field of magnetotelluric denoising, and particularly discloses a magnetotelluric denoising method and system based on a hybrid model. The method comprises the following steps: decomposing a magnetotelluric time sequence signal into a noise-containing contour modal signal and a noise-free contour modal signal by adopting a variational mode decomposition method; decoding the noise-containing contour modal signal, inputting the low-dimension vector set into an LSTM model, obtaining the position of the noise signal, and encoding the position; adopting a bidirectional long-short time memory network to cut off the position of the noise signal of the magnetotelluric time sequence signal, and performing bidirectional prediction on the cut time sequence signal to obtain the time sequence signal without noise at the noise position; and based on the noise-free contour modal signal and the noise-free time sequence signal at the noise position, reconstructing a noise-free magnetotelluric time sequence signal by using VMD. According to the invention, the low-loss denoising of the magnetotelluric signal can be realized, and the influence on a real signal is reduced.

Description

Technical Field

[0001] The present application belongs to the technical field of magnetotelluric denoising, and more specifically, to a magnetotelluric denoising method and system based on a hybrid model. Background Art

[0002] Magnetotelluric (MT) signals are susceptible to noise, and noise limits the accuracy of magnetotelluric restoration of the real underground structure, which has an adverse effect on the subsequent data interpretation. Nowadays, there are many magnetotelluric denoising methods, which can be roughly divided into two categories: traditional denoising methods and artificial intelligence denoising methods. Traditional denoising methods have poor performance in terms of large-scale data computing efficiency and denoising effect, and lack of adaptability, which is not conducive to the fine denoising of MT signals and has been gradually eliminated. Artificial intelligence denoising methods will cause overfitting to a certain extent, eliminate effective signals in the data, and cannot effectively remove noise. Therefore, it is crucial not to affect the effective signal during the denoising process. Summary of the invention

[0003] In view of the defects of the prior art, the purpose of the present application is to provide a magnetotelluric denoising method and system based on a hybrid model, aiming to solve the problem that the magnetotelluric denoising method based on neural network is prone to overfitting, which affects the effective signal during the denoising process.

[0004] To achieve the above objectives, in a first aspect, the present application provides a magnetotelluric denoising method based on a hybrid model, comprising the following steps: Step S1: using a variational modal decomposition method to perform modal decomposition on the magnetotelluric time series signal in the frequency domain, and decomposing the modal signal containing noise and the modal signal containing no noise; Step S2: Decode the noisy contour modal signal, input the obtained low-dimensional vector set into the LSTM model, obtain the location of the noise signal, and encode the noise signal; Step S3: Based on the noise signal and the location of the noise signal obtained in step S2, a bidirectional long short-term memory network is used to truncate the magnetotelluric time series signal at the location of the noise signal, and a bidirectional prediction is made on the truncated time series signal to obtain a noise-free time series signal at the noise location; Step S4: Based on the noise-free contour modal signal obtained in step S1 and the noise-free time series signal at the noise position obtained in step S3, VMD is used to reconstruct the noise-free magnetotelluric time series signal.

[0005] Further preferably, the modal component k and the quadratic penalty factor α in the variational modal decomposition method of step S1 are determined, and the magnetotelluric time series signal is modally decomposed in the frequency domain; wherein the modal component k determines the number of layers of VMD decomposition, and the quadratic penalty factor affects the bandwidth and center frequency of the modal component obtained by decomposition.

[0006] Further preferably, the noise signal includes: triangular noise, pulse noise, step noise and square wave noise.

[0007] Further preferably, step S2 specifically includes the following steps: Step S2.1: input the noisy contour modal signal obtained in step S1 into the VAE for decoding, and learn the latent variables in the noisy contour modal signal; Step S2.2: taking the noisy contour modal signal containing latent variables as the input of the LSTM model, detecting the noise position of the noisy contour modal signal, and marking the noise position information; Step S2.3: Use VAE to encode the noisy contour modal signal marked with noise position information to obtain a noisy contour modal signal containing noise position information.

[0008] In a second aspect, the present application provides a hybrid model-based magnetotelluric denoising system, comprising: The signal mode decomposition module is used to perform mode decomposition of the magnetotelluric time series signal in the frequency domain by using the variational mode decomposition method, and decompose the noise-containing contour mode signal and the noise-free contour mode signal; The noise signal localization module is used to decode the noisy contour modal signal, input the obtained low-dimensional vector set into the LSTM model, obtain the location of the noise signal, and encode the noise signal; A denoising module is used to truncate the location of the noise signal of the magnetotelluric time series signal based on the noise signal and the location of the noise signal, and to predict the truncated time series signal in both directions to obtain a noise-free time series signal at the noise location by using a bidirectional long short-term memory network; The signal reconstruction module is used to reconstruct the noise-free magnetotelluric time series signal using VMD based on the noise-free contour modal signal and the noise-free time series signal at the noise position.

