Ocean controllable source electromagnetic signal denoising method

Abstract

The invention discloses an ocean controllable source electromagnetic signal denoising method, and relates to the technical field of ocean geophysical exploration and signal processing, and the method comprises the following steps: reading an original ocean controllable source electromagnetic CSEM time sequence signal, carrying out the adaptive framing processing, and obtaining the current frame length and frame shift; according to the method, the variable step size minimum mean square self-adaptive filtering algorithm is adopted, the step size factor can be dynamically adjusted according to noise components, and the method has natural adaptive capacity to non-stationary and non-Gaussian ocean random noise; the method effectively solves the contradiction between the convergence speed and the steady-state precision of a conventional fixed step size algorithm, tracks the noise change in real time through an adaptive framing mechanism, dynamically adjusts the frame length, further improves the tracking capability of the algorithm for time-varying noise, and remarkably enhances the robustness and adaptability of denoising in a complex marine environment.

Description

Technical Field

[0001] This invention relates to the field of marine geophysical exploration and signal processing technology, and in particular to a method for denoising controllable source electromagnetic signals in the ocean. Background Technology

[0002] Controlled-source electromagnetic (CSEM) is a key geophysical technique for detecting subsea oil and gas reservoirs. It involves towing an artificial electromagnetic source and transmitting low-frequency electromagnetic signals to the seabed, then recording the electric and magnetic field components in a receiver array deployed on the seabed. Due to the complex marine environment (such as wave motion, ocean currents, instrument thermal noise, and ship interference), the received signals inevitably contain high-intensity, complex random noise. This noise severely obscures the primary field response signal, making the extraction of the effective signal extremely difficult.

[0003] In existing technologies, traditional frequency domain filtering (such as bandpass filtering) is simple and easy to implement, but it requires complete separation of the frequency bands of the effective signal and noise. However, the effective signal of CSEM and the frequency band of marine environmental noise overlap significantly. Using this method will result in the loss of a large amount of low-frequency effective signal and introduce the Gibbs phenomenon. Wavelet threshold denoising has good time-frequency localization characteristics, but its performance is highly dependent on the selection of wavelet basis functions, decomposition levels, and threshold functions. It lacks adaptability and its signal processing effect is unstable for different work areas and different noise levels. It is also prone to producing pseudo-Gibbs phenomenon near signal singularities. Methods based on principal component analysis (PCA) or singular value decomposition (SVD) are suitable for removing some coherent noise, but their denoising effect is limited for non-stationary and highly random marine environmental noise. Moreover, the computational load is large, which is not conducive to real-time processing.

[0004] Therefore, a method for denoising electromagnetic signals from a controllable marine source is proposed to solve the above problems. Summary of the Invention

[0005] The main objective of this invention is to provide a method for denoising controllable source electromagnetic signals in the ocean, in order to solve the problems mentioned in the background above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a method for denoising electromagnetic signals from a controllable marine source, comprising the following steps:

[0007] S1: Read the raw ocean-controlled source electromagnetic CSEM time series signal and perform adaptive framing processing to obtain the current frame length and frame shift;

[0008] S2: The signal segment divided based on the current frame length and frame shift, after a unit delay, is used as the reference input signal for the adaptive filter;

[0009] S3: The variable step size minimum mean square (VSS-LM) adaptive filtering algorithm is used to denoise each frame of signal. The step size factor is dynamically adjusted according to the noise components, and the denoised error signal of each frame is output.

[0010] S4: The error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal.

[0011] Preferably, in step one, the adaptive framing processing includes the following steps:

[0012] S11: The noise characterization parameters are calculated using a sliding window, and the noise characterization parameters include local variance and short-time signal-to-noise ratio;

[0013] S12: When the change in noise characterization parameters exceeds a preset threshold across multiple consecutive sampling points, frame length adjustment is triggered.

[0014] S13: Based on the current noise characterization parameters, dynamically determine the length of the current frame within the preset minimum and maximum frame length range.

