Transformer oil spectrum characteristic value extraction and light intensity correction method

Abstract

The present application relates to a kind of transformer oil spectrum characteristic value extraction and light intensity correction method, the method is by low-pass filtering to original spectrum absorption signal to remove noise;Then the current spectrum signal amplitude is extracted from non-absorption section, which reflects the size of light intensity in real time;While extracting the upper envelope of signal and filtering again to obtain smooth upper envelope, on this basis, the absorption starting point and end point are accurately identified, connect both to form baseline, and calculate the difference absolute value of smooth upper envelope and baseline in absorption interval, as the characteristic value related to gas concentration;When calibration, record calibration spectrum signal amplitude, when actual measurement, the characteristic value of actual measurement is divided by calibration spectrum absorption signal amplitude, then multiplied by current spectrum signal amplitude, to obtain the characteristic value after light intensity correction, the correction process directly eliminates the influence of light intensity fluctuation on characteristic value, thereby guaranteeing the accuracy of concentration inversion, significantly improve the stability and reliability of direct absorption method in long-term online monitoring.

Description

Technical Field

[0001] This invention relates to the field of online monitoring of transformer oil spectra, specifically to a method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil. Background Technology

[0002] Transformers are core equipment in power systems, and their operating status directly affects the safety and stability of the power grid. Statistics show that insulation faults are one of the main causes of transformer failure. Currently, oil-immersed transformers are widely used in power systems. Under the influence of electricity, heat, oxygen, and moisture, the insulating oil gradually ages and decomposes, producing small amounts of characteristic gases that dissolve in the oil. Research indicates that different fault types correspond to different gas compositions and concentrations: for example, oil overheating mainly produces H2, CH4, C2H4, and C2H6; solid insulation material overheating mainly produces CO and CO2; and arc discharge, spark discharge, and partial discharge produce gases such as C2H2, H2, and CH4. Acetylene (C2H2) is an important indicator gas for discharge faults; a concentration of 1 ppm in the oil should trigger an alarm. Therefore, online monitoring of dissolved gases in transformer oil can enable early warning and fault type identification.

[0003] Currently, the main technologies for detecting dissolved gases in transformer oil include gas chromatography, photoacoustic spectroscopy, and tunable diode laser absorption spectroscopy (TDLAS). Among these, TDLAS technology is widely used in industrial and environmental gas detection fields due to its advantages such as high sensitivity, high selectivity, fast response, and non-contact measurement. The core theoretical basis of TDLAS technology is the Lambert-Beer law. This law describes the attenuation of monochromatic light after passing through an absorbing medium: light intensity is... , frequency is Monochromatic laser light, through an absorption layer with a thickness of After passing through a uniform absorbing medium, the emitted light intensity can be expressed as:

[0004] In the formula, The intensity of the emitted light; The incident light intensity, i.e., the frequency is The light intensity of a monochromatic laser before it enters the absorbing medium; The concentration of the gas being measured (the number of molecules per unit volume). For frequency The absorption cross section of the medium is related to temperature and pressure; The absorbance is the thickness of the absorption layer, i.e., the effective path length of light through the gas. Defined as the logarithm of the ratio of incident light intensity to emitted light intensity:

[0005] From the above formula, it can be seen that the thickness of the absorption layer (i.e. Given the known conditions, the gas concentration can be calculated by measuring the intensity of the emitted light.

[0006] Another foundation of TDLAS is harmonic detection theory. Harmonic detection theory utilizes a high-frequency cosine signal added to a current to modulate the output of a semiconductor laser.

[0007] in, The injection current varies with time. This is the center current value, i.e., the operating bias current of the laser. The modulation amplitude refers to the AC modulation amplitude of the injected current. The modulation frequency is the angular frequency of the injected current. For time. Correspondingly, the instantaneous output frequency of the laser. for:

[0008] In the formula, The laser center frequency is the laser frequency without modulation. The frequency modulation amplitude is the amplitude of the laser frequency change caused by current modulation. The light intensity after gas absorption can be expanded into a Fourier series:

[0009] In the formula, The light intensity signal changes over time after being absorbed by the gas. For the first The amplitude of the second harmonic component is the center frequency. The function; For harmonic order; For the first The cosine function of the second harmonic.