[0009] Further preferably, the signal modal decomposition module is used to determine the modal component k and the quadratic penalty factor α in the variational modal decomposition method, and perform modal decomposition on the magnetotelluric time series signal in the frequency domain; wherein the modal component k determines the number of layers of VMD decomposition, and the quadratic penalty factor α affects the bandwidth and center frequency of the modal component obtained by decomposition.

[0010] Further preferably, the noise signal includes: triangular noise, pulse noise, step noise and square wave noise.

[0011] Further preferably, the noise signal positioning module includes: A decoding unit, used for inputting the noisy contour modal signal into the VAE for decoding, and learning the latent variables in the noisy contour modal signal; A noise detection unit is used to use the noisy contour modal signal containing latent variables as the input of the LSTM model, detect the noise position of the noisy contour modal signal, and mark the noise position signal; The encoding unit is used to encode the noisy contour modal signal marked with the noise position signal using VAE to obtain the noisy contour modal signal containing noise position information.

[0012] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory, wherein when the programs stored in the memory are executed, the processor is used to execute the method described in the first aspect or any further preferred aspect of the first aspect.

[0013] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, which, when executed on a processor, enables the processor to execute the method described in the first aspect or any further preferred aspect of the first aspect.

[0014] In a fifth aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method described in the first aspect or any further preferred aspect of the first aspect.

[0015] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0016] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the prior art: The present application provides a magnetotelluric denoising method based on a hybrid model, which uses a modal decomposition method to decompose magnetotelluric signals into different modes, that is, it can decompose noisy contour modal signals and noise-free contour modal signals, thereby achieving low-loss denoising of magnetotelluric signals, selecting noisy signals for denoising, and reducing the impact on real signals.

[0017] The present application provides a magnetotelluric denoising method based on a hybrid model, which uses a hybrid model (VAE-LSTM model) for anomaly detection, reduces the dimension of noisy contour modal signal data by decoding, and uses a convolutional neural network model (LSTM model) to analyze the reduced dimensionality data, thereby identifying the noise location more accurately and quickly.

[0018] The present application provides a magnetotelluric denoising method based on a hybrid model, which uses an improved neural network model (BiLSTM model) for bidirectional time series prediction, thereby obtaining more comprehensive data analysis and prediction, and can batch process time series of different lengths to generate prediction results in the most efficient way. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a magnetotelluric denoising method based on a hybrid model provided in an embodiment of the present application; Figure 2 is a simulated mixed noise time series signal diagram provided in an embodiment of the present application; Figure 3 : is a VMD decomposition result diagram for k=2, 3, 4, 5 provided in the embodiments of the present application; Figure 4 α=100, 500, 1000, 2000 are the time series components and spectrograms provided in the embodiments of the present application; Figure 5 is a diagram of VMD decomposition results of a mixed noise time series signal provided in an embodiment of the present application; Figure 6 is a detection result diagram of the anomaly detection model provided in an embodiment of the present application; Figure 7 is a BiLSTM structure diagram provided in an embodiment of the present application; Figure 8 is a denoising result diagram provided in an embodiment of the present application; Fig. 9 is a comparison chart of denoising effects of different methods provided in the embodiments of the present application; Fig.10 It is an impedance and phase comparison diagram of a simulated mixed noise signal using different denoising methods provided in an embodiment of the present application; Fig.11 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0021] The terms "first", "second" and the like in the description and claims of the present application are used to distinguish different objects rather than to describe a specific order of the objects.

[0022] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0023] In the description of the embodiments of the present application, unless otherwise specified, “plurality” means two or more than two.

[0024] The present application provides a magnetotelluric denoising method based on a hybrid model. First, the magnetotelluric time series signal is decomposed into different modes (noise-containing contour boundary modes and noise-free contour boundary modes) according to different frequencies, and a hybrid anomaly detection model is used to perform anomaly detection on the noisy contour boundary mode. The dimension of the calculation model is reduced by decoding, the robustness is enhanced, and the effectiveness of anomaly detection is further improved. At the same time, the impact on the effective signal during the denoising process can be effectively reduced. Then, the time series prediction model is set as a bidirectional neural network model, which can process large-scale time series such as magnetotelluric time series signals and can effectively realize the prediction of high-precision time series.

