Terahertz time-domain spectral signal noise reduction method and system for transformer aging insulating oil

Abstract

The invention provides a terahertz time-domain spectral signal noise reduction method and system for transformer aged insulating oil, and the method comprises the steps: collecting a terahertz time-domain spectral signal of an aged insulating oil sample, and carrying out the parameter optimization of variational mode decomposition through a whale optimization algorithm, according to the method, an optimized variational mode decomposition algorithm is adopted to decompose terahertz time-domain spectral signals of aged insulating oil into a series of intrinsic mode function components, then independent component analysis is adopted to separate the intrinsic mode function components into noise and effective signals, and finally the effective signals are screened out by calculating information entropy of independent components. And reconstructing the terahertz time-domain spectral signal of the aged insulating oil after noise reduction. The technical problems that the wavelet basis function and the decomposition layer number of wavelet noise reduction are difficult to determine, the noise reduction effect is restricted by the endpoint effect and mode aliasing problem of empirical mode decomposition, and the noise reduction effect of variational mode decomposition is poor are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of terahertz detection and signal processing, and in particular to a method and system for denoising terahertz time-domain spectroscopy signals of transformer aging insulating oil. BACKGROUND

[0002] Terahertz waves refer to electromagnetic waves with a frequency range of 0.1 THz to 10 THz, between microwaves and infrared rays, and have many unique properties. Terahertz waves can penetrate non-metallic materials and perform high-resolution imaging, so they have wide application prospects in communication, imaging, detection, material analysis, etc.

[0003] Transformer fault diagnosis is an important research field in power systems, aiming to prevent and reduce the occurrence of faults by timely monitoring and analyzing the operating state of transformers. Transformer faults mainly include mechanical faults, electrical faults, insulation faults, and thermal faults, etc. Traditional fault diagnosis methods usually require long detection and analysis times. This makes them have a certain lag in real-time monitoring and early warning, and they cannot quickly respond to changes in the operating state of the transformer, which may miss the best maintenance opportunity for the transformer.

[0004] In recent years, some scholars have combined terahertz time-domain spectroscopy technology with transformer fault diagnosis, using terahertz time-domain spectroscopy technology to detect transformer insulating oil, and analyzing the characteristics of the terahertz time-domain spectroscopy signals of the insulating oil to determine the operating state of the transformer in real time, solving the problem of the lag of traditional diagnosis. However, the terahertz signal generation device will be disturbed by noise during the transmission and sampling of terahertz signals, such as thermal noise due to the random motion of free electrons in the device; shot noise due to the discreteness of electrons in semiconductor devices and photodetectors. In addition, the generation and detection of terahertz signals rely on precise optoelectronic devices such as cascade lasers and photodetectors, and the non-ideal working characteristics of these devices may also introduce noise, further affecting the quality of the signals. Therefore, in order to accurately extract the characteristics of the terahertz time-domain spectroscopy signals of transformer aging insulating oil, it is necessary to perform denoising processing on the terahertz signals.

[0005] The prior art patent application document with the publication number CN118051758A discloses a method for denoising a terahertz time-domain spectrum signal, which comprises the following steps: performing processing on system echoes in the terahertz time-domain spectrum signal by using a double-Gaussian inverse filtering algorithm to obtain a processed terahertz time-domain spectrum signal; decomposing the processed terahertz time-domain spectrum signal by using an optimized variational mode decomposition method to obtain a plurality of target intrinsic mode function components; and filtering out noise modes from each target intrinsic mode component to determine a denoised terahertz time-domain spectrum signal. The prior art uses a sparrow search algorithm (SSA) to optimize VMD parameters (K and Alpha), but the signal-to-noise ratio (SNR), mean square error (MSE), and waveform similarity coefficient (NCC) indexes of the denoised signal of the prior art still need to be improved.

[0006] At present, the main methods for denoising the terahertz time-domain spectrum signal of the aging insulating oil of a transformer include wavelet denoising and empirical mode decomposition denoising. The wavelet denoising in the prior art has the problem that it is difficult to determine the wavelet basis function and the decomposition level. The empirical mode decomposition has the problems of end effect and mode aliasing, which greatly affect the denoising effect. Therefore, some scholars have proposed a variational mode decomposition denoising method, which solves the problems of end effect and mode aliasing, but the denoising effect of the variational mode decomposition depends on the appropriate decomposition mode number K, and the penalty factor Alpha also affects the denoising effect. Moreover, for the complex signal of the terahertz time-domain spectrum with multiple sources, each intrinsic mode function obtained by the variational mode decomposition may contain both effective signals and noise, resulting in poor denoising effect.

