Insulating oil Raman spectrum denoising method, system, device and storage medium

Abstract

The present invention provides a method, system, device, and storage medium for denoising Raman spectra of insulating oil. The method comprises obtaining Raman spectra data of insulating oil; constructing a Raman spectra data denoising model based on the Raman spectra data, so that the Raman spectra data denoising model eliminates various interference signals in the Raman spectra data while maintaining the spectral peak characteristics of the Raman spectra data unchanged; and performing segmented denoising on the Raman spectra data using the Raman spectra data denoising model to obtain smoothed denoised Raman spectra data. The denoising method provided by the present invention eliminates the need for manual parameter adjustment, effectively avoiding the introduction of human error. Furthermore, the Raman spectra data denoising model performs segmented denoising on the Raman spectra data, reducing the interference of noise on the Raman spectra signal and improving the accuracy of insulating oil Raman spectra data for insulating oil analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of spectral data analysis, and in particular to a method, system, device and storage medium for denoising an insulating oil Raman spectrum. Background Art

[0002] Raman spectroscopy is a spectroscopic analysis method that uses molecular vibrational and rotational information. Due to its high sensitivity to subtle chemical and biochemical changes and its adaptability to non-invasive applications, this technique has been widely used for structural detection and property analysis of solid and liquid materials.

[0003] However, when Raman spectroscopy is used to detect substances, in addition to obtaining Raman signals containing the substance to be detected, it will also generate noise interference due to factors such as laser fluctuations. The noise signal of the Raman spectroscopy signal mainly comes from the CCD array detector, which includes photon shot noise and dark noise. Photon shot noise is a statistical error that occurs when the CCD detector collects photons. Its essence is that the light intensity measured by the CCD can give the average number of collected photons, but it cannot know the actual number of photons collected at any time. When the noise amplitude is large, it will cause jitter in the Raman spectrum, resulting in burrs and spikes, and the signal-to-noise ratio of the spectrum will be reduced, which will seriously affect the extraction of Raman characteristic peaks and the identification of the substance to be detected. In particular, in the analysis of substances such as insulating oil, it will significantly reduce the accuracy of Raman spectroscopy data analysis.

[0004] Currently, the main methods for reducing Raman spectroscopy signal noise include Savitzky-Golay smoothing filtering, FFT filtering, penalized least squares (PLS) and threshold wavelet denoising. However, in practical applications, these methods are prone to losing Raman information of the sample being measured, and parameter adjustment requires manual effort, which can lead to human errors in the final processing results and substantially affect the accuracy of Raman spectroscopy signal analysis. Summary of the Invention

[0005] The present invention aims to provide a method, system, device and storage medium for denoising insulating oil Raman spectra to solve the above-mentioned technical problems. The method can effectively avoid the introduction of human errors and reduce the interference of noise on Raman spectra signals, thereby improving the accuracy of insulating oil Raman spectra data for insulating oil analysis.

[0006] In order to solve the above technical problems, the present invention provides a method for denoising the Raman spectrum of insulating oil, comprising the following steps:

[0007] Obtain Raman spectrum data of insulating oil;

[0008] A Raman spectral data denoising model is constructed based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged;

[0009] The Raman spectrum data denoising model is used to perform segmented denoising on the Raman spectrum data to obtain smoothed denoised Raman spectrum data.

[0010] In the above scheme, the Raman spectral data of the insulating oil is smoothed and denoised by constructing a Raman spectral data denoising model that can maintain the spectral peak characteristics of the Raman spectral data and eliminate various interference signals in the Raman spectral data. The entire smoothing and denoising process can be achieved simply by inputting the Raman spectral data of the insulating oil into the Raman spectral data denoising model, without the need for manual parameter adjustment, thus effectively avoiding the introduction of human errors. At the same time, the Raman spectral data denoising model performs segmented denoising on the Raman spectral data, which can reduce the interference of noise on the Raman spectral signal and improve the accuracy of the insulating oil Raman spectral data for insulating oil analysis.

[0011] Furthermore, the Raman spectrum data denoising model is constructed based on the Raman spectrum data, so that the Raman spectrum data denoising model can eliminate various interference signals in the Raman spectrum data while maintaining the spectral peak characteristics of the Raman spectrum data unchanged, specifically:

[0012] The first n data points of the Raman spectrum data are used as input and the n+1th data point is used as output to establish a first denoising model; wherein n is a constant;

[0013] Take n data points to the n+2th data point as input and the n+3th data point as output to establish the second denoising model;

[0014] This process is repeated until the last data point of the denoising model is reached, and several denoising models are established.

