Transformer bushing frequency domain dielectric spectrum measurement method and device and computer program product

Abstract

The invention discloses a frequency domain dielectric spectrum measurement method and device for a transformer bushing and a computer program product, and the method comprises the steps: S1, carrying out the sampling and preprocessing of a low-frequency signal of the transformer bushing; s2, performing intelligent data processing on the preprocessed signal through signal prediction, data reconstruction and sliding window regression; s3, filling sparse or discontinuous measurement data under low frequency by adopting an interpolation method, and processing the filled data through a low-pass filter and a moving average filter; s4, shortening the measurement time by using a time extension technology, and enhancing the low-frequency component of the signal; and S5, extracting the dielectric constant and the dielectric loss of the transformer bushing, comparing the dielectric constant and the dielectric loss with actual measurement results, and performing error analysis. According to the invention, the speed and accuracy of frequency domain dielectric spectrum measurement of the transformer bushing under low frequency are obviously improved, the signal acquisition and processing process is optimized, and high-efficiency and high-fidelity measurement under the low-frequency condition is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, a device and a computer program product for measuring a transformer bushing frequency domain dielectric spectrum. Background Art

[0002] In the operation of power systems, the stable operation of transformers is crucial, and the insulation performance of transformer bushings directly affects transformer safety. As the operating time increases, the insulation performance of bushings will deteriorate due to various factors, bringing hidden dangers to the power system.

[0003] Frequency domain dielectric spectroscopy (FDS) technology is currently the main method for evaluating the insulation status of transformer bushings. It reflects the microstructural changes of the insulating material by applying AC voltages of different frequencies to the bushing and measuring the dielectric response parameters, providing a basis for insulation status evaluation.

[0004] However, FDS technology faces many challenges when applied in low frequency bands (especially 1MHz and below). In terms of data acquisition, due to the slow polarization response of insulating materials at low frequencies, traditional methods require long-term application of test signals and multi-point sampling, resulting in time-consuming data acquisition and failure to meet the needs of rapid measurement, which seriously affects diagnostic efficiency, especially in the online monitoring scenario of power equipment.

[0005] In terms of data processing, existing technologies mostly rely on complex mathematical models, which have a large amount of calculations and are affected by factors such as measurement noise, system errors and non-uniformity of insulation materials. The calculation accuracy is difficult to guarantee, resulting in deviations and uncertainties in data processing results, reducing the accuracy and reliability of insulation status assessment.

[0006] In addition, the non-ideality of the measuring equipment and the attenuation of low-frequency signals seriously affect the measurement accuracy when measuring objects with large capacitance and complex structures such as transformer bushings, while traditional signal processing methods are difficult to effectively compensate for signal attenuation and extract weak signals.

[0007] At the same time, with the development of data collection methods, the amount of FDS measurement data has increased dramatically. Traditional methods have not fully utilized advanced signal processing and data compression technologies, and the data processing efficiency is low, which cannot meet the power system's needs for real-time perception and intelligent diagnosis of equipment status. Summary of the invention

[0008] The technical problem to be solved by the present invention is to provide a method, device and computer program product for measuring the frequency domain dielectric spectrum of a transformer bushing, so as to realize fast and accurate measurement of the dielectric spectrum data of the transformer bushing under low frequency conditions, and to efficiently process and analyze the data to accurately evaluate the insulation status.

[0009] In order to solve the above technical problems, the present invention provides a method for measuring the frequency domain dielectric spectrum of a transformer bushing, comprising:

[0010] Step S1, sampling and preprocessing the transformer bushing low-frequency signal;

[0011] Step S2, performing intelligent data processing on the preprocessed signal through signal prediction, data reconstruction and sliding window regression;

[0012] Step S3, using an interpolation method to fill in the sparse or discontinuous measurement data at low frequency, and processing the filled data through a low-pass filter and a moving average filter;

[0013] Step S4, using time stretching technology to shorten the measurement time and enhance the low-frequency components of the signal;

[0014] Step S5, extracting the dielectric constant and dielectric loss of the transformer bushing, comparing them with the actual measurement results and performing error analysis.