[0010] Harmonic components It can be measured by a lock-in amplifier, and its expression is:

[0011] In the formula, For integration variables, ; The instantaneous incident light intensity of the laser output at the modulation frequency; The gas absorption cross section at the instantaneous frequency; the absorption of light intensity by trace gases in the White cell. In this case, the above formula can be simplified to:

[0012] in, It is the product of gas concentration and absorption path length.

[0013] Therefore, the amplitude of each harmonic component Proportional to trace gases. When the linewidth is much smaller than the absorption linewidth, perform a Taylor expansion on the above equation:

[0014] in, for factorial; These are constant coefficients related to the harmonic order; For modulation amplitude Power; Absorption cross section At the center frequency place The first derivative.

[0015] It can be seen that the first The subharmonic component is proportional to of Second derivative. Because of the factorial in the denominator of the above equation, the amplitude of the absorption spectral line decreases rapidly with increasing harmonic order. Therefore, harmonics with excessively high orders should not be selected when calculating concentration. Furthermore, odd-order harmonics are odd-symmetric about the central absorption position, with an amplitude of 0 at the absorption center. However, the presence of baseline interference makes it difficult to locate the central absorption position. Even-order harmonics are even-symmetric about the central absorption position, with the largest amplitude at the central absorption position, making them more advantageous for searching the central absorption position. Generally, the second harmonic is selected for concentration calculation based on factors such as anti-interference capability and amplitude magnitude. See [link to relevant documentation]. Figure 1 .

[0016] In practical applications, TDLAS often incorporates harmonic detection technology to improve signal-to-noise ratio and anti-interference capabilities. By superimposing a high-frequency cosine modulation signal onto the laser's injected current, the laser output frequency changes periodically with the modulation signal. After gas absorption, the transmitted light intensity contains harmonic components. Among these, the second harmonic (2f) signal is the most commonly used basis for concentration calculation due to its maximum amplitude at the absorption center, ease of peak search, and strong anti-interference capability. Specifically, during calibration, a standard gas of known concentration is introduced into a gas chamber (such as a White cell) to establish a linear relationship between the second harmonic peak-valley value and the concentration. During actual measurement, the second harmonic peak-valley value of the gas to be measured is measured, and the concentration is obtained by substituting it into the calibration curve.

[0017] However, during long-term operation, the laser output power may attenuate, or the emitted light intensity may fluctuate due to factors such as contamination of the gas chamber window or optical path offset. Changes in light intensity directly affect the amplitude of the second harmonic signal, leading to deviations in concentration measurement results from the true value. To suppress the influence of light intensity fluctuations, existing technologies often use a normalization method by dividing the second harmonic signal by the first harmonic signal, thereby partially offsetting the error caused by light intensity changes. This method improves measurement stability to some extent, but still has the following limitations: the first harmonic signal itself is affected by factors such as circuit noise and modulation depth, and the normalization effect is not entirely ideal; furthermore, this method is only applicable to harmonic detection modes and cannot be directly applied to direct absorption methods.

[0018] The direct absorption method calculates absorbance by directly measuring the absolute attenuation of transmitted light intensity, and then inversely determines concentration. It has advantages such as simple principle, no need for demodulation, low computational load, and avoidance of absorption peak misinterpretation, making it suitable for real-time online monitoring systems. However, the direct absorption method is extremely sensitive to light intensity fluctuations: light source attenuation or gas cell contamination can lead to a decrease in emitted light intensity, resulting in a lower calculated absorbance and a lower measured concentration; conversely, if the light source power is unexpectedly increased, the measured concentration will be higher. Currently, there is a lack of mature and effective solutions for correcting light intensity fluctuations in the direct absorption method, limiting its stability and reliability in long-term online monitoring.

[0019] In summary, existing TDLAS gas detection technology, when applied to online monitoring of dissolved gases in transformer oil, especially in the direct absorption mode, suffers from the technical problem of unstable measurement results due to light intensity fluctuations. There is an urgent need for a feature value extraction and correction method that can effectively suppress the influence of light intensity and improve measurement stability. Summary of the Invention

[0020] This invention provides a method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil, aiming to solve the problem that the existing direct absorption method suffers from decreased measurement accuracy due to light intensity fluctuations in online monitoring of transformer oil spectra and lacks effective correction methods.