[0015] Preferably, in step S13, the method for dynamically determining the length of the current frame is as follows:

[0016] If the noise changes drastically, the minimum frame length should be used;

[0017] When the noise is stable, the maximum frame length is used;

[0018] When the noise varies moderately, the frame length is dynamically determined between the minimum and maximum frame lengths through linear interpolation.

[0019] The minimum frame length is no less than twice the order of the adaptive filter, the maximum frame length is no more than eight times the order of the adaptive filter, and the frame shift is always half the current frame length.

[0020] Preferably, in step S3, the denoising process for each frame of signal using the variable step size minimum mean square (VSS-LM) adaptive filtering algorithm includes the following steps:

[0021] S31: At each sampling time, the output signal of the adaptive filter is calculated using the adaptive filter weight coefficients updated at that time;

[0022] S32: Subtract the original ocean-controlled source electromagnetic time series signal from the output signal of the adaptive filter to obtain the error signal;

[0023] S33: Perform zero-phase FIR low-pass filtering on the error signal to separate the effective signal component from the noise component;

[0024] S34: Update the step size factor based on the power function form of the noise component, and use the step size factor to update the weight coefficients of the adaptive filter for calculation at the next sampling time.

[0025] S35: The low-pass filter cutoff frequency is calibrated via STFT every 10 frames.

[0026] Preferably, in S33, the zero-phase FIR low-pass filter is designed as follows:

[0027] The cutoff frequency of the zero-phase FIR low-pass filter needs to dynamically track the lowest frequency of the effective electromagnetic signal from the marine controllable source and be set to half of that lowest frequency.

[0028] The order of the low-pass filter is set to half the order of the adaptive filter;

[0029] The filtering process is performed using a combination of forward and backward methods.

[0030] Preferably, in step S35, the cutoff frequency of the low-pass filter is calibrated every 10 frames via STFT. The specific steps are as follows:

[0031] S351: Perform a short-time Fourier transform on the error signal, using a time window whose length is consistent with the current adaptive framing frame length, and calculate the power spectral density of the signal.

[0032] S352: Extract the peak frequency from the power spectral density, determine the lowest frequency of the effective signal based on the peak frequency, and update the cutoff frequency of the low-pass filter accordingly.

[0033] S353: The cutoff frequency calibration is performed in fixed units of 10 frames, and an update is performed after processing a preset number of data frames.

[0034] Preferably, the adaptive filter is a finite impulse response filter with a transverse structure.

[0035] Preferably, in step S4, the error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal, specifically as follows:

[0036] When the frame length changes from a long frame to a short frame, several sampling points at the end of the previous long frame are overlapped with an equal number of sampling points at the beginning of the current short frame, and the signal values ​​of the overlapping part are weighted and averaged, with the weights transitioning linearly from the previous frame to the current frame.

[0037] When the frame length changes from short to long, the above method is used for overlapping and weighted averaging; finally, a complete denoised time series signal is obtained.

[0038] Preferably, in S12, the number of consecutive sampling points is equal to half the order of the adaptive filter.

[0039] Preferably, in step S34, a minimum threshold constraint is set for the step size factor to ensure that the step size factor is not lower than the minimum threshold constraint.

[0040] The present invention has the following beneficial effects:

[0041] 1. In this invention, a variable step-size least mean square adaptive filtering algorithm is adopted. Its step-size factor can be dynamically adjusted according to the noise components without the need to preset the noise statistical characteristics. It has a natural adaptability to non-stationary and non-Gaussian marine random noise. By introducing a variable step-size mechanism based on the noise power function, the step-size is increased to achieve fast convergence when the noise is severe, and the step-size is decreased to reduce steady-state error when the noise is stable. This effectively solves the contradiction between convergence speed and steady-state accuracy in traditional fixed step-size algorithms. Furthermore, through an adaptive framing mechanism, the frame length is dynamically adjusted in real time to track noise changes, further improving the algorithm's ability to track time-varying noise and significantly enhancing the robustness and adaptability of denoising in complex marine environments.