[0015] Preferably, the preprocessing in step S1 specifically includes using sliding window technology, denoising, detrending and frequency domain analysis;

[0016] The sliding window technique uses a rectangular window, sets the window length and a predetermined step size, and performs weighted averaging on the data within the window, with each data point having a specified weight;

[0017] The denoising process adopts a combination of Kalman filtering and adaptive filtering, wherein the adaptive filtering uses an RLS filter to adjust the parameters of the Kalman filter;

[0018] Detrending was done using polynomial detrending, where the data were fitted with a polynomial and the fitted trend portion was subtracted from the original data;

[0019] Frequency domain analysis extracts and analyzes useful frequency components in low-frequency signals through Fourier transform, and processes frequency domain data using low-pass, high-pass or band-pass filters.

[0020] Preferably, in the sliding window technique, a window length T is set ω , the window starts from the beginning of the data sequence and slides forward step by step, each time moving a predetermined step size and processing a subset of the data in the window;

[0021] For time series, the signal after sliding window is expressed as:

[0022] x ω (t) = x(t)·ω(t)

[0023] where ω(t) is the window function, and

[0024]

[0025] Perform a weighted average of the data points in the window, where each data point has a specified weight:

[0026]

[0027] Where N is the window size, x(ti) is the i-th data point in the window, and ω i is the weight coefficient.

[0028] Preferably, the signal prediction in step S2 adopts a combination of an autoregressive model and a Kalman filter, and the autoregressive model is used as a state transition model of the Kalman filter; the data reconstruction adopts compressed sensing technology, and the sparsity of the signal is used to reconstruct the signal through a small amount of measurements.

[0029] Preferably, the data reconstruction specifically includes:

[0030] Using compressed sensing technology, the sparsity of the signal is exploited to reconstruct the signal through a small number of measurements:

[0031]

[0032] in, is the estimated value of the reconstructed signal, arg min||·|| indicates the use of sparse representation, Φ is the sparse dictionary matrix, y is the actual measurement value vector, and λ is the regularization parameter.

[0033] Preferably, linear interpolation is used to complete the data in step S3:

[0034]

[0035] Among them, x(t) represents the data value completed at time t after linear interpolation, t0 and t1 are the times of adjacent known data points, and Δt is the time interval.

[0036] Preferably, the time stretching technology in step S4 extends the original signal to a longer time series to improve the frequency resolution by splitting and parallel processing the long-time signal, and enhances the low-frequency component of the signal by an enhancement factor.

[0037] Preferably, the original signal is extended to a longer time series to improve the frequency resolution by:

[0038] x extended (t)=Interp(x(t),new time points)

[0039] The way to enhance the low-frequency components of the signal to improve the resolution of the analysis is:

[0040] X enhanced(f) = X(f)·E(f)

[0041] Among them, X(f) is the frequency domain representation of the original signal, E(f) is the enhancement factor, and the value of E(f) is greater than 1 in the low frequency band and approaches 0 in the high frequency band.

[0042] The present invention also provides a transformer bushing frequency domain dielectric spectrum measuring device, comprising:

[0043] one or more processors;

[0044] Memory;

[0045] One or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more application programs are configured to execute the transformer bushing frequency domain dielectric spectrum measurement method.

[0046] The present invention also provides a computer program product, comprising computer instructions, wherein the computer instructions instruct a computer device to execute operations corresponding to the method.

[0047] The implementation of the present invention has the following beneficial effects: on the one hand, by comprehensively using cutting-edge technologies such as sliding windows, sparse sampling and compressed sensing, the measurement speed and accuracy are greatly improved, the low-frequency signal acquisition and processing process is optimized, and the data processing efficiency is significantly improved. For example, in the signal sampling stage, low-frequency signals can be accurately captured to reduce information omissions. On the other hand, relying on the intelligent data processing link, with the help of compressed sensing and interpolation means, the data is flexibly adjusted according to the actual measurement conditions, the measurement error is effectively controlled within a reasonable range, and the measurement accuracy is guaranteed. Furthermore, with the help of layered signal processing technology, the measurement time is shortened by time extension, and the frequency domain analysis and interpolation are used to achieve efficient signal processing and accurate reconstruction, ensuring that the measurement system operates efficiently and with high fidelity at low frequencies. The present invention not only solves the problems of long measurement time and complex data processing of traditional technologies, but also lays a solid foundation for the early diagnosis and maintenance of transformer bushings, effectively promotes the stable and reliable operation of the power system, and reduces the potential risks caused by bushing failures. It has important practical value and innovative significance in the field of power equipment detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1It is a flow chart of a method for measuring the frequency domain dielectric spectrum of a transformer bushing according to an embodiment of the present invention.