[0007] The prior art has the technical problems that the wavelet denoising in the prior art has the problem that it is difficult to determine the wavelet basis function and the decomposition level, the empirical mode decomposition has the problems of end effect and mode aliasing, which greatly affect the denoising effect, and the denoising effect of the variational mode decomposition is poor. SUMMARY

[0008] The technical problem to be solved by the present application is how to solve the technical problems that the wavelet denoising in the prior art has the problem that it is difficult to determine the wavelet basis function and the decomposition level, the empirical mode decomposition has the problems of end effect and mode aliasing, which greatly affect the denoising effect, and the denoising effect of the variational mode decomposition is poor.

[0009] The present application solves the above technical problems by using the following technical solution: a method for denoising a terahertz time-domain spectrum signal of a transformer aging insulating oil comprises the following steps:

[0010] S1, using a whale optimization algorithm to optimize the parameters of the variational mode decomposition algorithm to obtain an applicable decomposition mode number K and an applicable penalty factor Alpha, and obtaining an optimized variational mode decomposition algorithm;

[0011] S2. Using the optimized variational mode decomposition algorithm, decompose the noisy terahertz time-domain spectral signal to obtain the intrinsic mode function;

[0012] S3, using independent component analysis to separate the intrinsic mode function into noise signal and effective signal;

[0013] S4. Calculate the information entropy of each independent component, set the sum of the mean and standard deviation of the information entropy as a threshold, and obtain the effective component of the terahertz time-domain spectral signal;

[0014] S5. Use the effective ingredients to reconstruct the noise-reduced terahertz time-domain spectral signal of aged insulating oil.

[0015] This paper combines VMD with independent component analysis (ICA) to complement their advantages. VMD has the advantage of simplifying the signal structure, but using VMD alone may result in an ineffective separation of noise and feature information, and the number of IMFs may be redundant. ICA, on the other hand, can further extract independent components from the IMF signal, effectively removing the noise component.

[0016] In a more specific technical solution, in S1, initial parameters are set, where the initial parameters include: population size, maximum number of iterations, and upper and lower limits of the search space.

[0017] In a more specific technical solution, in S1, a random initialization operation is performed in the search space to obtain an initial solution, which is used as the whale position;

[0018] Perform fitness evaluation on each whale position. If the current fitness value is better than the historical fitness value, the current fitness value is set as the optimal solution.

[0019] Update the whale positions. When all whales have completed their movement and evaluation, update the positions of all whales and repeat the random initialization, fitness evaluation, and update operations until the optimal solution is found.

[0020] The whale optimization algorithm adopted in the present invention simulates the predation behavior of whales, performs a global search in a given parameter space, optimizes the mode number K and penalty factor Alpha in the variational mode decomposition, optimizes the signal decomposition effect, improves the accuracy and robustness of the variational mode decomposition, and thus improves the noise reduction effect of the signal.

[0021] In a more specific technical solution, in S2, the variational mode decomposition algorithm is parameter-tuned according to the applicable decomposition mode number K and the applicable penalty factor Alpha.

[0022] In a more specific technical solution, in S2, the noisy terahertz time-domain spectrum signal is subjected to mirror expansion and Hilbert transform to remove the negative frequency part;

[0023] performing parameter initialization operation, initializing modal function center frequency Lagrange multiplier operator Initialize the number of cycles: n = 0;

[0024] Performing cycle: n = n + 1;

[0025] Performing parameter updating operation on modal function, center frequency and Lagrange multiplier operator;

[0026] Set the discrimination accuracy ε>0, repeatedly perform mirror expansion, Hilbert transform, parameter initialization operation and parameter updating operation until the iteration stopping condition is met:

[0027]

[0028] In a more specific technical solution, the parameter updating operation is performed according to the following formula:

[0029]

[0030]

[0031] In a more specific technical solution, in S3, the terahertz time domain spectrum data of the transformer aging insulating oil is subjected to decentralization operation and whitening operation.