[0042] 2. In this invention, the reference input is constructed by utilizing the signal's own delay. The noisy signal is delayed by one sampling point using a unit delay unit and then used as the reference noise input for the adaptive filter. This avoids the limitation of relying on additional hardware reference sensors. Noise estimation and suppression are achieved solely through algorithms, which significantly reduces the hardware complexity and deployment cost of the marine CSEM observation system and improves the versatility and portability of the method.

[0043] 3. In this invention, a real-time frequency tracking strategy based on short-time Fourier transform is used to dynamically calibrate the cutoff frequency of the low-pass filter every 10 frames, ensuring that it always matches the current signal's dominant frequency characteristics. This ensures that the effective signal frequency band is fully passed and the noise frequency band is suppressed. At the same time, a zero-phase FIR design with forward-backward filtering is adopted to avoid the phase distortion introduced by conventional filtering. This fully preserves the amplitude and phase information reflecting the underground electrical structure in the CSEM signal, providing a high-quality data foundation for subsequent high-precision inversion. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for denoising electromagnetic signals from a controllable marine source according to the present invention;

[0045] Figure 2 This is a comparison chart of noise pollution and the original signal;

[0046] Figure 3 The image shows the noise reduction effect of LMS.

[0047] Figure 4 The graph shows the relationship between the number of iterations and w and e. Detailed Implementation

[0048] The technical solutions of the embodiments 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Please see Figure 1 This invention provides a technical solution: a method for denoising controllable source electromagnetic signals in the ocean, comprising the following steps:

[0050] S1: Read the raw ocean-controlled source electromagnetic CSEM time series signal and perform adaptive framing processing to obtain the current frame length and frame shift;

[0051] S2: The signal segment divided based on the current frame length and frame shift, after a unit delay, is used as the reference input signal for the adaptive filter;

[0052] S3: The variable step size minimum mean square (VSS-LM) adaptive filtering algorithm is used to denoise each frame of signal. The step size factor is dynamically adjusted according to the noise components, and the denoised error signal of each frame is output.

[0053] S4: The error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal.

[0054] In step one, adaptive frame segmentation processing includes the following steps:

[0055] S11: The noise characterization parameters are calculated using a sliding window. The noise characterization parameters include local variance and short-time signal-to-noise ratio.

[0056] Specifically, the formula for calculating local variance is:

[0057] ;

[0058] In the formula, This represents the length of the sliding window, which is equal to the order of the adaptive filter. Represents the original, controllable ocean electromagnetic time-series signal. The index representing the current sampling time. Represents the signal within the sliding window The local average value, Represents the signal within the sliding window Local variance;

[0059] The formula for calculating the short-time signal-to-noise ratio is:

[0060] ;

[0061] In the formula, This represents the short-time signal-to-noise ratio.

[0062] S12: When the change in noise characterization parameters exceeds a preset threshold across multiple consecutive sampling points, frame length adjustment is triggered.

[0063] Specifically, the threshold for variance variation: , ( =0.2 , (The initial noise variance is calculated from the first 100 sampling points).

[0064] Signal-to-noise ratio change threshold: (TSNR=3dB, which conforms to engineering practice);

[0065] Adjustment is triggered when either the local variance or the short-time signal-to-noise ratio exceeds the variance change threshold or the signal-to-noise ratio change threshold.

[0066] in, The change at point K represents the local variance. Represents the current sampling time The local variance, This represents the local variance before K sampling points, where K represents the number of consecutive sampling points. Represents the variance threshold. The change at point K represents the short-time signal-to-noise ratio. Represents the current sampling time The short-time signal-to-noise ratio, Represents the signal-to-noise ratio threshold

[0067] S13: Based on the current noise characterization parameters, dynamically determine the length of the current frame within the preset minimum and maximum frame length range.

[0068] In step S13, the method for dynamically determining the length of the current frame is as follows:

[0069] If the noise changes drastically, the minimum frame length should be used;

[0070] When the noise is stable, the maximum frame length is used;

[0071] When the noise varies moderately, the frame length is dynamically determined between the minimum and maximum frame lengths through linear interpolation.