[0050] Figure 2 The figure is a schematic diagram of the principle of a method for measuring the frequency domain dielectric spectrum of a transformer bushing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following descriptions of the embodiments refer to the accompanying drawings to illustrate specific embodiments in which the present invention may be implemented.

[0052] Please refer to Figure 1 As shown, an embodiment of the present invention provides a method for measuring a transformer bushing frequency domain dielectric spectrum, comprising:

[0053] Step S1, sampling and preprocessing the transformer bushing low-frequency signal;

[0054] Step S2, performing intelligent data processing on the preprocessed signal through signal prediction, data reconstruction and sliding window regression;

[0055] Step S3, using an interpolation method to fill in the sparse or discontinuous measurement data at low frequency, and processing the filled data through a low-pass filter and a moving average filter;

[0056] Step S4, using time stretching technology to shorten the measurement time and enhance the low-frequency components of the signal;

[0057] Step S5, extracting the dielectric constant and dielectric loss of the transformer bushing, comparing them with the actual measurement results and performing error analysis.

[0058] It can be seen from the above steps that the embodiment of the present invention effectively removes noise and trend interference in the signal through sliding window technology, denoising, detrending and frequency domain analysis, accurately extracts low-frequency useful information, and improves signal quality. Secondly, signal prediction, data reconstruction and sliding window regression in intelligent data processing enhance the generalization and prediction capabilities of the model, making data analysis more reliable. Furthermore, the use of interpolation and filters to process sparse data, combined with time extension technology to shorten the measurement time and enhance the low-frequency component, not only improves the integrity and stability of the data, but also optimizes the overall measurement efficiency, and ultimately achieves accurate extraction and error analysis of the dielectric constant and dielectric loss of the transformer bushing, providing an efficient and accurate solution for low-frequency measurement.

[0059] Specifically, please combine Figure 2 As shown, in the embodiment of the present invention, step S1 first samples and pre-processes the low-frequency signal, which specifically includes the following aspects:

[0060] 1) Low frequency signal sampling device

[0061] Choose a high-precision low-frequency sampling instrument to ensure that the 1mHz signal can be captured. According to the sampling theorem (Nyquist sampling theorem), the sampling frequency f s Must be greater than twice the signal frequency f; but to improve accuracy, the sampling frequency must be at least 10 times the signal frequency:

[0062] f s >10f (1)

[0063] Assume the signal is x(t), at t i After sampling at the moment, the sampling point is:

[0064] x(t i )=x(iΔt) (2)

[0065] Where Δt = 1 / f s is the sampling interval.

[0066] 2) Sliding Window Technology

[0067] Set a window length T ω , the window starts from the beginning of the data sequence and slides forward step by step, each time moving a predetermined step size and processing a subset of the data in the window, and finally summarizing the processing results of the window and keeping the window sliding. For time series, the signal after sliding the window can be expressed as:

[0068] x ω (t) = x(t)·ω(t) (3)

[0069] Among them, ω(t) is the window function, and this method uses a rectangular window.

[0070]

[0071] To emphasize the importance of certain measured data, a weighted moving average window is used for signal processing. The data points within the window are weighted averaged, where each data point has a specified weight.

[0072]

[0073] Where N is the window size, x(ti) is the i-th data point in the window, and ω i is the weight coefficient.