[0025] The hybrid model-based magnetotelluric denoising method provided in the present application involves a time series signal decomposition model, a time series signal anomaly detection model and a time series prediction model; the magnetotelluric denoising method specifically includes the following steps: Step S1: Using the variational mode decomposition (VMD) method, the magnetotelluric time series signal is decomposed into multiple intrinsic modes (IMFs) to achieve the frequency domain division of the signal, so that each mode has local characteristics in the time domain and is independent of each other in the frequency domain; wherein the time series signal decomposition model adopts the variational mode decomposition method; Step S2: Decode the noisy contour modal signal through the decoder in the VAE to obtain a low-dimensional vector set of the noisy contour modal signal; input the low-dimensional vector set into the LSTM model, determine the location of the anomaly through the LSTM model and mark it, and finally encode the marked anomaly data through the VAE; among them, the time series anomaly detection model combines the variational auto-encoders (VAE) and the long short-term memory network (LSTM); the anomaly here is noise; Step S3: Using a bidirectional long short-term memory (Bi-LSTM) model, when processing magnetotelluric signals, the time series signal with a determined noise position can be noise-truncated, and bidirectional prediction can be performed directly on the truncated time series signal; this improved model not only maintains the accuracy of time prediction of the traditional LSTM model, but also can better adapt to magnetotelluric time series signals; wherein, the time series prediction model is a bidirectional long short-term memory network, rather than a traditional unidirectional long short-term memory network.

[0026] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0027] Example 1 The hybrid model-based magnetotelluric denoising method provided in this application is as follows: Figure 1 As shown, the specific steps include: Step S1: First, add mixed noises such as triangle noise, pulse noise, step noise and square wave noise to the MT pure signal to obtain Figure 2 The figure shows a simulated mixed noise signal containing various noises. The simulated signal is a wide-band nonlinear signal with no obvious features in the frequency domain. The modal component K determines the number of layers of VMD decomposition. Too large or too small a K will cause the modal component to contain more noise, affecting subsequent signal processing. The selection of the quadratic penalty factor α will affect the bandwidth and center frequency of the modal component obtained by decomposition. K and α are determined by simulating time series signals. α is fixed to 2000, and K is selected to 2, 3, 4, and 5 to verify the effect of decomposing the low-frequency component containing noise contour information. The decomposition effect is shown in the figure. Figure 3 shown; from Figure 3 It can be seen that when the k value is selected as 2, the low-frequency component can clearly reflect the noise profile information and can also reduce the redundant components generated when k is greater than 2; to verify whether α affects the low-frequency noise profile component, keep k=2, select α as 100, 500, 1000, 2000, and obtain the time series and spectrum of the two components as follows Figure 4 shown; from Figure 4 It can be seen that when k=2 and α=100, the low-frequency noise contour component has the clearest contour and the most complete noise contour. Therefore, it can be determined that when the modal component K in the VMD decomposition is 2 and the secondary penalty factor α is 100, the magnetotelluric time series signal can be well decomposed. Finally, VMD decomposition is used for processing Figure 2 The simulated mixed noise signal in is obtained as Figure 5 The VMD decomposition result of Figure 5 Various types of noise can be well decomposed and low-frequency components with clear outlines can be obtained.

[0028] More specifically, the magnetotelluric time series simulation signal is Figure 2 The mixed noisy simulated signal shown contains various noise signals.

[0029] from Figure 5 It can be seen that IMF1 is the high-frequency component of the decomposed profile without noise, and IMF2 is the low-frequency component of the decomposed profile with noise.

[0030] Step S2: Input the low-frequency component containing the noise contour obtained in step S1 into VAE for decoding, learn the latent variables in the low-frequency component, and store the latent variables. Then, use the low-frequency component data containing the latent variables as the input of the LSTM model, detect the noise position of the low-frequency component, and mark the abnormal position information; finally, use VAE to encode the low-frequency component marked with abnormal position information to obtain the low-frequency component containing the abnormal position. The abnormal detection result is as follows: Figure 6 shown; from Figure 6 It can be seen that when the distance between two abnormal locations is small, the anomaly detection model cannot obviously detect a single anomaly, but can only detect the approximate range of the anomaly.

[0031] The LSTM training parameters are as follows: solver is adam; number of trainings: 250; gradient threshold: 1; initial learning rate: 0.01; after 100 rounds of training, multiply by 0.2 to reduce the learning rate.