[0021] To achieve the above objectives, the first aspect of the present invention provides a method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil, comprising the following steps: The acquired spectral absorption signal is low-pass filtered to obtain the filtered spectral absorption signal; The current spectral signal amplitude is calculated based on the amplitude of the spectral signal in the non-absorption segment of the filtered spectral absorption signal. Extract the upper envelope of the filtered spectral absorption signal and filter the upper envelope to obtain a smooth upper envelope; The absorption start point and absorption end point are determined on the smooth upper envelope; Connect the absorption start point and the absorption end point to form a baseline straight line; Within the interval between the absorption start point and the absorption end point, the difference between the smooth upper envelope and the baseline line is calculated, and the absolute value of the difference is summed to obtain the characteristic value. During the calibration phase, the amplitude of the calibration spectral absorption signal is obtained; During the actual measurement phase, the measured characteristic value is divided by the amplitude of the calibrated spectral absorption signal and then multiplied by the amplitude of the current spectral signal to obtain the characteristic value after light intensity correction.

[0022] Furthermore, the method for extracting the upper envelope of the filtered spectral absorption signal includes: In the filtered spectral absorption signal, find all local maxima. Draw the upper envelope based on the local maxima.

[0023] Furthermore, the method for determining the absorption start and end points on the smooth upper envelope includes: Find the maximum and minimum points on the smooth upper envelope; Starting from the minimum point, search for envelope poles sequentially to the left. When the current pole value is less than the value of its adjacent pole to the right, the current pole is determined as the absorption start point. Starting from the maximum value point, search for envelope poles sequentially to the right. When the current pole value is greater than the value of its adjacent pole to the left, the current pole is determined as the absorption endpoint.

[0024] Furthermore, the low-pass filter adopts a Butterworth filter with a passband frequency of 20000Hz, a stopband frequency of 150000Hz, a passband attenuation of 0.1dB, and a stopband attenuation of 60dB.

[0025] Furthermore, the method for calculating the amplitude of the current spectral signal is as follows: in the filtered spectral absorption signal, select the sampling point interval of the non-absorption segment, search for all maxima and minima in the interval, calculate the mean of the maxima and the mean of the minima respectively, and take the difference between the two as the amplitude of the current spectral signal.

[0026] Furthermore, the sampling point range of the unabsorbed segment is between 20,000 and 25,000 points.

[0027] Furthermore, when filtering the upper envelope, a Savitzky-Golay filter with an order of 5 and a filter window length of 51 is used.

[0028] Furthermore, the light intensity-corrected eigenvalues ​​are used to invert gas concentrations, wherein the gas includes one or more of acetylene, methane, ethylene, carbon monoxide, and carbon dioxide dissolved in transformer oil.

[0029] To achieve the above objectives, a second aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the transformer oil spectral feature value extraction and light intensity correction method, and the processor is configured to execute the program stored in the memory.

[0030] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the steps of the transformer oil spectral feature value extraction and light intensity correction method.

[0031] The beneficial effects of this invention are: Compared with existing technologies, the present invention provides a method for extracting spectral feature values ​​and correcting light intensity of transformer oil. This method removes noise by low-pass filtering the original spectral absorption signal; then, it extracts the amplitude of the current spectral signal from the non-absorption segment, which reflects the real-time light intensity; simultaneously, it extracts the upper envelope of the signal and filters it again to obtain a smooth upper envelope. Based on this, it accurately identifies the absorption start and end points, connects them to form a baseline, and calculates the cumulative absolute value of the difference between the smooth upper envelope and the baseline within the absorption interval, using this as a feature value related to gas concentration. During calibration, the amplitude of the calibration spectral signal is recorded. During actual measurement, the measured feature value is divided by the amplitude of the calibration spectral absorption signal and then multiplied by the amplitude of the current spectral signal to obtain the light intensity-corrected feature value. This correction process directly eliminates the influence of light intensity fluctuations on the feature value, thereby ensuring the accuracy of concentration inversion and significantly improving the stability and reliability of the direct absorption method in long-term online monitoring. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0033] Figure 1 This is a methane second harmonic interception signal diagram disclosed in an embodiment of the present invention.

[0034] Figure 2 This is a flowchart of a method for extracting spectral feature values ​​and correcting light intensity of transformer oil, as disclosed in an embodiment of the present invention.

[0035] Figure 3 This is an overall comparison diagram before and after filtering, as disclosed in an embodiment of the present invention.

[0036] Figure 4This is a partial comparison image before and after filtering disclosed in an embodiment of the present invention.