[0032] The non-Gaussian maximization operation is performed on the terahertz time domain spectrum data.

[0033] The non-Gaussianity of the signal is measured by using the negative entropy and kurtosis to update the weight matrix to obtain the independent components of the source signal.

[0034] The present application further processes the IMFs obtained by VMD decomposition by using independent component analysis (ICA). The design basis comes from the independence principle in statistics, which assumes that all elements constituting the mixed signal are independent of each other and have no dependent relationship. Independent component analysis (ICA) is a blind signal separation method, and its core function is to decompose the mixed signal data into multiple independent and mutually statistically unrelated components.

[0035] In a more specific technical solution, the method of non-Gaussian maximization operation includes gradient descent iteration and fixed point iteration algorithm.

[0036] The system noise existing in the terahertz sampling signal is close to Gaussian distribution, but the part containing the characteristic signal often shows non-Gaussian property. The independent component analysis (ICA) used in the present application can extract the useful signal containing characteristic information by maximizing the non-Gaussianity of the output signal, thereby further improving the noise reduction effect.

[0037] In a more specific technical solution, in S4, the independent component with information entropy greater than the threshold is regarded as noise, and the independent component with information entropy less than the threshold is regarded as an effective component.

[0038] In a more specific technical solution, the terahertz time-domain spectrum signal denoising system of the transformer aging insulating oil comprises:

[0039] The parameter optimization module is used to use the whale optimization algorithm to perform parameter optimization on the variational mode decomposition algorithm, obtain the applicable decomposition mode number K and the applicable penalty factor Alpha, and obtain the optimized variational mode decomposition algorithm.

[0040] The signal decomposition module is used to decompose the noisy terahertz time-domain spectrum signal by using the optimized variational mode decomposition algorithm to obtain the intrinsic mode function, and the signal decomposition module is connected with the parameter optimization module.

[0041] The component analysis module is used to separate the intrinsic mode function into noise signals and effective signals by using independent component analysis, and the component analysis module is connected with the signal decomposition module.

[0042] The effective component acquisition module is used to calculate the information entropy of each independent component, set the sum of the mean value and the standard deviation of the information entropy as a threshold, and acquire the effective component of the terahertz time-domain spectrum signal, and the effective component acquisition module is connected with the component analysis module.

[0043] The signal reconstruction module is used to reconstruct the denoised terahertz time-domain spectrum signal of the aging insulating oil by using the effective component, and the signal reconstruction module is connected with the effective component acquisition module.

[0044] Compared with the prior art, the present application has the following advantages:

[0045] The present application combines VMD and independent component analysis (ICA) to complement each other. The advantage of VMD is that it can simplify the signal structure, but if VMD is used alone, it may cause noise and feature information to be effectively separated, and the number of IMFs may also be redundant, while ICA can further extract independent components from the IMF signal and effectively separate noise components.

[0046] The whale optimization algorithm used in the present application performs global search in a given parameter space by simulating the hunting behavior of whales, optimizes the mode number K and the penalty factor Alpha in the variational mode decomposition, makes the signal decomposition effect optimal, improves the accuracy and robustness of the variational mode decomposition, and thus improves the signal denoising effect.

[0047] The independent component analysis (ICA) adopted in the application can extract useful signals containing feature information by maximizing the non-Gaussianity of output signals, thereby further improving the noise reduction effect.

[0048] The independent component analysis (ICA) adopted in the application can extract useful signals containing feature information by maximizing the non-Gaussianity of output signals, thereby further improving the noise reduction effect.

