A dry-type air-core reactor optical fiber sensing noise reduction method based on improved SVD

Abstract

The application discloses a kind of based on improved SVD's dry-type air-core reactor optical fiber sensing noise reduction method, and improved SVD is proposed to be applied to distributed optical fiber sensing detection dry-type air-core reactor fault alarm system.Establish the backscattering model of temperature measurement system, add Gaussian white noise and utilize traditional threshold denoising and improved SVD denoising analysis signal-to-noise ratio, provide theoretical basis for experiment.Under existing conditions in laboratory, collect original anti-Stokes light with temperature information and Stokes light and obtain ratio data matrix by ratio, after remodeling data matrix, singular value decomposition reconstruction is carried out, then after remodeling matrix, the ratio signal with temperature information after denoising is obtained, into two-way demodulation formula to obtain demodulation temperature information, finally according to demodulation error, compare denoising effect, the present application can effectively reduce demodulation temperature error by noise reduction algorithm, improve system temperature measurement precision, compared with traditional wavelet denoising, improve system signal-to-noise ratio, avoid threshold selection uncertainty and other problems.

Description

A Noise Reduction Method for Fiber Optic Sensing in Dry-Type Air-Core Reactors Based on Improved SVD Technical Field

[0001] This invention relates to a noise reduction method for distributed sensing systems, and more particularly to a noise reduction method for fiber optic sensing based on an improved SVD for dry air-core reactors. Background Technology

[0002] With the continuous development of fiber optic sensing technology, fiber optic materials possess excellent resistance to oxidation, electromagnetic interference, and corrosion, as well as low cost, simple installation, no need for regular maintenance, and long service life. Therefore, fiber optic sensing technology is widely used in special scenarios such as power systems, tunnels and mines, and transportation.

[0003] Reactors are essential equipment in power systems, used to compensate for reactive power and reduce dynamic voltage. Dry-type air-core reactors are commonly used today, offering advantages such as low loss, low noise, good reactance linearity, long design life, and simple maintenance. In the system, these devices primarily function to limit inrush current, limit short-circuit current, compensate for stray capacitive current, and filter components.

[0004] During operation, reactors often experience localized overheating due to impurities in the coil wires or poor insulation of the epoxy resin-coated glass fiber tape. This can lead to reactor burnout and severe damage to the power system. Since conventional temperature sensing systems are unsuitable for accurately measuring the temperature field of reactors, distributed fiber optic sensing is proposed to address this issue. Distributed sensing systems primarily utilize Rayleigh scattering, Brillouin scattering, and Raman scattering. Raman scattering signals are highly sensitive to temperature and have strong anti-interference capabilities, making them widely used in temperature measurement. However, Raman scattering signals are weak and easily affected by system noise, resulting in a low signal-to-noise ratio and impacting measurement accuracy. Therefore, continuously improving the system's signal-to-noise ratio is essential. Currently, noise reduction is mainly achieved through hardware denoising and data processing. Hardware denoising offers significant results but is costly and has limitations in practical applications. Furthermore, existing data processing-based noise reduction methods are complex to operate and rely on human experience, introducing uncertainty. Summary of the Invention

[0005] The technical problem to be solved by the invention is: how to improve the signal-to-noise ratio of the temperature measurement system, reduce the temperature demodulation error in the reactor fault detection process based on the temperature measurement signal, and provide an improved noise reduction method for the SVD distributed optical fiber temperature measurement system.

[0006] The present invention solves the above-mentioned technical problems through the following technical solution, and the present invention includes the following steps:

[0007] S0: Establish a simulation mathematical model

[0008] Based on the distributed optical fiber sensing mechanism and the component configuration, a mathematical model of backscattering is established. Different levels of Gaussian white noise are added to the mathematical model of backscattering, and the signal-to-noise ratio before and after the noise reduction algorithm is calculated.

[0009] S1: Laying distributed sensing optical fibers on dry-type air-core reactors

[0010] The optical fiber is bonded to the reactor encapsulation surface using high-temperature resistant epoxy resin for real-time temperature monitoring.