[0037] Figure 5 This is an upper envelope diagram disclosed in an embodiment of the present invention.

[0038] Figure 6 This is an overall diagram of an upper envelope filter disclosed in an embodiment of the present invention.

[0039] Figure 7 This is a partial view of an upper envelope filter disclosed in an embodiment of the present invention. Detailed Implementation

[0040] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0041] According to embodiments of the present invention, it should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the following methods, in some cases the steps shown or described may be executed in a different order than that shown here.

[0042] This invention is applied to an online monitoring system for transformer oil spectrometry. Its core lies in accurately measuring the concentration of trace gases dissolved in transformer oil, and then inferring the potential fault types of the transformer based on the types and amounts of these gases. Faults in transformer operation mainly include mechanical, electrical, and thermal faults, with electrical and thermal faults being the most prevalent. Over long-term operation, the insulating oil and solid insulating materials inside the transformer gradually age under the influence of multiple factors such as temperature, electric field, and oxidation, producing hydrogen (H2) and low-molecular-weight hydrocarbon gases, such as methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2). Long-term research on the generation mechanism of transformer insulating oil has revealed a correlation between the types and amounts of dissolved gases in transformer oil and the potential faults of the transformer, as shown in Table 1.

[0043] Table 1. Correspondence between dissolved gases in transformer oil and operating conditions

[0044] As shown in the table above, CO and CO2 are produced by the overheating decomposition of fixed insulation materials, while the production of C2H2 indicates a transformer discharge fault. Therefore, potential transformer faults can be identified by analyzing dissolved gases in the oil.

[0045] Specifically, the Roger ratio method used in this system uses the ratios of CH4 / H2, C2H2 / C2H4, and C2H4 / C2H6 as features for fault identification. The correspondence between faults and ratios is shown in Table 2 below.

[0046] Table 2 Roger's proportional boundary conditions

[0047] As shown in the table, when a certain ratio reaches the boundary condition, the fault can be inferred by referring to the table. The magnitude of gas concentration directly affects the accuracy of fault identification, and the magnitude of gas concentration depends on the correct identification of gas absorption peaks. Therefore, the transformer oil spectral feature value extraction and light intensity correction method proposed in this invention aims to obtain stable and reliable feature values ​​by accurately processing the spectral signal obtained by the direct absorption method, and then infer the accurate gas concentration, providing key data support for transformer fault diagnosis. It has important application value in transformer oil spectral online monitoring systems.

[0048] like Figure 2 As shown, this invention provides a method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil, which includes the following steps: Step S100: Perform low-pass filtering on the acquired spectral absorption signal to obtain the filtered spectral absorption signal; In the online monitoring system for transformer oil spectra, TDLAS technology scans the gas absorption spectrum using a tunable diode laser. It typically employs a sawtooth wave superimposed with a high-frequency sine wave as the driving method. The sawtooth wave frequency is 10Hz for slow wavelength scanning, while the high-frequency sine wave frequency is 10kHz for fast wavelength modulation. The sampling frequency is set to 1,000,000Hz to meet the high-resolution acquisition requirements.

[0049] Under such conditions of high-frequency modulation and high-speed sampling, the original spectral absorption signal will inevitably be mixed with high-frequency noise components from various sources such as circuit noise, environmental interference, and laser noise itself. If these noises are not processed, they will directly affect the accuracy and stability of subsequent feature value extraction.

[0050] Therefore, a low-pass filter was used to preprocess the original signal. Specifically, a Butterworth low-pass filter was selected, with the following design parameters: passband frequency 20000Hz, stopband frequency 150000Hz, passband attenuation 0.1dB, and stopband attenuation 60dB. This filter can effectively suppress high-frequency noise while preserving the useful signal components at and near the 10kHz modulation frequency to the maximum extent, thus ensuring the integrity of the absorption spectrum. The signal-to-noise ratio of the spectral absorption signal obtained after this filtering process is significantly improved. Figure 3 For the overall comparison before and after filtering, Figure 4 This is a local comparison before and after filtering.

[0051] The low-pass filtering mentioned here refers to a filtering method that allows signals below the set cutoff frequency to pass through while attenuating signals above the cutoff frequency; the Butterworth filter is a filter with the largest flat amplitude-frequency characteristic in the passband, and is widely used in the field of signal preprocessing due to its small phase distortion and simple design.