[0015] A Raman spectral data denoising model is constructed based on several denoising models, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while keeping the spectral peak characteristics of the Raman spectral data unchanged.

[0016] In this approach, multiple denoising models are established by traversing all Raman spectral data points. This ensures that the Raman spectrum remains linearly continuous during the denoising process, enhancing the denoising effect. This makes the Raman spectral data denoising model constructed from multiple denoising models more effective for denoising insulating oil Raman spectral data.

[0017] Furthermore, the Raman spectrum data denoising model is constructed based on the Raman spectrum data, so that the Raman spectrum data denoising model can eliminate various interference signals in the Raman spectrum data while maintaining the spectral peak characteristics of the Raman spectrum data unchanged, specifically:

[0018] A Raman spectral data denoising model is constructed based on the Raman spectral data. The Raman spectral data denoising model is specifically as follows:

[0019] y(k)=g(w3x(k)+b2)

[0020] x(k)=f(w1x C (k)+w2(u(k-1))+b1)

[0021] x C (k) = x(k-1)

[0022] Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons;

[0023] The Raman spectral data denoising model eliminates various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged.

[0024] In this solution, the Raman spectroscopy data denoising model incorporates a special bridging layer, which enables internal feedback, storage, and utilization of past output information. Combined with the consistent linear characteristics of all data points in an ideal spectral segment, this bridging layer allows the Raman spectroscopy data denoising model to eliminate various interfering signals while maintaining the original spectral peaks. This model eliminates the need for human intervention and effectively reduces the impact of noise.

[0025] Furthermore, the transfer function of the output layer neurons adopts purelin function, and the transfer function of the hidden layer neurons adopts tansig function.

[0026] In the above scheme, the purelin function is a linear transfer function. Since its input and output can take arbitrary values, it is highly compatible with the characteristics of Raman spectroscopy data denoising models, improving their efficiency. The tansig function, with its input taking arbitrary values and its output between -1 and +1, has a low error, effectively ensuring the accuracy of the Raman spectroscopy data denoising model.

[0027] The present invention also provides an insulating oil Raman spectrum denoising system, comprising a data acquisition module, a denoising model building module and a smoothing denoising module; wherein:

[0028] The data acquisition module is used to acquire Raman spectrum data of insulating oil;

[0029] The denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged;

[0030] The smoothing and denoising module is used to perform segmented denoising on the Raman spectrum data using a Raman spectrum data denoising model to obtain smoothed and denoised Raman spectrum data.

[0031] Furthermore, the denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged, specifically:

[0032] The first n data points of the Raman spectrum data are used as input and the n+1th data point is used as output to establish a first denoising model; wherein n is a constant;

[0033] Take n data points to the n+2th data point as input and the n+3th data point as output to establish the second denoising model;

[0034] This process is repeated until the last data point of the denoising model is reached, and several denoising models are established.

[0035] A Raman spectral data denoising model is constructed based on several denoising models, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while keeping the spectral peak characteristics of the Raman spectral data unchanged.

[0036] Furthermore, the denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged, specifically:

[0037] A Raman spectral data denoising model is constructed based on the Raman spectral data. The Raman spectral data denoising model is specifically as follows:

[0038] y(k)=g(w3x(k)+b2)

[0039] x(k)=f(w1x C (k)+w2(u(k-1))+b1)

[0040] x C (k) = x(k-1)

[0041] Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons;

[0042] The Raman spectral data denoising model eliminates various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged.

[0043] Furthermore, in the denoising model building module, a purelin function is used as a transfer function of the output layer neurons.

[0044] The present invention also provides an insulating oil Raman spectrum denoising device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, an insulating oil Raman spectrum denoising method is implemented.

[0045] The present invention also provides a storage medium, which includes a stored computer program. When the computer program is running, the device where the storage medium is located is controlled to execute a method for denoising the Raman spectrum of insulating oil. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic flow chart of a method for denoising the Raman spectrum of insulating oil proposed in one embodiment of the present invention;

[0047] Figure 2 This is a graph of original Raman spectrum data proposed in one embodiment of the present invention;

[0048] Figure 3 A schematic diagram of the network structure of a Raman spectroscopy data denoising model proposed in one embodiment of the present invention;

[0049] Figure 4This is a graph of Raman spectrum data after noise reduction proposed in one embodiment of the present invention;

[0050] Figure 5 This is a module connection diagram of an insulating oil Raman spectrum denoising system proposed in one embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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.

