Power quality disturbance detection method based on improved adaptive s-transform

Abstract

The application belongs to the technical field of power quality disturbance detection, and particularly relates to a power quality disturbance detection method based on improved adaptive S transform, which comprises the following steps: designing window width control factors of two Gaussian windows, flexibly adjusting time domain window width and spectrum characteristics of the Gaussian windows, and constituting a global adaptive Gaussian window; taking upper and lower limits of time-frequency domain resolution as constraint conditions, taking signal energy concentration degree as an objective function of an optimization scheme, and adopting an interior point method to solve parameters; performing fast Fourier transform on a signal and setting a threshold value, defining a part higher than the threshold value as a characteristic frequency point k i ; screening out calculation at non-characteristic frequency points, performing time-frequency transform only at the characteristic frequency point k i , and constructing a fast algorithm of the improved adaptive S transform; and obtaining a characteristic information value of a final disturbance. The application constructs a global adaptive Gaussian window as a kernel function of the improved adaptive S transform, can automatically adjust effective window length and spectrum of a window function with detection frequency, avoids frequent switching of window parameters for improving time-frequency resolution, and efficiently and accurately detects various power quality disturbances.

Description

Technical Field

[0001] This invention belongs to the field of power quality disturbance detection technology, specifically relating to a power quality disturbance detection method based on an improved adaptive S-transform. Background Technology

[0002] The increasing penetration rate of renewable energy in power grids and the connection of numerous power electronic devices and nonlinear loads to the power system have led to power quality disturbances exhibiting strong randomness, wide bandwidth, and multiple coupling characteristics. These disturbances result in complex and diverse power quality problems, severely impacting industrial production and residential life. Simultaneously, with increasing informatization, various sophisticated electronic devices are placing ever higher demands on power quality. Rapid and accurate detection of various power quality disturbances and extraction of disturbance data, followed by long-term monitoring or intelligent analysis, can provide timely and accurate information on the characteristics of various disturbance events, offering crucial data for power quality assessment and mitigation. However, existing detection methods often struggle to simultaneously meet the requirements of rapid response and high accuracy. Therefore, proposing a power quality disturbance detection method capable of efficiently detecting various disturbances is of great significance for improving the reliability of power supply and the development of smart grid technology.

[0003] Currently, domestic and international methods for detecting power quality disturbances mainly rely on time-frequency analysis methods based on signal processing techniques, such as Short-Time Fourier Transform (SFT), Wavelet Transform, Hilbert Transform, and S-Transform. The SFT has a fixed window function and cannot distinguish disturbances of different frequency components. Wavelet Transform detects and extracts features from different types of disturbance signals through multi-scale decomposition, but it is highly susceptible to noise, leading to large detection errors. Hilbert Transform is prone to mode aliasing when extracting features such as the instantaneous amplitude of disturbance signals, resulting in poor adaptability. The S-Transform can be considered a combination of SFT and wavelet transforms, capable of detecting various types of disturbances. However, it requires traversing all frequency points when calculating the Hadamard product of the window function and the signal, resulting in high computational complexity and a large amount of computation, leading to poor real-time performance and making it difficult to apply in practical engineering. Summary of the Invention

[0004] In response to the increased diversity and randomness of power quality disturbances in power systems due to the penetration of renewable energy, existing detection methods are unable to adequately adapt to the characteristics of power quality disturbances, resulting in poor detection performance, low real-time performance, and an inability to quickly and accurately reflect the specific characteristics of power quality disturbances. This invention provides a power quality disturbance detection method based on an improved adaptive S-transform.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] The power quality disturbance detection method based on the improved adaptive S-transform includes the following steps:

[0007] S1. Design the window width control factors of two Gaussian windows, flexibly adjust the time-domain window width and spectral characteristics of the Gaussian windows, and construct a global adaptive Gaussian window ω(τ―t,f) as the kernel function of the improved adaptive S-transform to fully adapt to the transient power quality disturbance characteristics in the power system.

[0008] S2. A parameter optimization scheme is proposed with the objective function of enhancing energy concentration ECM. The upper and lower limits of the resolution in the time and frequency domains are used as constraints. The interior point method is used to solve the parameters. The results are used as the values ​​of the global adaptive Gaussian window ω(τ―t,f) window width control factor in S1.

[0009] S3. Perform a Fast Fourier Transform on the signal and set a threshold. Points above the threshold are defined as characteristic frequency points k. i ;

[0010] S4. Calculations at non-characteristic frequency points are filtered out, and only the characteristic frequency point k is considered. i We perform time-frequency transformation at the point and construct a fast algorithm for improved adaptive S-transform to obtain a time-frequency transformation matrix containing feature information at the feature frequency point.