[0049] The application solves the technical problems that the wavelet basis function and the number of decomposition layers are difficult to determine in the wavelet noise reduction, the end effect and mode aliasing of empirical mode decomposition restrict the noise reduction effect, and the noise reduction effect of variational mode decomposition is poor. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The figure is a basic step schematic diagram of the transformer aging insulating oil terahertz time-domain spectrum signal noise reduction method of embodiment 1 of the application;

[0051] Figure 2 The figure is a transformer aging insulating oil terahertz time-domain spectrum original signal diagram of embodiment 2 of the application;

[0052] Figure 3 The figure is a terahertz signal diagram after noise reduction obtained by the wavelet threshold noise reduction method of embodiment 2 of the application;

[0053] Figure 4 The figure is a terahertz signal diagram after noise reduction obtained by the variational mode decomposition noise reduction method of embodiment 2 of the application;

[0054] Figure 5 The figure is an intrinsic mode function diagram decomposed by the variational mode decomposition of embodiment 2 of the application;

[0055] Figure 6 The figure is an independent component diagram obtained by the independent component analysis of embodiment 2 of the application;

[0056] Figure 7 The figure is a terahertz time-domain spectrum signal diagram after noise reduction reconstructed by selecting appropriate independent components of embodiment 2 of the application;

[0057] Figure 8 The figure is a comparison diagram of the original terahertz time-domain spectrum signal and the terahertz time-domain spectrum signal after noise reduction of embodiment 2 of the application. DETAILED DESCRIPTION

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, the terahertz time-domain spectroscopy signal noise reduction method for transformer aged insulating oil provided by the present invention includes the following basic steps:

[0061] S1. Use the whale optimization algorithm to optimize the parameters of the variational mode decomposition to obtain the optimal decomposition mode number K and penalty factor Alpha;

[0062] In this embodiment, initial parameters are set, including but not limited to: population size, maximum number of iterations, upper and lower limits of the search space, and a number of solutions are randomly initialized in the search space. The solutions are the positions of the whales.

[0063] In this embodiment, the fitness of each whale's position is evaluated, and if the current fitness value is better than the previous fitness value, the current fitness value is set as the optimal solution;

[0064] In this embodiment, the positions of the whales are updated. When all whales have completed their movement and evaluation, the positions of all whales are updated, and the above steps are repeated until the optimal solution is found.

[0065] S2. Decompose the noisy terahertz time-domain spectral signal using the optimized variational mode decomposition algorithm to obtain a series of intrinsic mode functions;

[0066] In this embodiment, the eigenmode number K and penalty factor Alpha of the variational mode decomposition are set to the optimal parameter combination obtained in S1;

[0067] In this embodiment, the signal is mirror-extended and Hilbert transformed, and the negative frequency portion is removed;

[0068] In this embodiment, the initialization mode function Center frequency Lagrange multiplication operator Initialization loop times: n=0;

[0069] In this embodiment, a loop is executed, n=n+1;

[0070] In this embodiment, the following formula is used to update

[0071]

[0072] Set the discrimination accuracy ε>0 and repeat the above steps until the following iteration stop condition is met:

[0073]

[0074] S3, using independent component analysis to separate a series of intrinsic mode functions into noise and effective signals;

[0075] In this embodiment, independent component analysis is used to extract independent components from the intrinsic mode functions (IMFs) obtained by variational mode decomposition. The noise component exhibits strong Gaussianity and low independence, while the valid signal is highly non-Gaussian and independent. Through processing in step S3, the noise component and the valid signal component in the signal are separated.

[0076] In this embodiment, the terahertz time-domain spectroscopy data of transformer aged insulating oil is decentralized and whitened;

[0077] In this embodiment, the non-Gaussianity of the signal is maximized by gradient descent iteration or fixed point iteration algorithm;

[0078] In this embodiment, negative entropy or kurtosis is used as a means to measure the non-Gaussianity of the signal to update the weight matrix, thereby obtaining the independent components of the source signal.

[0079] S4. Calculate the information entropy of each independent component, set the mean plus the standard deviation of the information entropy of the independent component as the threshold, and regard the independent components with information entropy greater than the threshold as noise, and the independent components with information entropy less than the threshold as effective components;

[0080] In this embodiment, the information entropy of each independent component is calculated;

[0081] In this embodiment, the threshold is calculated by adding the mean of the independent component information entropy and the standard deviation to set the threshold;

[0082] In this embodiment, independent components with information entropy greater than a threshold are regarded as noise, and independent components with information entropy less than the threshold are regarded as effective components.

[0083] S5. Use effective ingredients to reconstruct the terahertz time-domain spectral signal of aged insulating oil after noise reduction.

[0084] Example 2

[0085] like Figure 2As shown, in order to target practical applications, in this embodiment, the transformer insulating oil subjected to 196h thermal aging is used as a sample, and experimental data is obtained using a terahertz transmission device. Observation of the local magnified image shows that the signal is mixed with obvious noise.