[0052] Step S200: Calculate the current spectral signal amplitude based on the amplitude of the spectral signal in the non-absorption segment of the filtered spectral absorption signal; In a TDLAS gas detection system, the laser continuously outputs light covering the gas absorption wavelength within one scan cycle. When the laser wavelength scans to the characteristic absorption position of the gas, the light intensity attenuates due to gas absorption, forming an absorption depression. At the beginning or end of the scan cycle, the laser wavelength is outside the gas absorption range, and the light intensity is unaffected by gas absorption; this portion of the signal is called the non-absorption segment. The intensity amplitude of the non-absorption segment directly reflects the output light intensity of the laser and the transmittance of the optical path, serving as a crucial indicator of the current light intensity state.

[0053] In this embodiment, the sampling point interval between 20,000 and 25,000 points in the filtered spectral absorption signal is selected as the non-absorption segment. The selection of this interval is based on the correspondence between the system sampling rate and the scanning period: the sampling frequency is 1,000,000 Hz, and the sawtooth scanning frequency is 10 Hz, that is, each scanning period contains 100,000 sampling points. The 20,000 to 25,000 points are located at the beginning of the scanning period, which is far away from the gas absorption position and can reliably reflect the light intensity reference that is not affected by gas absorption.

[0054] Within this interval, first search for all local maxima and local minima. Local maxima correspond to the upper peaks in the signal waveform, and local minima correspond to the lower valleys in the waveform. Then calculate the average of all maxima and the average of all minima respectively. Finally, subtract the average minima from the average maxima, and the difference is the current spectral signal amplitude.

[0055] By subtracting the mean of the maxima from the mean of the minima, the influence of residual high-frequency noise or modulation ripples in the signal on amplitude calculation can be effectively eliminated, obtaining a stable and representative light intensity amplitude. The sampling point mentioned here refers to each numerical point in the numerical sequence obtained after the analog-to-digital converter discretely samples the continuous analog signal; local maxima and local minima refer to the peak and valley values ​​of the signal within a certain neighborhood. Finding these extreme points can describe the amplitude of signal fluctuations. The current spectral signal amplitude calculated in step S200 will be used for light intensity correction in subsequent steps to eliminate the influence of light intensity fluctuations on the concentration measurement results.

[0056] Step S300: Extract the upper envelope of the filtered spectral absorption signal and filter the upper envelope to obtain a smooth upper envelope; In TDLAS gas absorption signals, useful information is mainly contained in the shape and amplitude of the absorption depression. However, due to the influence of high-frequency modulation and noise, the original signal often presents as a rippled oscillating waveform, and its true absorption profile is obscured by these details. In order to restore the intrinsic morphology of the absorption depression, it is necessary to extract the envelope of the signal.

[0057] In this embodiment, the method for extracting the upper envelope is as follows: First, in the filtered spectral absorption signal obtained in step S100, all local maxima are searched point by point; a local maxima is a point in the signal waveform where the value of a certain point is greater than the values ​​of its adjacent points, and these points correspond to the upper peak of the oscillating waveform; then, these local maxima are connected sequentially in chronological order to draw the upper envelope of the original signal. This upper envelope reflects the trend line of the signal amplitude changing with time, that is, the overall outline of the absorption peak, such as... Figure 5 As shown in the figure, although the original upper envelope outlines the approximate shape of the absorption peak, it still exhibits noticeable spikes and irregular fluctuations due to noise and modulation residue. If these spikes are not addressed, they will interfere with the accurate identification of the absorption start and end points, as well as the stable calculation of eigenvalues ​​in subsequent steps.

[0058] In this embodiment, a Savitzky-Golay filter is used to smooth the upper envelope. Specifically, the parameters are set to order 5 and filter window length 51. The Savitzky-Golay filter is a smoothing algorithm based on local polynomial fitting. Its characteristic is that it can maintain the shape and width of the signal well while smoothing noise, without over-broadening or flattening peaks like ordinary moving averages. Order 5 indicates that a fifth-order polynomial is used to fit the data within each window, and the window length 51 indicates that 51 points before and after the current point are used in the calculation for each fitting. The smoothed upper envelope obtained after this filtering process is as follows: Figure 6As shown in "Overall after upper envelope filtering", its overall contour remains consistent with the original envelope, but burrs have been effectively removed; further from... Figure 7 The magnified view of the "local area after upper envelope filtering" clearly shows that the originally jagged fluctuations have become smooth and continuous, and the rising edge, peak, and falling edge of the absorption peak have been well preserved.