[0072] The minimum frame length is no less than twice the order of the adaptive filter, the maximum frame length is no more than eight times the order of the adaptive filter, and the frame shift is always half the current frame length.

[0073] Specifically, the noise undergoes drastic changes (strong time-varying): or Frame length set to It can quickly track changes in noise.

[0074] Moderate noise variation (weak time-varying): or Linear interpolation adjustment: ;

[0075] The noise level is stable (no time variation): or Frame length set to Improve processing efficiency;

[0076] The threshold values ​​for each are:

[0077] ;

[0078] In the above formula, Represents the maximum frame length. Represents the minimum frame length. This represents the current frame length that needs to be determined. The upper limit threshold representing the local variance. The lower bound threshold representing the local variance. The upper threshold representing the short-time signal-to-noise ratio. The lower threshold representing the short-time signal-to-noise ratio;

[0079] In S3, the variable step size minimum mean square (VSS-LM) adaptive filtering algorithm is used to denoise each frame of signal, including the following steps:

[0080] S31: At each sampling time, use the updated adaptive filter weight coefficients at that time to calculate the output signal of the adaptive filter;

[0081] The output signal is:

[0082] ;

[0083] In the formula, This represents the adaptive filter at the sampling time. The output signal, Represents the moment The adaptive filter weight coefficient vector, The transpose operator for vectors or matrices. Represents the moment The reference input vector input to the adaptive filter is the original noisy signal after a unit delay. Represents the original noisy signal. Represents the moment The Each filter weight coefficient Represents the reference input signal at time... The value of .

[0084] S32: Subtract the original ocean-controlled source electromagnetic time series signal from the output signal of the adaptive filter to obtain the error signal;

[0085] S33: Perform zero-phase FIR low-pass filtering on the error signal to separate the effective signal component from the noise component;

[0086] The separation formula is as follows:

[0087] ;

[0088] - ;

[0089] In the formula, Represents the error signal. Represents a zero-phase finite impulse response low-pass filter operation. The estimated value representing the effective signal component. Estimated values ​​representing noise components;

[0090] S34: Update the step size factor in the form of a power function based on the noise component, and use the step size factor to update the weight coefficients of the adaptive filter for calculation at the next sampling time.

[0091] The formula for calculating the step size factor based on the power function form of the noise component is as follows:

[0092] ;

[0093] In the formula, Represents the moment Adaptive step size factor, Represents a constant greater than 0, the amplitude gain coefficient of the step size factor. This represents the shape control coefficient of the step size factor, and is a constant greater than 0. Represents the natural exponential function. The instantaneous power representing the estimated noise component;

[0094] The effect is that when the noise increases... Increase, μ(n) Increasing the size accelerates convergence;

[0095] When the effective signal changes abruptly Increase, but Unchanged, μ(n) Maintain stability and avoid distortion of effective signals;

[0096] When the noise decreases Decrease, μ(n) Reduce, thereby lowering the steady-state error.

[0097] S35: The low-pass filter cutoff frequency is calibrated via STFT every 10 frames.

[0098] In S33, the zero-phase FIR low-pass filter is designed as follows:

[0099] The cutoff frequency of the zero-phase FIR low-pass filter needs to dynamically track the lowest frequency of the effective electromagnetic signal from the marine controllable source and be set to half of that lowest frequency.

[0100] The order of the low-pass filter is set to half that of the adaptive filter;

[0101] The filtering process employs a combination of forward and backward methods.

[0102] In S35, the low-pass filter cutoff frequency is calibrated every 10 frames via STFT. The specific steps are as follows:

[0103] S351: Perform a short-time Fourier transform on the error signal, using a time window whose length is consistent with the current adaptive framing length, and calculate the power spectral density of the signal.

[0104] The calculation formula is:

[0105] ;

[0106] In the formula: Represents the window function. Represents frame shift, Represents the number of FFT points. Represents the frame index. Represents frequency index, This represents the error signal for the current frame. Representing the Frame, First Complex spectral values ​​at each frequency point;

[0107] Power spectral density (PSD) calculation:

[0108] ;

[0109] In the formula, Represents power spectral density;

[0110] S352: Extract the peak frequency from the power spectral density, determine the lowest frequency of the effective signal based on the peak frequency, and update the cutoff frequency of the low-pass filter accordingly.