[0052] See Figure 1 This embodiment provides a method for denoising an insulating oil Raman spectrum, comprising the following steps:

[0053] S1: Acquire Raman spectrum data of insulating oil;

[0054] S2: constructing a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged;

[0055] S3: Using the Raman spectrum data denoising model to perform segmented denoising on the Raman spectrum data to obtain smoothed denoised Raman spectrum data.

[0056] In this embodiment, the Raman spectral data of the insulating oil is smoothed and denoised by constructing a Raman spectral data denoising model that can maintain the spectral peak characteristics of the Raman spectral data and eliminate various interference signals in the Raman spectral data. The entire smoothing and denoising process can be achieved simply by inputting the Raman spectral data of the insulating oil into the Raman spectral data denoising model, eliminating the need for manual parameter adjustment and effectively avoiding the introduction of human errors. At the same time, the Raman spectral data denoising model performs segmented denoising on the Raman spectral data, which can reduce the interference of noise on the Raman spectral signal and improve the accuracy of the insulating oil Raman spectral data for insulating oil analysis.

[0057] Furthermore, the Raman spectrum data denoising model is constructed based on the Raman spectrum data, so that the Raman spectrum data denoising model can eliminate various interference signals in the Raman spectrum data while maintaining the spectral peak characteristics of the Raman spectrum data unchanged, specifically:

[0058] The first n data points of the Raman spectrum data are used as input and the n+1th data point is used as output to establish a first denoising model; wherein n is a constant;

[0059] Take n data points to the n+2th data point as input and the n+3th data point as output to establish the second denoising model;

[0060] This process is repeated until the last data point of the denoising model is reached, and several denoising models are established.

[0061] A Raman spectral data denoising model is constructed based on several denoising models, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while keeping the spectral peak characteristics of the Raman spectral data unchanged.

[0062] In this embodiment, multiple denoising models are established by traversing all Raman spectral data points. This ensures that the Raman spectrum remains linearly continuous during the denoising process, thereby enhancing the denoising effect. This makes the Raman spectral data denoising model constructed from multiple denoising models more effective for denoising insulating oil Raman spectral data.

[0063] Furthermore, the Raman spectrum data denoising model is constructed based on the Raman spectrum data, so that the Raman spectrum data denoising model can eliminate various interference signals in the Raman spectrum data while maintaining the spectral peak characteristics of the Raman spectrum data unchanged, specifically:

[0064] A Raman spectral data denoising model is constructed based on the Raman spectral data. The Raman spectral data denoising model is specifically as follows:

[0065] y(k)=g(w3x(k)+b2)

[0066] x(k)=f(w1x C (k)+w2(u(k-1))+b1)

[0067] x C (k) = x(k-1)

[0068] Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons;

[0069] The Raman spectral data denoising model eliminates various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged.

[0070] In this embodiment, the Raman spectroscopy data denoising model incorporates a special bridging layer, which enables internal feedback, storage, and utilization of past output information. Combined with the consistent linear characteristics of all data points in an ideal spectral segment, this bridging layer allows the Raman spectroscopy data denoising model to eliminate various interfering signals in the Raman spectroscopy data while maintaining the original spectral peaks. This model eliminates the need for human intervention and effectively reduces the impact of noise.

[0071] Furthermore, the transfer function of the output layer neurons adopts purelin function, and the transfer function of the hidden layer neurons adopts tansig function.

[0072] In this embodiment, the purelin function is a linear transfer function. Because its input and output can take arbitrary values, it is highly compatible with the characteristics of Raman spectroscopy data denoising models, improving the efficiency of building Raman spectroscopy data denoising models. The tansig function, with its input and output values between -1 and +1, has a low error, effectively ensuring the accuracy of the Raman spectroscopy data denoising model.

[0073] In this embodiment, the tansig function expression is:

[0074]

[0075] In this embodiment, the purelin function expression is:

[0076] purelin(x)=x

[0077] In order to further illustrate the technical solution of the present invention and highlight its purpose and technical effects, this embodiment provides a specific implementation process of the present invention.