[0074] 3) Denoising

[0075] The Kalman filter is combined with the adaptive filter to achieve better denoising effect. The Kalman filter provides system state estimation and noise prediction, while the adaptive filter can dynamically adjust the parameters of the Kalman filter (such as the covariance of process noise and measurement noise) to adapt to different noise environments.

[0076] The Kalman filter assumes that the state of the system can be described by a linear system model and that the observed data is affected by Gaussian noise. The Kalman filter is based on two equations:

[0077] x k+1 =F k +B k u k +ω k (7)

[0078] z k =H k x k +v k (8)

[0079] Among them, F k is the state transfer matrix, B k is the control input matrix, H k is the observation matrix, ω k and v k They are process noise and observation noise respectively. In each iteration, state prediction and update are performed to finally obtain the optimal state estimate.

[0080] Adaptive filtering is a technique that can dynamically adjust filter coefficients based on the statistical characteristics of the input signal. It usually includes a filter with adjustable coefficients and an adaptive algorithm to update these coefficients.

[0081] The embodiment of the present invention adopts an RLS filter to adjust the parameters of the Kalman filter.

[0082]

[0083] Among them, ω n is the filter coefficient, μ is the step size factor, d n It is a signal of expectation. is the actual signal.

[0084] 4) Detrending

[0085] Use polynomial detrending to remove long-term trends from data in order to analyze short-term fluctuations or other characteristics. Fit a polynomial to the data and then subtract the trend from the original data.

[0086] x(t)=p(t)+Residual (10)

[0087] Among them, p(t) is the trend term of the polynomial fitting, which makes the short-term characteristics of the data more obvious after processing.

[0088] 5) Frequency domain analysis

[0089] Through Fourier transform, useful frequency components in low-frequency signals are extracted and analyzed.

[0090]

[0091] In the frequency domain, noise usually appears as high-frequency or random components, while effective signals appear as low-frequency components. By filtering in the frequency domain, noise interference can be eliminated and useful low-frequency information can be retained. Apply low-pass, high-pass or band-pass filters to process frequency domain data.

[0092] X filtered (f) = X(f)·H(f) (12)

[0093] Among them, X filtered (f) represents the frequency domain representation of the signal after the filter processing, X(f) represents the frequency domain representation of the original signal, and H(f) is the filter response function. The filtered signal X filtered (f) Only the frequency components that meet the filter characteristics (low-pass, high-pass or band-pass) are retained. For example, when a low-pass filter is used, X filtered (f) Only low-frequency components below the filter cutoff frequency will be included, and high-frequency noise components will be effectively suppressed.

[0094] So far, the embodiment of the present invention has completed all preprocessing operations on the measurement data, and fully reserved the low-frequency effective information.

[0095] Step S2 of the embodiment of the present invention performs intelligent data processing through signal prediction, data reconstruction and sliding window regression.

[0096] 1) Signal prediction

[0097] The linear trend modeling capability of the autoregressive model is combined with the dynamic estimation and noise processing capabilities of the Kalman filter. The autoregressive model is used as the state transition model of the Kalman filter. The autoregressive model is

[0098] X t =φ1X t-1 +φ2X t-2 +∈ t (13)

[0099] Among them, X t represents the signal value at time t, which is a point in the time series signal to be predicted or analyzed; φ1 and φ2 are the coefficients of the autoregressive model, which determine the past signal value (X t-1 and X t-2) contributes to the current signal value. By adjusting these coefficients, the linear trend of the signal can be fitted. For example, if φ1 is large, it means that the signal value at the previous moment has a greater impact on the signal value at the current moment; if φ2 is also large, then the signal value at the previous moment also affects the current signal value to a certain extent. ∈ t It is a random error term, which represents the part that the model cannot fully explain through past signal values. This error term is usually assumed to be a random variable with a mean of zero and a constant variance. It is used to consider the uncertainty factors in the signal, such as noise and small fluctuations not captured by the model.

[0100] Use it to predict the state at the next moment, update the state estimate through Kalman filtering, and correct the prediction based on the observed data.