[0111] The calculation formula is: Extract peak frequency :

[0112]

[0113] In the formula: Represents the sampling frequency. This represents the frequency index where the maximum power spectrum value is found.

[0114] The formula for calculating the lowest frequency of the effective signal is as follows:

[0115] ;

[0116] In the formula, The lowest frequency representing a valid signal. Represents frequency offset;

[0117] Then update the low-pass filter cutoff frequency: ;

[0118] S353: The cutoff frequency calibration is performed every 10 frames, and an update is performed after a preset number of data frames have been processed.

[0119] The adaptive filter is a finite impulse response filter with a transverse structure.

[0120] In S4, the error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal, specifically:

[0121] When the frame length changes from a long frame to a short frame, several sampling points at the end of the previous long frame are overlapped with an equal number of sampling points at the beginning of the current short frame, and the signal values ​​of the overlapping part are weighted and averaged, with the weights transitioning linearly from the previous frame to the current frame.

[0122] When the frame length changes from short to long, the above method is used for overlapping and weighted averaging; finally, a complete denoised time series signal is obtained.

[0123] In S12, the number of consecutive sampling points is equal to half the order of the adaptive filter.

[0124] In S34, a minimum threshold constraint is set for the step size factor to ensure that the step size factor is not lower than the minimum threshold constraint.

[0125] The formula for calculating the step size factor after adding a minimum threshold constraint is as follows:

[0126] ;

[0127] In the formula, Represents the moment Constrained adaptive step size factor This represents the minimum threshold of the step size factor, used to prevent the filter from stalling due to an excessively small step size factor.

[0128] The synthesized noisy CSEM signal was processed using the steps of this invention. Calculations showed that the signal-to-noise ratio (SNR) of the denoised signal increased from the original 5dB to 15dB, an improvement of approximately 10dB. Figure 2 As shown, random spike noise in the waveform diagram is effectively suppressed. At the same time, the key waveform features of the CSEM signal (such as amplitude envelope and zero crossing) are well preserved, and no obvious waveform distortion is observed. This indicates that the method of the present invention is very suitable for denoising marine CSEM signals.

[0129] like Figure 3 and Figure 4 As shown, the denoised signal after processing the contaminated signal under LMS parameter conditions retains its details clearly, and the energy recovery is relatively complete. It can be seen that the spikes caused by noise in the signal have been effectively suppressed. The curve is smooth and the consistency is very good, achieving the denoising effect. After 500 iterations, the weight coefficient W is basically close to stable after the 200th iteration. The entire system converges stably, and the weight coefficient update speed is relatively fast. The convergence curve of the error e shows that the curve changes significantly before the 200th iteration, with relatively large oscillations. After the 200th iteration, it tends to stabilize, and the overall curve change is relatively gentle. This indicates that the parameter values ​​in the LMS algorithm are relatively reasonable, further proving the feasibility of using the LMS algorithm to denoise noisy marine CSEM signals.

[0130] The above embodiments demonstrate that the denoising method proposed in this invention can effectively suppress random noise and maintain signal waveforms in both real-time processing at sea and post-processing indoors, and has engineering feasibility.

[0131] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0132] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for denoising controllable marine electromagnetic signals, characterized in that, Includes the following steps: S1: Read the raw ocean-controlled source electromagnetic CSEM time series signal and perform adaptive framing processing to obtain the current frame length and frame shift; S2: The signal segment divided based on the current frame length and frame shift, after a unit delay, is used as the reference input signal for the adaptive filter; S3: The variable step size minimum mean square (VSS-LM) adaptive filtering algorithm is used to denoise each frame of signal. The step size factor is dynamically adjusted according to the noise components, and the denoised error signal of each frame is output. S4: The error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal.