[0086] like Figure 3 As shown in FIG, the denoised signal is obtained by using wavelet threshold denoising. In this embodiment, the db5 wavelet basis is used, the number of decomposition layers is 5, and the soft threshold method is selected for denoising. Due to the difficulty in selecting the wavelet basis and the number of decomposition layers, the wavelet denoising effect in this experiment is poor, and obvious noise still exists locally.

[0087] like Figure 4 As shown in Figure 1, the denoised signal is obtained by using only the variational mode decomposition denoising method. Since some intrinsic mode functions may contain noise signals, this part of the noise is reconstructed into the denoised signal when reconstructing the signal, resulting in obvious noise still being mixed in the denoised signal.

[0088] like Figure 5 As shown, the noise reduction method proposed in the present invention is adopted: first, the whale optimization algorithm is used to optimize the mode number K and the penalty factor Alpha, and the optimization result is K=5, Alpha=3381; the optimized variational mode decomposition is used to decompose the noisy terahertz signal, and the decomposition result is shown in FIG. Figure 5 ;

[0089] like Figure 6 As shown, in this embodiment, independent component analysis is used to separate a series of intrinsic mode functions into noise and effective signals, independent components;

[0090] like Figure 7 As shown, in this embodiment, the independent component information entropy and the threshold are calculated, and the independent components greater than the threshold are selected to reconstruct the terahertz signal of the transformer aging insulating oil after noise reduction.

[0091] like Figure 8 As shown in FIG. 1 , in this embodiment, the terahertz time domain spectroscopy signals before and after noise reduction are compared. As can be seen from the above figures, the noise reduction method of the present invention has a good noise reduction effect on the terahertz time domain spectroscopy signal of the aged insulating oil of the transformer.

[0092] In this embodiment, the sparrow search algorithm (SSA) is used to optimize the VMD parameters (K and Alpha), while the whale optimization algorithm (WOA) is used in this application. Although both are swarm intelligence optimization algorithms, the differences in the algorithm mechanisms will lead to different parameter optimization effects. In this embodiment, the optimization effects of SSA and WOA are compared on the same signal (see the table below). The results show that the VMD parameter combination optimized by WOA has a better noise reduction effect in terms of signal-to-noise ratio (SNR), mean square error (MSE), and waveform similarity coefficient (NCC) after noise reduction than the SSA optimization parameter combination.

[0093]

[0094] In summary, this paper combines VMD with independent component analysis (ICA) to complement their advantages. VMD has the advantage of simplifying the signal structure, but using VMD alone may result in an ineffective separation of noise and feature information, and the number of IMFs may be redundant. ICA, on the other hand, can further extract independent components from the IMF signal, effectively removing the noise component.

[0095] The whale optimization algorithm adopted in the present invention simulates the predation behavior of whales, performs a global search in a given parameter space, optimizes the mode number K and penalty factor Alpha in the variational mode decomposition, optimizes the signal decomposition effect, improves the accuracy and robustness of the variational mode decomposition, and thus improves the noise reduction effect of the signal.

[0096] The present invention further processes the IMFs derived from VMD decomposition using independent component analysis (ICA). Its design is based on the independence principle in statistics, which assumes that all elements that make up a mixed signal are independent of each other and have no dependencies. Independent component analysis (ICA) is a blind signal separation method whose core function is to decompose mixed signal data into multiple independent and statistically uncorrelated components.

[0097] The independent component analysis (ICA) adopted in the present invention can extract useful signals containing feature information by maximizing the non-Gaussianity of the output signal, thereby further improving the noise reduction effect.

[0098] The present invention solves the technical problems in the prior art that the wavelet basis function and the number of decomposition layers of wavelet denoising are difficult to determine, the endpoint effect and modal aliasing of empirical mode decomposition restrict the denoising effect, and the variational mode decomposition denoising effect is poor.