[0078] This embodiment provides a method for denoising an insulating oil Raman spectrum, comprising:

[0079] S1: Obtain Raman spectrum data of insulating oil, specifically:

[0080] Based on the actual operating insulating oil samples collected, the insulating oil samples were monitored through the laser Raman spectroscopy detection platform equipped by the State Key Laboratory of Special Power Transmission and Distribution Equipment and System Safety and New Technologies to obtain the Raman spectral data of the insulating oil. Please see the case for the diagram. Figure 2 .

[0081] Since the Raman spectrum data of insulating oil contains noise signals, which affects the accuracy of spectrum data analysis, in order to eliminate or weaken the influence of noise, it is necessary to perform smoothing and noise reduction on the original Raman spectrum data.

[0082] S2: constructing a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged;

[0083] In this embodiment, a Raman spectrum data denoising model can be constructed based on the Raman spectrum data on the basis of the Elman network structure. The first four data points in the Raman spectrum data (a total of 1024 data points) are used as input, and the fifth data point is used as output to establish a first denoising model with a topological structure of 4-1-1-1. Then the third, fourth, fifth, and sixth data points (wherein the fifth data point is the output of the first denoising model) are used as input, and the seventh data point is used as output to establish a second denoising model with a topological structure of 4-1-1-1. Then the fifth, sixth, seventh, and eighth data points (wherein the fifth data point is the output of the first denoising model, and the seventh data point is the output of the second denoising model) are used as input, and the ninth data point is used as output to establish a third denoising model with a topological structure of 3-1-1-1. By analogy, a total of 511 denoising models are established; a Raman spectrum data denoising model is constructed based on the established denoising models.

[0084] It should be noted that the Elman network structure used in this embodiment adds a carrier layer to the BP neural network. Therefore, the Raman spectroscopy data denoising model constructed by it can internally feedback, store, and utilize output information from past time periods. Combined with the fact that all points in an ideal spectral segment have consistent linear characteristics, the Raman spectroscopy data denoising model can be used to eliminate various interference signals in the Raman spectroscopy data while maintaining the original spectral peak characteristics.

[0085] Furthermore, the network structure of the Raman spectroscopy data denoising model constructed in this embodiment can be found in Figure 3 The specific model expression is:

[0086] y(k)=g(w3x(k)+b2)

[0087] x(k)=f(w1x C (k)+w2(u(k-1))+b1)

[0088] x C (k) = x(k-1)

[0089] S3: Using the Raman spectrum data denoising model to perform segmented denoising on the Raman spectrum data to obtain smoothed denoised Raman spectrum data.

[0090] In this embodiment, the first four data points of the Raman spectrum to be denoised (a total of 1024 data points) are used as input and substituted into the first denoising model to obtain the fifth data point. Then the third, fourth, fifth, and sixth data points (where the fifth data point is the output of the first denoising model) are used as input and substituted into the second denoising model to obtain the seventh data point. Then the fifth, sixth, seventh, and eighth data points (where the fifth data point is the output of the first denoising model and the seventh data point is the output of the second denoising model) are used as input and substituted into the third denoising model to obtain the ninth data point. And so on, the Raman spectrum is denoised. Finally, the data points are integrated to obtain the final smoothed and denoised Raman spectrum data, for which a diagram can be seen. Figure 4 .

[0091] from Figure 2 、 Figure 4 Comparison shows that the insulating oil Raman spectrum denoising method proposed in this example effectively removes noise interference signals from Raman spectral data, highlights the characteristic characteristics of the measured component signals, ensures the validity of the spectral data, improves the signal-to-noise ratio of the original spectral signal, and effectively enhances the accuracy of insulating oil spectral data analysis. This method is simple, fast, requires a small number of parameters, and avoids the introduction of human errors during the calculation process.

[0092] See Figure 5 This embodiment provides an insulating oil Raman spectrum denoising system for implementing an insulating oil Raman spectrum denoising method, comprising a data acquisition module, a denoising model construction module, and a smoothing denoising module; wherein:

[0093] The data acquisition module is used to acquire Raman spectrum data of insulating oil;

[0094] The denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged;

[0095] The smoothing and denoising module is used to perform segmented denoising on the Raman spectrum data using a Raman spectrum data denoising model to obtain smoothed and denoised Raman spectrum data.