[0011] S5. Based on the time-frequency transformation matrix in S4, feature extraction is performed to obtain the final perturbation feature information value.

[0012] The structure of the globally adaptive Gaussian window in S1 is as follows:

[0013]

[0014] Here, the standard deviation (f) is a function of the signal frequency f, with the specific structure σ(f)=1 / |λ1f+λ2|. The standard deviation can be flexibly controlled by two window width adjustment factors λ1 and λ2, which is the basis for enabling the window function to adapt to various disturbance signals. The parameter τ is the time shift factor that controls its position in the time domain, t is the time variable in the time domain signal, and f is the frequency of the detected signal.

[0015] The improved adaptive S-transform in S1 is defined as follows:

[0016]

[0017] Where x(t) is any time-domain signal.

[0018] The parameter optimization scheme in S2, which uses enhancing energy concentration (ECM) as the objective function, includes the following steps:

[0019] S21. After performing an improved adaptive S-transform on the signal, the energy concentration modulus (ECM) is defined as:

[0020]

[0021] The normalized IAST matrix is:

[0022]

[0023] The goal of the parameter tuning scheme is to achieve the highest possible energy concentration (ECM) of the detected signal. IAST (λ1,λ2) is the largest;

[0024] S22, using the trade-off between time and frequency resolution as a constraint;

[0025] S23. Obtain parameter values ​​that can enhance energy concentration while taking into account the trade-off between time and frequency resolution.

[0026] The method in S22 that uses the trade-off between time and frequency resolution as a constraint is as follows:

[0027] The overall constraints are:

[0028] σ low ≤σ(f)≤σ up

[0029] σ low To be the smallest resolvable time interval, for a discrete signal sampling sequence, σ can be... low Replace with nT s T s Let σ be the sampling period of the signal, where n is an integer greater than or equal to 1. low ≤σ(f) is transformed into constraint condition one:

[0030] nT s *(λ1f min +λ2)―1≤0

[0031] Among them, f min T is the minimum frequency to be detected. s The sampling period of the signal;

[0032] σ up This determines the upper limit of frequency resolution; similarly, σ(f) ≤ σ up Transformed into constraint condition two:

[0033] 1 / (λ1f max +λ2)―uT s ≤0

[0034] Among them, f max Based on the sampling theorem, T is determined by the sampling frequency. s The sampling period of the signal.

[0035] The method for determining the parameter values ​​in S23 that can enhance energy concentration while also considering the trade-off between time and frequency resolution is as follows:

[0036] To improve the speed of parameter optimization, it is also necessary to impose reasonable limits on the parameters themselves, which serves as constraint condition three. The overall parameter optimization scheme is as follows:

[0037]

[0038] stnT s *t(λ1f min +λ2)―1≤0

[0039] 1 / (λ1f max +λ2)―uT s ≤0

[0040] λ1∈(―0.1,―0.0015)

[0041] λ2∈(15,35)

[0042] The interior point method is used to solve for the parameters of the above optimization scheme, and the parameter values ​​that can enhance energy concentration and take into account the trade-off between time and frequency resolution are obtained.

[0043] The method for determining the characteristic frequency point of the disturbance characteristic signal in S3 includes the following steps:

[0044] S31. Perform a fast Fourier transform on the signal to obtain the spectrum of the disturbance signal, preliminarily determine the frequency distribution, and determine the main energy component contained in the actual disturbance signal.

[0045] S32. Set a threshold based on the characteristics of the actual disturbance signal, define the frequency components with amplitudes higher than this threshold in the spectrum as the main frequency points, and automatically filter out the remaining non-main frequency points. The number of main frequency points v determined by the automatic threshold method is much smaller than N, thereby reducing the computation time complexity to O(Nlog2N) and reducing the amount of computation.

[0046] The calculation method for filtering out non-characteristic frequency points in S4 includes the following steps:

[0047] S41. First, discretize it, and discretize the time variables τ and f in the improved adaptive S-transform into mT respectively. s and k / NT s ;

[0048] S42. During the two-dimensional time-frequency matrix calculation of the disturbance signal, all rows containing non-dominant frequency points are calculated as zero, and only the dominant frequency point k is calculated. i The discretized expression for the relevant part is:

[0049]

[0050] Where, ki The dominant frequency point of the signal is selected by automatic thresholding, where m and l have values ​​of 0, 1, 2, ..., N-1, and i has values ​​of 1, 2, ..., v,X((l+k) i ) / NT s ) is the discrete form of X(α+f) containing only the dominant frequency, W(l / NT) s ,k i / NT s ) is the discrete Fourier transform of the adaptive Gaussian window.