[0101] x t =AX t-1 +Hu t +w t (14)

[0102] Among them, x t is the state vector, A is the state transfer matrix, H is the observation matrix, u t represents the control input vector, w t represents process noise. In the framework of Kalman filtering, if the system is affected by external control input, u t It is used to describe the changes in the system state caused by these external inputs.

[0103] 2) Data reconstruction

[0104] Data reconstruction is used to improve the completeness of the predicted value and the accuracy of the prediction model. The embodiment of the present invention adopts compressed sensing technology and utilizes the sparsity of the signal to reconstruct the signal through a small amount of measurement.

[0105]

[0106] in, is the estimated value of the reconstructed signal, arg min||·|| indicates the use of sparse representation, Φ is the sparse dictionary matrix, y is the actual measurement value vector, and λ is the regularization parameter.

[0107] In the data reconstruction process, the signal estimate is calculated by using the known measurement information (represented by y) and related parameters (such as Φ and λ) through compressed sensing technology. It is an approximate reconstruction of the original signal, and its purpose is to improve the integrity and accuracy of signal prediction. In addition, in the compressed sensing scenario, since the signal is sparse, information about the signal can be obtained through a small number of measurements, and these measurements form the vector y. For example, when processing the low-frequency signal of the transformer bushing, these measurements are obtained after some sampling or transformation, and they are used as known information for signal reconstruction.

[0108] 3) Sliding Window Regression

[0109] Sliding window regression is a method commonly used in time series analysis. It trains a regression model by sliding a window of fixed length on the time series. The embodiment of the present invention adopts local weighted regression technology.

[0110]

[0111] Among them, y(t) represents the predicted output of the regression model at time, which is the result of local weighted regression calculation based on the data in the sliding window, and is used to estimate or predict the value of the time series at that moment. t is the local weighting coefficient. In local weighted regression, different weights are assigned to different data points in the window. These weights reflect the relative importance of each data point to the current predicted output y(t). The calculation of weights is usually related to factors such as the distance between the data point and the current predicted point, and will change dynamically as the window slides to achieve the effect of local weighting. β0 is the intercept term of the regression model, which indicates the baseline value of the predicted output when the independent variable x(ti) is zero; β1 is the slope coefficient of the regression model, which indicates the change in the predicted output y(t) for each unit change of the independent variable x(ti). β0 and β1 are parameters obtained by regression fitting the data in the sliding window, which can be used to describe the linear relationship between the independent variable and the dependent variable.

[0112] The combination of technologies can improve the generalization ability of the model, enhance the accuracy of prediction, and make data analysis results more reliable.

[0113] Since the measured data at low frequencies may be sparse or discontinuous, an interpolation method (such as linear interpolation, spline interpolation or more complex interpolation techniques) is required to fill in the data points in the frequency domain. In step S3 of the embodiment of the present invention, linear interpolation is used to complete the data:

[0114]

[0115] Among them, x(t) represents the data value completed at time t after linear interpolation, t0 and t1 are the times of adjacent known data points, and Δt is the time interval.

[0116] Specifically, t0 and t1 are the times of known data points, and Δt is the time interval (t1 - t0). x(t0) and x(t1) are the known data values at times t0 and t1. Through formula (17), using the information of known data points, the data at time t (t0 < t < t1) is linearly interpolated and completed to handle the possible data sparsity or discontinuity at low frequencies.

[0117] The high-frequency noise in the completed data is removed using a low-pass filter, and then the data is smoothed using a moving average filter, which can make the dielectric spectrogram smoother and have less noise.

[0118]

[0119] Among them, f c is the cut-off frequency, and n is the order of the filter.

[0120] In step S4, to achieve fast measurement at low frequencies, time stretching technology is used to shorten the measurement time. The long-time signal is split and processed in parallel, and the original signal is extended to a longer time series to improve the frequency resolution.

[0121] x extended (t) = Interp(x(t), new time points) (19)

[0122] The low-frequency components of the signal are enhanced to improve the analysis resolution.

[0123] X enhanced (f) = X(f) · E(f) (20)

[0124] Among them, E(f) is the enhancement factor, and the value of E(f) is greater than 1 in the low-frequency band and approaches 0 in the high-frequency band.