[0096] Furthermore, the denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged, specifically:

[0097] The first n data points of the Raman spectrum data are used as input and the n+1th data point is used as output to establish a first denoising model; wherein n is a constant;

[0098] Take n data points to the n+2th data point as input and the n+3th data point as output to establish the second denoising model;

[0099] This process is repeated until the last data point of the denoising model is reached, and several denoising models are established.

[0100] A Raman spectral data denoising model is constructed based on several denoising models, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while keeping the spectral peak characteristics of the Raman spectral data unchanged.

[0101] Furthermore, the denoising model building module is used to build a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged, specifically:

[0102] A Raman spectral data denoising model is constructed based on the Raman spectral data. The Raman spectral data denoising model is specifically as follows:

[0103] y(k)=g(w3x(k)+b2)

[0104] x(k)=f(w1x C (k)+w2(u(k-1))+b1)

[0105] x C (k) = x(k-1)

[0106] Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons;

[0107] The Raman spectral data denoising model eliminates various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged.

[0108] Furthermore, in the denoising model building module, a purelin function is used as a transfer function of the output layer neurons.

[0109] This embodiment provides an insulating oil Raman spectrum denoising device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, an insulating oil Raman spectrum denoising method is implemented.

[0110] This embodiment provides a storage medium, which includes a stored computer program. When the computer program is executed, the device where the storage medium is located is controlled to execute a method for denoising insulating oil Raman spectra.

[0111] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for denoising insulating oil Raman spectra, characterized in that: The following steps are involved: Obtain Raman spectrum data of insulating oil; A Raman spectral data denoising model is constructed based on Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged. Specifically, the first n data points of the Raman spectral data are taken as input and the n+1th data point is taken as output to establish a first denoising model; n is a constant; n data points to the n+2th data point are taken as input and the n+3th data point is taken as output to establish a second denoising model; and so on until the last data point of the denoising model, and several denoising models are established; the Raman spectral data denoising model is specifically: y(k)=g(w3x(k)+b2) x(k)=f(w1x C (k)+w2(u(k-1))+b1) x C (k)=x(k-1) Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons; The Raman spectrum data denoising model is used to perform segmented denoising on the Raman spectrum data to obtain smoothed denoised Raman spectrum data.

2. The insulating oil Raman spectrum denoising method according to claim 1, characterized in that: The transfer function of the output layer neurons adopts purelin function.

3. An insulating oil Raman spectrum denoising system, characterized in that: It includes data acquisition module, denoising model construction module and smoothing denoising module; among which: The data acquisition module is used to acquire Raman spectrum data of insulating oil; The denoising model construction module is used to construct a Raman spectral data denoising model based on the Raman spectral data, so that the Raman spectral data denoising model can eliminate various interference signals in the Raman spectral data while maintaining the spectral peak characteristics of the Raman spectral data unchanged. Specifically, the first n data points of the Raman spectral data are taken as input and the n+1th data point is taken as output to establish a first denoising model; n is a constant; n data points to the n+2th data point are taken as input and the n+3th data point is taken as output to establish a second denoising model; and so on until the last data point of the denoising model is reached, and several denoising models are established. The Raman spectral data denoising model is specifically: y(k)=g(w3x(k)+b2) x(k)=f(w1x C (k)+w2(u(k-1))+b1) x C (k)=x(k-1) Where: y(k) represents the output at time k; u represents the input; x represents the output of the hidden layer; x C is the feedback vector; w1 is the connection weight between the hidden layer and the receiving layer; w2 is the connection weight between the hidden layer and the input layer; w3 is the connection weight between the hidden layer and the output layer; b1 and b2 are the thresholds of the output layer and the hidden layer respectively; g(·) is the transfer function of the output layer neurons; f(·) is the transfer function of the hidden layer neurons; The smoothing and denoising module is used to perform segmented denoising on the Raman spectrum data using a Raman spectrum data denoising model to obtain smoothed and denoised Raman spectrum data.

4. The insulating oil Raman spectrum denoising system according to claim 3, characterized in that: In the denoising model building module, the purelin function is used as the transfer function of the output layer neurons.

5. An insulating oil Raman spectrum denoising device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, characterized in that: When the processor executes the computer program, the insulating oil Raman spectrum denoising method according to any one of claims 1 to 2 is implemented.

6. A storage medium comprising a stored computer program, characterized in that: When the computer program is running, the device where the storage medium is located is controlled to execute the insulating oil Raman spectrum denoising method according to any one of claims 1 to 2.

Images