2. The method for denoising marine controllable source electromagnetic signals according to claim 1, characterized in that, In step one, the adaptive frame segmentation processing includes the following steps: S11: The noise characterization parameters are calculated using a sliding window, and the noise characterization parameters include local variance and short-time signal-to-noise ratio; S12: When the change in noise characterization parameters exceeds a preset threshold across multiple consecutive sampling points, frame length adjustment is triggered. S13: Based on the current noise characterization parameters, dynamically determine the length of the current frame within the preset minimum and maximum frame length range.

3. The method for denoising marine controllable source electromagnetic signals according to claim 2, characterized in that, In step S13, the method for dynamically determining the length of the current frame is as follows: If the noise changes drastically, the minimum frame length should be used; When the noise is stable, the maximum frame length is used; When the noise varies moderately, the frame length is dynamically determined between the minimum and maximum frame lengths through linear interpolation. The minimum frame length is no less than twice the order of the adaptive filter, the maximum frame length is no more than eight times the order of the adaptive filter, and the frame shift is always half the current frame length.

4. The method for denoising controllable marine electromagnetic signals according to claim 1, characterized in that, In step S3, the variable step size minimum mean square (VSS-LM) adaptive filtering algorithm is used to denoise each frame of signal, including the following steps: S31: At each sampling time, the output signal of the adaptive filter is calculated using the adaptive filter weight coefficients updated at that time; S32: Subtract the original ocean-controlled source electromagnetic time series signal from the output signal of the adaptive filter to obtain the error signal; S33: Perform zero-phase FIR low-pass filtering on the error signal to separate the effective signal components and noise components; S34: Update the step size factor based on the power function form of the noise component, and use the step size factor to update the weight coefficients of the adaptive filter for calculation at the next sampling time. S35: The low-pass filter cutoff frequency is calibrated via STFT every 10 frames.

5. A method for denoising controllable marine electromagnetic signals according to claim 4, characterized in that, In S33, the zero-phase FIR low-pass filter is designed as follows: The cutoff frequency of the zero-phase FIR low-pass filter needs to dynamically track the lowest frequency of the effective electromagnetic signal from the marine controllable source and be set to half of that lowest frequency. The order of the low-pass filter is set to half the order of the adaptive filter; The filtering process employs a combination of forward and backward methods.

6. The method for denoising marine controllable source electromagnetic signals according to claim 4, characterized in that, In step S35, the low-pass filter cutoff frequency is calibrated every 10 frames via STFT. The specific steps are as follows: S351: Perform a short-time Fourier transform on the error signal, using a time window whose length is consistent with the current adaptive framing frame length, and calculate the power spectral density of the signal. S352: Extract the peak frequency from the power spectral density, determine the lowest frequency of the effective signal based on the peak frequency, and update the cutoff frequency of the low-pass filter accordingly. S353: The cutoff frequency calibration is performed in fixed units of 10 frames, and an update is performed after processing a preset number of data frames.

7. The method for denoising marine controllable source electromagnetic signals according to claim 1, characterized in that, The adaptive filter is a finite impulse response filter with a transverse structure.

8. A method for denoising controllable marine electromagnetic signals according to claim 1, characterized in that, In step S4, the error signals after processing all frames are superimposed and reconstructed according to the frame shift to obtain the complete denoised time series signal, specifically: When the frame length changes from a long frame to a short frame, several sampling points at the end of the previous long frame are overlapped with an equal number of sampling points at the beginning of the current short frame, and the signal values ​​of the overlapping part are weighted and averaged, with the weights transitioning linearly from the previous frame to the current frame. When the frame length changes from short to long, the above method is used for overlapping and weighted averaging; finally, a complete denoised time series signal is obtained.

9. A method for denoising controllable marine electromagnetic signals according to claim 2, characterized in that, In S12, the number of consecutive sampling points is equal to half the order of the adaptive filter.

10. A method for denoising controllable source electromagnetic signals in marine environments according to claim 4, characterized in that, In step S34, a minimum threshold constraint is set for the step size factor to ensure that the step size factor is not lower than the minimum threshold constraint.

Images