[0099] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A terahertz time-domain spectroscopy signal noise reduction method for transformer aging insulating oil, characterized in that: The method comprises: S1. Use the whale optimization algorithm to optimize the parameters of the variational mode decomposition algorithm, obtain the applicable decomposition mode number K and the applicable penalty factor Alpha, and obtain the optimized variational mode decomposition algorithm; S2. Decomposing the noisy terahertz time-domain spectrum signal using the optimized variational mode decomposition algorithm to obtain an intrinsic mode function; S3. Using independent component analysis, the intrinsic mode function is separated into a noise signal and a valid signal; S4. Calculate the information entropy of each independent component, set the sum of the mean and standard deviation of the information entropy as a threshold, and obtain the effective component of the terahertz time-domain spectroscopy signal; S5. Reconstruct a noise-reduced terahertz time-domain spectrum signal of the aged insulating oil using the effective component.

2. The method for denoising terahertz time-domain spectroscopy signals of transformer aged insulating oil according to claim 1, characterized in that: In S1, initial parameters are set, wherein the initial parameters include: population size, maximum number of iterations, and upper and lower limits of the search space.

3. The method for denoising terahertz time-domain spectroscopy signals of transformer aged insulating oil according to claim 1, characterized in that: In S1, a random initialization operation is performed in the search space to obtain an initial solution, and the initial solution is used as the whale position; Performing a fitness evaluation operation on each of the whale positions, and if a current fitness value is better than a historical fitness value, setting the current fitness value as the optimal solution; The whale positions are updated. When all whales have completed movement and evaluation, the positions of all whales are updated, and the random initialization operation, the fitness evaluation operation, and the update operation are repeated until the optimal solution is found.

4. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 1, characterized in that: In S2, the variational mode decomposition algorithm is parameter-tuned according to the applicable decomposition mode number K and the applicable penalty factor Alpha.

5. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 1, characterized in that: In S2, the noisy terahertz time-domain spectrum signal is subjected to mirror expansion and Hilbert transformation to remove the negative frequency portion; Perform parameter initialization and initialize the modal function Center frequency {w 1 k}, Lagrange multiplication operator Initialization loop times: n=0; Execute loop: n=n+1; performing a parameter update operation on the mode function, the center frequency, and the Lagrange multiplication operator; Set the discrimination accuracy ε>0, and repeatedly perform the mirror expansion, the Hilbert transform, the parameter initialization operation, and the parameter update operation until the iteration stop condition is met:

6. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 5, characterized in that: The parameter update operation is performed according to the following formula:

7. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 1, characterized in that: In the S3, a decentralization operation and a whitening operation are performed on the terahertz time-domain spectrum data of the transformer aged insulating oil; performing a non-Gaussian maximization operation on the terahertz time-domain spectroscopy data; Negative entropy and kurtosis are used to measure the non-Gaussianity of the signal to update the weight matrix and obtain the independent components of the source signal.

8. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 7, characterized in that: The method of the non-Gaussian maximization operation includes: gradient descent iteration and fixed point iteration algorithm.

9. The method for denoising terahertz time-domain spectroscopy signals of aged transformer insulating oil according to claim 1, characterized in that: In S4, the independent components whose information entropy is greater than the threshold are regarded as noise, and the independent components whose information entropy is less than the threshold are regarded as effective components.

10. A terahertz time-domain spectroscopy signal noise reduction system for transformer aging insulating oil, characterized in that: The system comprises: The parameter optimization module is used to use the whale optimization algorithm to optimize the parameters of the variational mode decomposition algorithm, obtain the applicable decomposition mode number K, the applicable penalty factor Alpha, and obtain the optimized variational mode decomposition algorithm; A signal decomposition module, configured to use the optimized variational mode decomposition algorithm to decompose the noisy terahertz time-domain spectrum signal to obtain an intrinsic mode function, wherein the signal decomposition module is connected to the parameter optimization module; a component analysis module, configured to separate the intrinsic mode function into a noise signal and a valid signal by using independent component analysis, wherein the component analysis module is connected to the signal decomposition module; an effective component acquisition module, configured to calculate the information entropy of each independent component, set the sum of the mean and standard deviation of the information entropy as a threshold, and obtain the effective component of the terahertz time-domain spectroscopy signal; the effective component acquisition module is connected to the component analysis module; A signal reconstruction module is used to reconstruct the noise-reduced terahertz time-domain spectrum signal of the aged insulating oil using the effective component, and the signal reconstruction module is connected to the effective component acquisition module.

Images