[0059] The upper envelope mentioned here refers to the curve formed by connecting the local maxima of the signal, used to characterize the amplitude change trend of the signal; a local maximum is a point in the signal waveform whose value is greater than the value of its neighboring points; the Savitzky-Golay filter is a digital smoothing filter that achieves smoothing by performing polynomial least squares fitting on the data within a sliding window.

[0060] Step S400: Determine the absorption start point and absorption end point on the smooth upper envelope; In this embodiment, the determination of the absorption start point and absorption end point is based on the morphological characteristics of the smooth upper envelope, specifically using an extreme point search and comparison method. First, on the smooth upper envelope obtained in step S300, all data points are traversed to find the global maximum and global minimum points. Since gas absorption usually exhibits a concave waveform, the absorption area on the smooth upper envelope corresponds to the concave part of the envelope. Therefore, the minimum point is generally located near the center of absorption, i.e., the strongest absorption point, while the maximum point is located in the unabsorbed areas on both sides of the absorption, usually near the absorption edge.

[0061] After determining the maximum and minimum points, a bidirectional search is performed: starting from the minimum point, the search proceeds to the left, point by point, for the extreme points on the envelope. Here, the extreme point refers to the turning point of the envelope within a local range, i.e., the local maximum or local minimum point. When the current extreme point value is less than the value of its adjacent extreme point to the right, it means that the envelope has changed from a downward trend to an upward trend as it extends to the left. This turning point is the left boundary of absorption and is determined as the starting point of absorption.

[0062] Similarly, starting from the maximum point, we search for the extreme point point by point to the right. When the current extreme point value is greater than the value of its adjacent extreme point to the left, it means that the envelope changes from an upward trend to a downward trend as it extends to the right. This turning point is the right boundary of the absorption depression and is determined to be the absorption endpoint.

[0063] The principle behind this determination method is as follows: Ideally, when the envelope on the left side of the absorption depression enters the absorption region from the non-absorption region, its amplitude decreases from high to low, showing a downward trend. After reaching the absorption starting point, it begins to decrease even more rapidly, thus often exhibiting a curvature change at the starting point. Conversely, when the envelope on the right side of the absorption depression exits the absorption region and returns to the non-absorption region, its amplitude increases from low to high, showing an upward trend. After reaching the absorption endpoint, it tends to stabilize, thus also exhibiting a shape inflection point at the endpoint. By searching for the changing trends of extreme values, these inflection points can be effectively captured, thereby accurately locating the absorption boundary and avoiding misjudgments caused by noise or envelope fluctuations. Here, extreme points refer to local extreme points on the function curve, including local maxima and local minima.

[0064] Step S500: Connect the absorption start point and the absorption end point to form a baseline straight line; The baseline represents the linear trend of the light intensity signal under ideal conditions where there is no gas absorption or the absorption is negligible. In actual measurements, due to factors such as fluctuations in laser output power, optical path transmission loss, and non-resonant absorption of background gas, the spectral signal often has a slowly changing background component. By connecting the absorption start point and the end point to construct the baseline, this background influence can be effectively deducted, thereby separating the absorption from the background signal.

[0065] Step S600: Within the interval between the absorption start point and the absorption end point, calculate the difference between the smooth upper envelope and the baseline straight line, and sum the absolute values ​​of the differences to obtain the characteristic value; In this step, all sampling points between the absorption start point and the absorption end point are first determined. For each point in this interval, the difference between the value of the smooth upper envelope at that point and the value of the baseline line at the same horizontal coordinate position is calculated. Since gas absorption is characterized by a downward concave envelope, the envelope value is usually less than the baseline value, and the difference itself is negative. However, in order to uniformly measure the absolute magnitude of absorption, the absolute value of each difference is taken, thereby converting the absorption intensity into a positive number.

[0066] Subsequently, all absolute values ​​are summed, and the sum is the characteristic value of the absorption. This summation process is actually a numerical integration of the area enclosed by the absorption, reflecting the total energy or total number of molecules absorbed by the gas. Its physical meaning is inherently consistent with the absorbance integral in Lambert-Beer's law, that is, the magnitude of the characteristic value is proportional to the gas concentration.