[0125] For a long-time signal, it is split into multiple parts. The advantage of this splitting is that these parts can be processed simultaneously, just like multiple people doing different parts of the work at the same time, thereby improving the overall processing efficiency and achieving the purpose of shortening the measurement time. For example, assume there is a signal with a duration of 10 minutes, which is split into 5 sub-signals with a duration of 2 minutes each, and then multiple processing units (such as multiple processor cores) are used to process these 5 sub-signals simultaneously. Compared with processing the entire 10-minute signal sequentially, the total processing time can be greatly reduced.

[0126] After splitting the signal, the original signal is expanded to a longer time series. The frequency resolution is related to the duration of the signal. The longer the duration, the higher the frequency resolution. For example, by adding new time points to the time axis of the signal (such as the interpolation operation reflected by Interp(x(t), new time points) in formula (19)), the signal can be made more refined in the time dimension, and then the different frequency components can be more accurately distinguished in the frequency domain. This is especially important for the analysis of low-frequency signals, because the frequency components of low-frequency signals are relatively close and require higher resolution to distinguish.

[0127] For low-frequency signal analysis, more attention is paid to the low-frequency component. By multiplying the enhancement factor E(f), the amplitude of the low-frequency component can be increased. For example, if the value of E(f) in the low-frequency band is greater than 1, the amplitude of the low-frequency component will increase, which is equivalent to amplifying the low-frequency part, so that the characteristics of the low-frequency signal can be observed and analyzed more clearly, and the resolution of the analysis can be improved. This helps to more accurately extract key parameters such as the dielectric constant and dielectric loss of the transformer bushing in subsequent steps.

[0128] Finally, in step S5, the dielectric constant and dielectric loss of the transformer bushing are extracted, compared with the actual measurement results, and error analysis is performed to optimize the fast measurement method at low frequency.

[0129] Corresponding to the transformer bushing frequency domain dielectric spectrum measurement method described in the first embodiment of the present invention, the second embodiment of the present invention further provides a transformer bushing frequency domain dielectric spectrum measurement device, including:

[0130] one or more processors;

[0131] Memory;

[0132] One or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more application programs are configured to execute the transformer bushing frequency domain dielectric spectrum measurement method.

[0133] Corresponding to the transformer bushing frequency domain dielectric spectrum measurement method described in the aforementioned embodiment 1 of the present invention, embodiment 3 of the present invention further provides a computer program product, including computer instructions, and the computer instructions instruct a computer device to perform operations corresponding to the transformer bushing frequency domain dielectric spectrum measurement method described in the aforementioned embodiment 1 of the present invention.

[0134] Preferably, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor. The processor is the control center of the device, and various parts of the device are connected using various interfaces and lines.

[0135] The memory mainly includes a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (Flash Card), etc., or the memory can also be other volatile solid-state storage devices.

[0136] It should be noted that the above-mentioned device may include but is not limited to a processor and a memory, which can be understood by those skilled in the art.

[0137] For the working principle and process of this embodiment, please refer to the description of the aforementioned embodiment 1 of the present invention, which will not be repeated here.

[0138] Compared with the prior art, the beneficial effects brought by the embodiments of the present invention are that, on the one hand, by comprehensively using cutting-edge technologies such as sliding windows, sparse sampling and compressed sensing, the measurement speed and accuracy are greatly improved, the low-frequency signal acquisition and processing process is optimized, and the data processing efficiency is significantly improved. For example, in the signal sampling stage, low-frequency signals can be accurately captured to reduce information omissions. On the other hand, relying on the intelligent data processing link, with the help of compressed sensing and interpolation means, the data is flexibly adjusted according to the actual measurement conditions, the measurement error is effectively controlled within a reasonable range, and the measurement accuracy is guaranteed. Furthermore, with the help of layered signal processing technology, the measurement time is shortened by time extension, and the frequency domain analysis and interpolation are used to achieve efficient signal processing and accurate reconstruction, ensuring that the measurement system operates efficiently and with high fidelity at low frequencies. The present invention not only solves the problems of long measurement time and complex data processing of traditional technologies, but also lays a solid foundation for the early diagnosis and maintenance of transformer bushings, effectively promotes the stable and reliable operation of the power system, and reduces the potential risks caused by bushing failures. It has important practical value and innovative significance in the field of power equipment detection.