[0032] Step S3: First, use the abnormal low-frequency component detected in step S2 to input the BiLSTM model. The BiLSTM model is an improved network of the long short-term memory network. It processes data in the forward and reverse directions through two independent LSTMs to obtain more comprehensive data analysis and prediction. The BiLSTM structure is as follows: Figure 7 As shown; the BiLSTM model is used to predict the noise position of the low-frequency component, and then the high-frequency component that is basically noise-free obtained by decomposition in step S1 and the low-frequency component with noise contour after the noise position prediction are used to reconstruct the MT simulation time series signal using VMD, as shown Figure 8 As shown in the figure, the denoised signal is very close to the original clean signal, but some high-amplitude pulse signals are retained at some high-amplitude noise locations, and the pulse signals have little impact on the signal. These residual pulse signals have little impact on the overall signal, indicating that the proposed hybrid model denoising algorithm can effectively denoise MT time series data.

[0033] Step S4: The denoised time series obtained by this algorithm is compared with LSTM, KSVD, wavelet transform and maximum-minimum fair algorithm. The comparison of the denoising effect of the time series is shown in Figure 9. Then the impedance and phase analysis of the time series signal is performed to obtain Fig.10 The phase impedance comparison results are shown.

[0034] Example 2 The present application provides a hybrid model-based magnetotelluric denoising system, comprising: The signal mode decomposition module is used to perform mode decomposition of the magnetotelluric time series signal in the frequency domain by using the variational mode decomposition method, and decompose the noise-containing contour mode signal and the noise-free contour mode signal; The noise signal localization module is used to decode the noisy contour modal signal, input the obtained low-dimensional vector set into the LSTM model, obtain the location of the noise signal, and encode the noise signal; A denoising module is used to truncate the location of the noise signal of the magnetotelluric time series signal based on the noise signal and the location of the noise signal, and to predict the truncated time series signal in both directions to obtain a noise-free time series signal at the noise location by using a bidirectional long short-term memory network; The signal reconstruction module is used to reconstruct the noise-free magnetotelluric time series signal using VMD based on the noise-free contour modal signal and the noise-free time series signal at the noise position.

[0035] Further preferably, the signal modal decomposition module is used to determine the modal component k and the quadratic penalty factor α in the variational modal decomposition method, and perform modal decomposition on the magnetotelluric time series signal in the frequency domain; wherein the modal component k determines the number of layers of VMD decomposition, and the quadratic penalty factor α affects the bandwidth and center frequency of the modal component obtained by decomposition.

[0036] Further preferably, the noise signal includes: triangular noise, pulse noise, step noise and square wave noise.

[0037] Further preferably, the noise signal positioning module includes: A decoding unit, used for inputting the noisy contour modal signal into the VAE for decoding, and learning the latent variables in the noisy contour modal signal; A noise detection unit is used to use the noisy contour modal signal containing latent variables as the input of the LSTM model, detect the noise position of the noisy contour modal signal, and mark the noise position signal; The encoding unit is used to encode the noisy contour modal signal marked with the noise position signal using VAE to obtain the noisy contour modal signal containing noise position information.

[0038] In summary, compared with the prior art, the present application has the following advantages: from Fig. 9 It can be clearly seen that this algorithm can perform MT signal denoising very well. Compared with the existing denoising methods, the hybrid model denoising method can detect the location of anomalies very well and remove anomalies well, and the denoising effect is better than the existing denoising methods. The impedance and phase analysis of the time series obtained by denoising various methods is carried out. Fig.10 It can be seen that the impedance and phase of the clean signal show smooth and continuous characteristics. After adding noise, the impedance and phase show different degrees of anomalies below 1 Hz. The results of denoising using MMF, KSVD and wavelet methods show limited improvement, and there are still obvious anomalies in the impedance and phase curves. Although the impedance and phase curves produced by the LSTM method are relatively close to the clean signal, there are still some anomalies below 1 Hz. In contrast, the curves produced by the hybrid model method are significantly improved compared with the noisy signal curves and are very similar to the curves of the clean signal. This shows that the proposed algorithm has excellent denoising performance for MT low-frequency signals.

[0039] This application uses a modal decomposition method to decompose the magnetotelluric signal into different modes, achieve low-loss denoising of the magnetotelluric signal, and select noisy signals for denoising to reduce the impact on the real signal; This application uses a hybrid model for anomaly detection, reduces the dimension of the data through decoding, and uses a convolutional neural network model to analyze the reduced-dimensional data, so as to identify the noise location more accurately and quickly.

[0040] This application uses an improved neural network model for bidirectional time series prediction, which can process time series of different lengths in batches to efficiently generate prediction results.

[0041] It can be understood that the detailed functional implementation of each of the above-mentioned units / modules can be found in the introduction of the aforementioned method embodiment, and will not be repeated here.