[0067] Compared to traditional methods that only use peak and valley values ​​to calculate concentration, this interval-based summation method utilizes all waveform information of the absorption, resulting in stronger noise resistance and stability. Even if the shape of the absorption depression is slightly distorted, the summation can still stably reflect the concentration change.

[0068] The summation of the absolute values ​​of the differences described here is an integral operation that quantifies the deviation between the waveform and the baseline, equivalent to calculating the area enclosed by the absorption. The characteristic values ​​obtained in step S600 will be combined with the light intensity correction factor in subsequent steps for the final gas concentration inversion.

[0069] Step S700: In the calibration stage, obtain the amplitude of the calibration spectral absorption signal; Before the online monitoring system for transformer oil spectra is put into actual operation, calibration is required in a known and stable environment to establish a reference benchmark for subsequent measured data. Specifically, in a laboratory or standard operating condition, background gas or a standard gas of known concentration without the gas to be measured is introduced into the White cell. Using the same acquisition parameters and optical path settings as in the actual measurement, one or more sets of spectral absorption signals are acquired. Subsequently, following the same method as in step S200—that is, searching for maxima and minima in the non-absorption segment of the filtered signal (e.g., within the range of 20,000 to 25,000 points), calculating the difference between the mean of the maxima and the mean of the minima—the amplitude of the spectral signal under this condition is obtained, and the average of multiple measurements is taken as the calibration spectral absorption signal amplitude. This calibration amplitude represents the light intensity level of the system under ideal or reference conditions, and is used as the denominator in step S800 to compare with the measured amplitude to calculate the correction coefficient caused by light intensity fluctuations.

[0070] Step S800: In the actual measurement stage, divide the measured characteristic value by the amplitude of the calibrated spectral absorption signal, and then multiply it by the amplitude of the current spectral signal to obtain the characteristic value after light intensity correction.

[0071] In the aforementioned steps, a characteristic value directly related to the gas concentration was obtained through step S600. This characteristic value is obtained by accumulating the difference between the smooth upper envelope and the baseline, reflecting the area enclosed by the absorption depression. However, the magnitude of this characteristic value depends not only on the gas concentration but also on the light intensity fluctuation. When the laser output power decreases or the gas chamber becomes contaminated, causing the light intensity to drop, the absorption amplitude will decrease even if the gas concentration remains unchanged, resulting in a lower characteristic value. Conversely, if the light intensity unexpectedly increases, the characteristic value will be higher.

[0072] To eliminate this effect, step S800 introduces a light intensity correction mechanism: First, the current spectral signal amplitude calculated in step S200 is used, which reflects the light intensity level of the non-absorption segment during this measurement in real time; second, the calibration spectral absorption signal amplitude obtained in step S700 during the calibration stage is called, which represents the light intensity level under the reference state; then, the ratio of the current spectral signal amplitude to the calibration spectral absorption signal amplitude is calculated, which is the correction coefficient caused by light intensity fluctuation; finally, the characteristic value calculated in step S600 is multiplied by the correction coefficient to obtain the characteristic value after light intensity correction.

[0073] The characteristic values ​​measured under the current light intensity conditions are converted to equivalent values ​​under the calibrated light intensity conditions, thereby eliminating the influence of light intensity fluctuations on the characteristic values ​​and making the corrected characteristic values ​​only related to the gas concentration. In practical applications, substituting the corrected characteristic values ​​into the characteristic value-concentration relationship curve established during the calibration stage allows for the accurate deduction of the concentration of the gas to be measured. The light intensity-corrected characteristic values ​​described here are essentially the result of normalizing the original characteristic values, and their numerical stability is significantly better than that of the uncorrected characteristic values, effectively solving the technical problem of decreased measurement accuracy due to light intensity fluctuations in long-term online monitoring using the direct absorption method. Through the correction process in step S800, this method achieves adaptive compensation for light intensity fluctuations, ensuring the accuracy and reliability of dissolved gas concentration measurement in transformer oil.