[0139] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for measuring the frequency domain dielectric spectrum of a transformer bushing, characterized in that: include: Step S1, sampling and preprocessing the transformer bushing low-frequency signal; Step S2, performing intelligent data processing on the preprocessed signal through signal prediction, data reconstruction and sliding window regression; Step S3, using an interpolation method to fill in the sparse or discontinuous measurement data at low frequency, and processing the filled data through a low-pass filter and a moving average filter; Step S4, using time stretching technology to shorten the measurement time and enhance the low-frequency components of the signal; Step S5, extracting the dielectric constant and dielectric loss of the transformer bushing, comparing them with the actual measurement results and performing error analysis.

2. The method according to claim 1, characterized in that: The preprocessing in step S1 specifically includes using sliding window technology, denoising, detrending and frequency domain analysis; The sliding window technique uses a rectangular window, sets the window length and a predetermined step size, and performs weighted averaging on the data within the window, with each data point having a specified weight; The denoising process adopts a combination of Kalman filtering and adaptive filtering, wherein the adaptive filtering uses an RLS filter to adjust the parameters of the Kalman filter; Detrending was done using polynomial detrending, where the data were fitted with a polynomial and the fitted trend portion was subtracted from the original data; Frequency domain analysis extracts and analyzes useful frequency components in low-frequency signals through Fourier transform, and processes frequency domain data using low-pass, high-pass or band-pass filters.

3. The method according to claim 2, characterized in that In the sliding window technique, a window length T is set. ω , the window starts from the beginning of the data sequence and slides forward step by step, each time moving a predetermined step size and processing a subset of the data in the window; For time series, the signal after sliding window is expressed as: x ω (t)=x(t)·ω(t) where ω(t) is the window function, and Perform a weighted average of the data points in the window, where each data point has a specified weight: Where N is the window size, x(ti) is the i-th data point in the window, and ω i is the weight coefficient.

4. The method according to claim 1, characterized in that In step S2, the signal prediction adopts a combination of an autoregressive model and a Kalman filter, and the autoregressive model is used as a state transition model of the Kalman filter; the data reconstruction adopts compressed sensing technology, and the signal is reconstructed through a small amount of measurements using the sparsity of the signal.

5. The method according to claim 4, characterized in that The data reconstruction specifically includes: Using compressed sensing technology, the sparsity of the signal is exploited to reconstruct the signal through a small number of measurements: in, is the estimated value of the reconstructed signal, arg min||1|| indicates the use of sparse representation, Φ is the sparse dictionary matrix, y is the actual measurement value vector, and λ is the regularization parameter.

6. The method according to claim 1, characterized in that In step S3, linear interpolation is used to complete the data: Among them, x(t) represents the data value completed at time t after linear interpolation, t0 and t1 are the times of adjacent known data points, and Δt is the time interval.

7. The method according to claim 1, characterized in that The time stretching technology in step S4 extends the original signal to a longer time series to improve the frequency resolution by splitting and parallel processing the long-time signal, and enhances the low-frequency components of the signal through the enhancement factor.

8. The method according to claim 7, characterized in that The way to extend the original signal to a longer time series to improve the frequency resolution is: x extended (t)=Interp(x(t),new time points) The way to enhance the low-frequency components of the signal to improve the resolution of the analysis is: X enhanced (f)=X(f)·E(f) Among them, X(f) is the frequency domain representation of the original signal, E(f) is the enhancement factor, and the value of E(f) is greater than 1 in the low frequency band and approaches 0 in the high frequency band.

9. A transformer bushing frequency domain dielectric spectrum measurement device, characterized in that: include: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to execute the transformer bushing frequency domain dielectric spectrum measurement method according to any one of claims 1 to 8.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions instruct a computer device to execute operations corresponding to the method according to any one of claims 1 to 8.

Images