[0011] S2: Acquire reactor temperature signal

[0012] Under the existing laboratory conditions, the anti-Stokes signal and Stokes signal in the fiber optic temperature measurement system were collected, and the ratio of the anti-Stokes signal to the Stokes signal was calculated according to the dual-path demodulation scheme.

[0013] S3: Reshape the temperature ratio signal

[0014] The calculated ratio signal is reshaped into a corresponding m×n dimensional matrix, where m and n represent the smallest common factor of the ratio signal L.

[0015] S4: Perform SVD decomposition on the matrix

[0016] Perform singular value decomposition on the matrix to obtain a left singular matrix, a singular value diagonal matrix, and a right singular matrix. Retain the first singular value in the singular value diagonal matrix and set the remaining singular values ​​to 0, forming a new singular value matrix.

[0017] S5: Reconstruct the SVD signal to obtain the denoised temperature data, complete demodulation to acquire temperature measurement results, and analyze the denoising effect through error analysis.

[0018] Reconstruct the left singular matrix, the right singular matrix, and the new singular value diagonal matrix obtained in step S4. Use the reconstructed signal to reshape the matrix and input it into the demodulation formula to obtain the temperature measurement result.

[0019] S6: Set alarm threshold temperature on the client side.

[0020] The system compares the temperature data of the dry-type air-core reactor after noise reduction and demodulation with the threshold temperature data set by the client. When the demodulated temperature data is greater than the threshold temperature data, the system alarms. When the demodulated temperature data is less than the threshold temperature data, the system continues to monitor the reactor status in real time.

[0021] Furthermore, in the aforementioned S0, the mathematical model for backscattering is specifically as follows:

[0022] P(t)=0.5Sα s vP0Te -αvt

[0023] Where S is the backscattering coefficient; α s α is the backscattering factor; v is the group velocity of light in the optical fiber; P0 is the incident light power; T is the optical pulse width; α is the transmission loss coefficient.

[0024] Furthermore, in S3, the ratio of the acquired anti-Stokes signal and Stokes signal is calculated and reconstructed into an m×n dimensional H matrix, specifically:

[0025]

[0026] Where m≥2 and n≥2 are common factors of L, and m×n=L.

[0027] Furthermore, in S4, after performing SVD processing on the H matrix, we obtain:

[0028] Y = USV T

[0029] Where U is an m×m left orthogonal matrix; V is an n×n right orthogonal matrix; and S is a singular value diagonal matrix. The diagonal elements are X1, X2, ..., X... min(m,n) ; where X1, X2, ... are the singular values ​​of matrix Y.

[0030] Furthermore, in S4, only the first singular value in the S singular value diagonal matrix is ​​retained, and the remaining singular values ​​are set to 0; S(2:i,2:i)=0 where i=min(m,n). The processed signal is reconstructed to obtain the denoised temperature ratio curve, which is then substituted into the demodulation formula to obtain the temperature curve.

[0031] Furthermore, the S5 dual-channel demodulation formula is as follows:

[0032]

[0033] Where h is Planck's constant, h = 6.626 × 10⁻³⁴ J·s; k is Boltzmann's constant, k = 1.38 × 10⁻²³ J / K; Δv is Raman frequency shift, Δv = 13.2 THz; K is the ratio of the denoised anti-Stokes signal to the Stokes signal, and T0 is the known temperature;

[0034] Furthermore, regarding the specific error in step S5...

[0035] Where T maxFor the demodulated maximum temperature measurement result, T min The minimum temperature measurement result obtained through demodulation; N represents all valid temperature measurement data obtained through demodulation.