[0042] It should be understood that the above-mentioned system is used to execute the methods in the above-mentioned embodiments. The implementation principles and technical effects of the corresponding program modules in the system are similar to those described in the above-mentioned methods. The working process of the system can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0043] Based on the method in the above embodiment, Fig.11As shown, an embodiment of the present application provides an electronic device, which may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call the logic instructions in the memory 830 to execute the method in the above embodiment.

[0044] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0045] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0046] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0047] It is understandable that the processor in the embodiment of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0048] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0049] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0050] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0051] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A magnetotelluric denoising method based on a hybrid model, characterized in that: The following steps are involved: Step S1: using a variational modal decomposition method to perform modal decomposition on the magnetotelluric time series signal in the frequency domain, and decomposing the modal signal containing noise and the modal signal containing no noise; Step S2: Decode the noisy contour modal signal, input the obtained low-dimensional vector set into the LSTM model, obtain the location of the noise signal, and encode the noise signal; Step S3: Based on the noise signal and the position of the noise signal obtained in step S2, a bidirectional long short-term memory network is used to truncate the noise position of the magnetotelluric time series signal, and a bidirectional prediction is made on the truncated time series signal to obtain a noise-free time series signal at the noise position; Step S4: Based on the noise-free contour modal signal obtained in step S1 and the noise-free time series signal at the noise position obtained in step S3, VMD is used to reconstruct the noise-free magnetotelluric time series signal.

2. The magnetotelluric denoising method according to claim 1, characterized in that: Determine the modal component k and the quadratic penalty factor α in the variational modal decomposition method in step S1, and perform modal decomposition on the magnetotelluric time series signal in the frequency domain; wherein the modal component k determines the number of layers of the variational modal decomposition, and the quadratic penalty factor α affects the bandwidth and center frequency of the modal component obtained by decomposition.

3. The magnetotelluric denoising method according to claim 1 or 2, characterized in that: Noise signals include: triangle noise, pulse noise, step noise and square wave noise.

4. The magnetotelluric denoising method according to claim 1 or 2, characterized in that: Step S2 specifically includes the following steps: Step S2.1: input the noisy contour modal signal obtained in step S1 into the VAE for decoding, and learn the latent variables in the noisy contour modal signal; Step S2.2: taking the noisy contour modal signal containing latent variables as the input of the LSTM model, detecting the noise position of the noisy contour modal signal, and marking the noise position information; Step S2.3: Use VAE to encode the noisy contour modal signal marked with noise position information to obtain a noisy contour modal signal containing noise position information.

5. A magnetotelluric denoising system based on a hybrid model, characterized in that: include: The signal mode decomposition module is used to perform mode decomposition of the magnetotelluric time series signal in the frequency domain by using the variational mode decomposition method, and decompose the noise-containing contour mode signal and the noise-free contour mode signal; The noise signal localization module is used to decode the noisy contour modal signal, input the obtained low-dimensional vector set into the LSTM model, obtain the location of the noise signal, and encode the noise signal; A denoising module is used to truncate the location of the noise signal of the magnetotelluric time series signal based on the noise signal and the location of the noise signal, and to predict the truncated time series signal in both directions to obtain a noise-free time series signal at the noise location by using a bidirectional long short-term memory network; The signal reconstruction module is used to reconstruct the noise-free magnetotelluric time series signal using VMD based on the noise-free contour modal signal and the noise-free time series signal at the noise position.

6. The magnetotelluric denoising system according to claim 5, characterized in that: The signal modal decomposition module is used to determine the modal component k and the quadratic penalty factor α in the variational modal decomposition method, and to perform modal decomposition of the magnetotelluric time series signal in the frequency domain; among them, the modal component k determines the number of layers of VMD decomposition, and the quadratic penalty factor α affects the bandwidth and center frequency of the modal component obtained by decomposition.

7. The magnetotelluric denoising system according to claim 5 or 6, characterized in that: Noise signals include: triangle noise, pulse noise, step noise and square wave noise.

8. The magnetotelluric denoising system according to claim 5, characterized in that: The noise signal positioning module includes: A decoding unit, used for inputting the noisy contour modal signal into the VAE for decoding, and learning the latent variables in the noisy contour modal signal; A noise detection unit is used to use the noisy contour modal signal containing latent variables as the input of the LSTM model, detect the noise position of the noisy contour modal signal, and mark the noise position signal; The encoding unit is used to encode the noisy contour modal signal marked with the noise position signal using VAE to obtain the noisy contour modal signal containing noise position information.

Images