[0051] The method for extracting the feature information of the final disturbance based on the final time-frequency feature map in S5 includes the following steps:

[0052] S51. First, extract the time-domain feature information of the disturbance signal. In the time-frequency matrix of the two, the time-domain variation curve of a certain fixed frequency component in the disturbance signal can be extracted. Calculate the difference curve of the curve to obtain the start time of the disturbance of the frequency component. Further extract the feature information of the magnitude of the disturbance signal amplitude change from the time-domain variation curve.

[0053] S52. Based on the final two-dimensional time-frequency matrix, extract the frequency domain feature information such as the number of frequency components contained in the disturbance signal. Then, extract the frequency normalization value of the vertical axis in the improved adaptive S-variation two-dimensional time-frequency matrix to accurately obtain the frequency value of each disturbance signal component.

[0054] Compared with the prior art, the beneficial effects of this invention are:

[0055] This invention constructs a globally adaptive Gaussian window as the kernel function of the improved adaptive S-transform. This window automatically adjusts its effective window length and spectrum according to changes in the detection frequency, avoiding frequent switching of window parameters to improve time-frequency resolution. Furthermore, the window parameters are selected with the goal of enhancing signal energy concentration, ensuring accurate time-frequency localization of various disturbances. Finally, a fast algorithm is constructed to perform time-frequency transformation on the disturbance signal, resulting in lower computational complexity, better real-time performance, and stronger time-frequency resolution, enabling rapid and accurate detection of various disturbances. Attached Figure Description

[0056] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0057] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0058] Figure 1 The flowchart of the improved fast algorithm for adaptive S-transform of this invention is shown below;

[0059] Figure 2 This is a time-frequency analysis diagram of IAST from the present invention;

[0060] Figure 3 The graphs show the fundamental frequency amplitude curve and the frequency amplitude curve of this invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. These descriptions are only for further illustrating the features and advantages of the present invention, and not for limiting the claims of the present invention. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0063] Step S1: Design the window width control factors of two Gaussian windows, flexibly adjust the time-domain window width and spectral characteristics of the Gaussian windows, and construct a global adaptive Gaussian window ω(τ―t,f) as the kernel function of the improved adaptive S-transform to fully adapt to the transient power quality disturbance characteristics in the power system.

[0064] Step S2: Propose a parameter tuning scheme with enhanced energy concentration ECM as the objective function, take the upper and lower limits of the time and frequency domain resolution as constraints, use the interior point method to solve the parameters, and use the result as the value of the global adaptive Gaussian window ω(τ―t,f) window width control factor in Step 1.

[0065] Step S3: Perform a Fast Fourier Transform on the signal and set a threshold. Points above the threshold are defined as characteristic frequency points k. i ;

[0066] Step S4: Filter out calculations at non-feature frequency points, only performing calculations at feature frequency points k.i We perform time-frequency transformation at the point and construct a fast algorithm for improved adaptive S-transform to obtain a time-frequency transformation matrix containing feature information at the feature frequency point.

[0067] Step S5: Based on the time-frequency transformation matrix in step 4, perform feature extraction to obtain the final perturbation feature information value;

[0068] Preferably, the structure of the globally adaptive Gaussian window is as follows:

[0069]

[0070] In the formula, the standard deviation σ(f) is a function of the signal frequency f, specifically σ(f) = 1 /

[0071] The standard deviation, |λ1f+λ2|, can be flexibly controlled by two window width adjustment factors, λ1 and λ2. This is the basis for enabling the window function to adapt to various disturbance signals. The parameter τ is a time-shift factor that controls its position in the time domain, t is the time variable in the time domain signal, and f is the frequency of the detected signal.