[0036] This invention verifies the feasibility of denoising algorithms in improving the signal-to-noise ratio of a distributed fiber optic sensing and detection dry-type air-core reactor fault alarm system from both simulation and experimental perspectives. Under laboratory conditions, the ratio of the acquired anti-Stokes and Stokes optical flux raw data is first calculated. This ratio data is then reconstructed and decomposed into a left singular matrix, a singular value diagonal matrix, and a right singular matrix using SVD. The first singular value in the singular value matrix is ​​retained to obtain a new singular value matrix. The left, right, and new singular value diagonal matrices are then reconstructed to obtain the denoised matrix. The denoised matrix is ​​then reconstructed and substituted into the demodulation formula to obtain the temperature measurement result. Finally, the experimental temperature measurement data demodulated under different denoising methods are compared, and the difference between the highest and lowest demodulated temperatures is used as an indicator to measure the denoising effect.

[0037] Compared with the prior art, the advantages of the present invention are as follows: The present invention proposes an improved SVD algorithm and applies it to the fault alarm system of dry air-core reactor, so as to improve the signal-to-noise ratio of the system and improve the temperature measurement accuracy from the software perspective without changing the hardware structure. Compared with traditional wavelet denoising and traditional SVD denoising, it avoids the randomness of selecting wavelets and singular values ​​based on empirical values, and reduces the denoising process and time. Attached Figure Description

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0039] Figure 1. Flowchart of the improved SVD algorithm applied to the fault alarm system of fiber optic sensing detection dry air-core reactor.

[0040] Figure 2. Schematic diagram of a distributed fiber optic temperature measurement system under laboratory conditions;

[0041] Figure 3. Schematic diagram of a fiber optic sensing noise reduction method for dry air-core reactors based on improved SVD under laboratory conditions.

[0042] Figure 4 shows the simulated backscattering mathematical model, the effect of adding Gaussian white noise, and the effect after noise reduction, respectively.

[0043] Figure 5. Comparison of signal-to-noise ratio improvement effects of different noise reduction methods under simulation angle;

[0044] Figure 6 shows a comparison of the noise reduction effects of different noise reduction methods at 30.0℃;

[0045] Figure 7 shows a comparison of the noise reduction effects of different noise reduction methods at 40.0℃;

[0046] Figure 8 shows a comparison of the noise reduction effects of different noise reduction methods at 50.0℃;

[0047] Figure 9 shows a comparison of the noise reduction effects of different noise reduction methods at 60.0℃;

[0048] Figure 10 shows a comparison of demodulation errors of different noise reduction methods at different temperatures; Detailed Implementation

[0049] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] As shown in Figures 1, 2, and 3, the main equipment of the distributed fiber optic temperature measurement system is as follows: pulsed light source, circulator, wavelength division multiplexer, sensing fiber, photodiode, amplifier, information acquisition module, and computer. The specific principle is as follows: The light source emits a fixed laser pulse, which enters the sensing fiber through the circulator. Backscattering occurs in the sensing fiber, generating backscattered light. After passing through the circulator and the wavelength division multiplexer, the backscattered light is split into anti-Stokes light and Stokes light. The two optical signals are converted into electrical signals by an avalanche photodiode, and then amplified by an amplifier. After passing through the signal acquisition module, the anti-Stokes signal and the Stokes signal are obtained, and finally transmitted to the computer for data processing.

[0051] A method for noise reduction of fiber optic sensors in dry air-core reactors based on improved SVD includes the following steps:

[0052] Step 1: Simulation to establish a mathematical model

[0053] Based on the distributed fiber optic sensing mechanism and referring to the component configuration, a mathematical model for backscattering is established; the details are as follows:

[0054] P(t)=0.5Sa s vP0Te -αvt

[0055] Where S is the backscattering coefficient; α s α is the backscattering factor; v is the group velocity of light in the optical fiber; P0 is the incident light power; T is the optical pulse width; α is the transmission loss coefficient.

[0056] Step 2: Add Gaussian white noise of different degrees to the mathematical model of backscattering, and calculate the signal-to-noise ratio before and after algorithm processing. The calculation formula is as follows:

[0057]

[0058] Wherein, SNR represents the signal-to-noise ratio, and the larger the value, the better the denoising effect; P(1) is the original signal, P(2) is the signal after processing by different denoising algorithms, and L is the length of the signal.