[0074] Compared with existing technologies, the present invention provides a method for extracting spectral feature values ​​and correcting light intensity of transformer oil, which, in response to the application requirements of direct absorption method in TDLAS gas monitoring, constructs a complete end-to-end feature value calculation and correction system. Specifically: by extracting and smoothing the upper envelope of the filtered spectral absorption signal (step S300), the interference of high-frequency modulation and noise on the absorption profile is effectively removed, restoring the intrinsic morphology of the absorption peak; subsequently, based on the extreme point trend analysis of the smoothed upper envelope, the absorption start and end points are intelligently determined (step S400), avoiding the uncertainty of traditional methods that rely on fixed thresholds or manual intervention; then, a dynamic baseline is constructed by connecting the start and end points (step S500), adaptively subtracting background signals; finally, within the precisely defined absorption range, the sum of the absolute values ​​of the difference between the smoothed upper envelope and the baseline is calculated as the feature value (step S600). This feature value is essentially a numerical integral of the absorption depression area, which, compared to the traditional direct absorption method that only relies on simple amplitude calculation, has stronger noise resistance and concentration response linearity.

[0075] These four steps are interconnected and progressive, together forming a new feature extraction method specifically for the direct absorption method. It retains the inherent advantages of the direct absorption method, such as no demodulation required and low computational cost, while significantly improving the stability and accuracy of the feature values ​​through refined waveform processing. At the same time, it provides a reliable input basis for subsequent light intensity correction, thus fundamentally solving the problem of decreased measurement accuracy caused by light intensity fluctuations and noise interference in long-term online monitoring using the existing direct absorption method.

[0076] According to another aspect of the embodiments of this application, an electronic device is also provided, including a processor and a memory, wherein the processor is configured to implement the steps of the method when executing a computer program stored in the memory.

[0077] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0079] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0080] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0081] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil, characterized in that, Includes the following steps: The acquired spectral absorption signal is low-pass filtered to obtain the filtered spectral absorption signal; The current spectral signal amplitude is calculated based on the amplitude of the spectral signal in the non-absorption segment of the filtered spectral absorption signal. Extract the upper envelope of the filtered spectral absorption signal and filter the upper envelope to obtain a smooth upper envelope; The absorption start point and absorption end point are determined on the smooth upper envelope; Connect the absorption start point and the absorption end point to form a baseline straight line; Within the interval between the absorption start point and the absorption end point, the difference between the smooth upper envelope and the baseline line is calculated, and the absolute value of the difference is summed to obtain the characteristic value. During the calibration phase, the amplitude of the calibration spectral absorption signal is obtained; During the actual measurement phase, the measured characteristic value is divided by the amplitude of the calibrated spectral absorption signal and then multiplied by the amplitude of the current spectral signal to obtain the characteristic value after light intensity correction.

2. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, The method for extracting the upper envelope of the filtered spectral absorption signal includes: In the filtered spectral absorption signal, find all local maxima. Draw the upper envelope based on the local maxima.

3. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, The method for determining the absorption start and end points on the smooth upper envelope includes: Find the maximum and minimum points on the smooth upper envelope; Starting from the minimum point, search for envelope poles sequentially to the left. When the current pole value is less than the value of its adjacent pole to the right, the current pole is determined as the absorption start point. Starting from the maximum value point, search for envelope poles sequentially to the right. When the current pole value is greater than the value of its adjacent pole to the left, the current pole is determined as the absorption endpoint.

4. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, The low-pass filter uses a Butterworth filter with a passband frequency of 20,000 Hz, a stopband frequency of 150,000 Hz, a passband attenuation of 0.1 dB, and a stopband attenuation of 60 dB.

5. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, The method for calculating the amplitude of the current spectral signal is as follows: In the filtered spectral absorption signal, select the sampling point interval of the non-absorption segment, search for all maxima and minima in the interval, calculate the mean of the maxima and the mean of the minima respectively, and take the difference between the two as the amplitude of the current spectral signal.

6. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 5, characterized in that, The sampling point range for the non-absorbed segment is between 20,000 and 25,000 points.

7. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, When filtering the upper envelope, a Savitzky-Golay filter with an order of 5 and a filter window length of 51 is used.

8. The method for extracting spectral characteristic values ​​and correcting light intensity of transformer oil as described in claim 1, characterized in that, The light intensity-corrected eigenvalues ​​are used to invert gas concentrations, and the gases include one or more of the following dissolved in transformer oil: acetylene, methane, ethylene, carbon monoxide, and carbon dioxide.

9. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the transformer oil spectral feature value extraction and light intensity correction method according to any one of claims 1-8, and the processor is configured to execute the program stored in the memory.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is run by the processor, it executes the steps of the transformer oil spectral feature value extraction and light intensity correction method according to any one of claims 1-8.

Images