[0059] Step 3: Use the temperature measurement system to collect the anti-Stokes signal and the Stokes signal, calculate the ratio, and reconstruct the obtained L signal into an m×n dimensional H matrix, specifically in the following form:

[0060]

[0061] Where m≥2 and n≥2 are common factors of L, and m×n=L.

[0062] Step 4: Perform SVD decomposition on the H matrix to obtain:

[0063] Y = USV T

[0064] In the formula, U is an m×m left orthogonal matrix; V is an n×n right orthogonal matrix; and S is a singular value diagonal matrix. The diagonal elements are X1, X2, ..., X... min(m,n) ; where X1, X2, ... are the singular values ​​of matrix Y.

[0065] In step 4, take the first singular value from the diagonal matrix of singular values ​​and set the rest to 0.

[0066] Step 5: Reconstruct the new singular value diagonal matrix obtained in Step 4 with the left and right singular values.

[0067] The reconstructed ratio data is then reshaped and substituted into the demodulation formula to obtain the temperature measurement curve. The demodulation formulas involved in this algorithm are as follows:

[0068]

[0069] In the formula, h is Planck's constant, h = 6.626 × 10⁻³⁴ J·s; k is Boltzmann's constant, k = 1.38 × 10⁻²³ J / K; Δv is Raman frequency shift, Δv = 13.2 THz; K is the ratio of the denoised anti-Stokes signal to the Stokes signal, and T0 is the known temperature.

[0070] Based on the temperature measurement results, the effectiveness of the denoising algorithm is measured using a temperature error index; the smaller the temperature error, the better the effect. Specifically:

[0071]

[0072] Where T max For the demodulated maximum temperature measurement result, T min The minimum temperature measurement result obtained through demodulation; N represents all valid temperature measurement data obtained through demodulation.

[0073] Example 2: Referring to Figures 1 to 10, and Figure 4, the figures respectively show the temperature signal curve obtained using the mathematical model of backscattering at the simulated angle, the effect after adding Gaussian white noise, and the effect after improved SVD denoising. A comparison of the three figures shows that the algorithm of this invention has a significant denoising effect.

[0074] Referring to Figure 5, this figure compares the improvement of the SVD denoising algorithm of the present invention with that of the traditional denoising algorithm in improving the signal-to-noise ratio of the simulated temperature measurement signal. The horizontal axis represents the signal-to-noise ratio of the simulated signal to the original signal. The three matrices represent the signal-to-noise ratios after wavelet hard threshold denoising, wavelet soft threshold denoising, and improved SVD denoising, respectively. The comparison shows that the denoising algorithm of the present invention has a higher signal-to-noise ratio than the traditional denoising algorithm.

[0075] Referring to Figure 6, taking an actual temperature of 30℃ as an example, we calculate the ratio of the collected anti-Stokes signal and the Stokes signal, and reshape its signal length L into an m×n dimensional H matrix, where L is 200 points.

[0076] The H matrix is ​​then decomposed into three matrices: left singular value matrix, singular value diagonal matrix, and right singular value matrix. The first singular value of the singular value diagonal matrix is ​​retained, and the other singular values ​​are set to 0. After obtaining the new singular value diagonal matrix, it is reconstructed with the left and right singular matrices to obtain a new matrix. The signal is then reconstructed and recorded as the ratio signal after noise reduction. Finally, it is substituted into the demodulation formula to obtain the temperature measurement curve.

[0077] Similarly, we repeated the above steps for the temperature curves collected at 40.0℃, 50.0℃, and 60.0℃, respectively, to obtain three sets of comparison charts of different measurements, as shown in Figures 7, 8, and 9. Each chart has four lines, representing the original data, the temperature curve after hard threshold denoising, soft threshold denoising, and the temperature curve after improved SVD denoising and demodulation.

[0078] Referring to Figure 10, which is a comparison of the errors of several denoising algorithms calculated based on the error index, it can be seen that under the conditions of 30.0℃, 40.0℃, 50.0℃ and 60.0℃, the denoising error of this method is controlled within 0.58℃, 0.71℃, 0.73℃ and 0.87℃ respectively, and the error is less than that of traditional threshold denoising.

Claims

1. A method for noise reduction of fiber optic sensing in dry air-core reactors based on improved SVD, characterized in that: Includes the following steps: S0: Establishing a simulation mathematical model. Based on the distributed optical fiber sensing mechanism and referring to the component configuration, a mathematical model for backscattering is established: In the formula, S is the backscattering coefficient; It is the backscattering factor; The group velocity of light in the optical fiber; The incident light power; The width of the light pulse; The transmission loss coefficient is used; different levels of Gaussian white noise are added to the mathematical model of backscattering, and the signal-to-noise ratio before and after the noise reduction algorithm is calculated respectively. S1: Lay distributed sensing optical fibers on the dry-type hollow reactor. The optical fibers are bonded to the reactor's encapsulation surface using high-temperature resistant epoxy resin for real-time temperature monitoring. S2: Collect reactor temperature signals. Under existing laboratory conditions, collect the anti-Stokes signal and Stokes signal from the optical fiber temperature measurement system. Calculate the ratio of the anti-Stokes signal to the Stokes signal using a dual-path demodulation scheme to obtain the temperature ratio signal. S3: Reshape the temperature ratio signal. Reshape the calculated ratio signal into a corresponding m×n dimensional matrix, specifically: in ， The common factors of L, The m and n represent the smallest common factors of the ratio signal L, where L represents the length of the ratio signal. S4: Perform SVD decomposition on the matrix. After SVD decomposition, the matrix is ​​obtained as a left singular matrix, a singular value diagonal matrix, and a right singular matrix. The first singular value in the singular value diagonal matrix is ​​retained, and the remaining singular values ​​are set to 0 to form a new singular value diagonal matrix. S5: Reconstruct the SVD signal to obtain the denoised temperature data. Demodulate and obtain the temperature measurement result. Through error analysis and denoising effect, the left singular matrix, the right singular matrix, and the new singular value diagonal matrix obtained in step S4 are reconstructed. The reconstructed signal is then used to reshape the signal and enter the demodulation formula to obtain the temperature measurement result. S6: The alarm threshold temperature is set on the client side. The temperature data of the dry-type air-core reactor after noise reduction and demodulation is compared with the threshold temperature data set on the client side. When the demodulated temperature data is greater than the threshold temperature data, the system alarms. When the demodulated temperature data is less than the threshold temperature data, the system continues to monitor the reactor status in real time.

2. The method for denoising fiber optic sensors in dry air-core reactors based on improved SVD according to claim 1, characterized in that: in step S4, after performing SVD processing on the H matrix, the following is obtained: In the formula, U is The left orthogonal matrix; V is S is a right orthogonal matrix; S is a singular value diagonal matrix. ;in 、 …and these are the singular values ​​of matrix Y.

3. The fiber optic sensing noise reduction method for dry air-core reactors based on improved SVD as described in claim 2 is characterized in that: in step S4, only the first singular value in the S singular value diagonal matrix is ​​retained, and the remaining singular values ​​are set to 0. in The processed signal is reconstructed to obtain the denoised temperature ratio curve, and the reconstructed curve is then substituted into the demodulation formula to obtain the temperature curve.

4. The fiber optic sensing noise reduction method for dry air-core reactors based on improved SVD according to claim 3, characterized in that: in step S5, the dual-path demodulation formula is: In the formula, h is Planck's constant, h = 6.626 × 10⁻³⁴ J·s; k is Boltzmann's constant, k = 1.38 × 10⁻²³ J / K; —Raman shift, =13.2 THz; K—ratio of the anti-Stokes signal to the Stokes signal after noise reduction, T0 is the known temperature.

5. The fiber optic sensing noise reduction method for dry air-core reactors based on improved SVD according to claim 4, characterized in that: the error in step S5 is specifically: In the formula The maximum temperature measurement result demodulated in claim 4. The minimum temperature measurement result obtained through demodulation; N represents all valid temperature measurement data obtained through demodulation.

Images