[0072] For any time-domain signal x(t), the constructed improved adaptive S-transform is defined as follows:

[0073]

[0074] Step S21: After performing an improved adaptive S-transform on the signal, the energy concentration (ECM) is defined as follows:

[0075]

[0076] The normalized IAST matrix is:

[0077]

[0078] The goal of the parameter tuning scheme is to achieve the highest possible energy concentration (ECM) of the detected signal. IAST (λ1,λ2) is the largest

[0079] Step S22: Using the trade-off between time and frequency resolution as a constraint, the overall constraint condition is:

[0080] σ low ≤σ(f)≤σ up

[0081] σ low To be the smallest resolvable time interval, for a discrete signal sampling sequence, σ can be... low Replace with nT s T s Let σ be the sampling period of the signal, where n is an integer greater than or equal to 1. low≤σ(f) is transformed into constraint condition one:

[0082] nT s *(λ1f min +λ2)―1≤0

[0083] In the formula, f min T is the minimum frequency to be detected. s The sampling period of the signal

[0084] σ up This determines the upper limit of frequency resolution; similarly, σ(f) ≤ σ up Transformed into constraint condition two:

[0085] 1 / (λ1f max +λ2)―uT s ≤0

[0086] In the formula, f max Based on the sampling theorem, T is determined by the sampling frequency. s The sampling period of the signal

[0087] Step S23: To improve the speed of parameter optimization, it is also necessary to impose reasonable limits on the parameters themselves, as constraint condition three. The overall parameter optimization scheme is as follows:

[0088]

[0089] stnT s *(λ1f min +λ2)―1≤0

[0090] 1 / (λ1f max +λ2)―uT s ≤0

[0091] λ1∈(―0.1,―0.0015)

[0092] λ2∈(15,35)

[0093] The interior point method is used to solve for the parameters of the above optimization scheme, and the parameter values ​​that can enhance energy concentration and take into account the trade-off between time and frequency resolution are obtained.

[0094] Preferably, the specific steps for determining the characteristic frequency points of the disturbance characteristic signal are as follows:

[0095] Step S31: Perform a fast Fourier transform on the signal to obtain the spectrum of the disturbance signal, preliminarily determine the frequency distribution, and determine the main energy component contained in the actual disturbance signal.

[0096] Step S32: Set a threshold according to the characteristics of the actual disturbance signal, define the frequency components with amplitudes higher than this threshold in the spectrum as the main frequency points, and automatically filter out the remaining non-main frequency points. The number of main frequency points v determined by the automatic threshold method is much smaller than N, thereby reducing the time complexity of the calculation to O(Nlog2N) and reducing the amount of computation.

[0097] Preferably, the calculation for filtering out non-feature frequency points is performed only on the feature frequency point k. i The specific steps for improving the time-frequency transform and constructing an improved adaptive S-transform fast algorithm are as follows:

[0098] Step S41: To facilitate the implementation of the IAFST algorithm in a digital signal processor, it is first discretized, and the time variables τ and f in the improved adaptive S-transform are discretized into mT respectively. s and k / NT s .

[0099] Step S42: During the two-dimensional time-frequency matrix calculation of the disturbance signal, rows containing non-dominant frequency points are calculated as zero, and only the dominant frequency point k is calculated. i The discretized expression for the relevant part is:

[0100]

[0101] In the formula, k i The dominant frequency point of the signal is selected by automatic thresholding, where m and l have values ​​of 0, 1, 2, ..., N-1, and i has values ​​of 1, 2, ..., v,X((l+k) i ) / NT s ) is the discrete form of X(α+f) in equation (9) containing only the main frequency, W(l / NT) s ,k i / NT s (IAFST) is the Discrete Fourier Transform of an Adaptive Gaussian Window. The specific implementation flow of the IAFST algorithm is as follows: Figure 1 As shown.

[0102] Preferably, the specific steps for extracting the feature information of the final disturbance based on the final time-frequency feature map are as follows:

[0103] Step S51: First, extract the time-domain feature information of the disturbance signal. The time-domain variation curve of a certain fixed frequency component in the disturbance signal can be extracted from the time-frequency matrix of the two. The difference curve of the curve is calculated to obtain the start time of the disturbance of the frequency component. Further feature information such as the magnitude of the disturbance signal amplitude change is extracted from the time-domain variation curve.

[0104] Step S52: Based on the final two-dimensional time-frequency matrix, extract the frequency domain feature information such as the number of frequency components contained in the disturbance signal. Then, extract the frequency normalization value of the vertical axis in the improved adaptive S-variation two-dimensional time-frequency matrix to accurately obtain the frequency value of each disturbance signal component.

[0105] The technical solution claimed in this invention will be further described below through a specific embodiment, which adopts the above-mentioned real-time power quality disturbance detection method based on the improved adaptive S-transform.

[0106] In this embodiment, the power quality disturbance signal to be detected is a composite disturbance signal containing multiple frequency components, and the specific mathematical model is shown in the following formula:

[0107]

[0108] Where f0 is the power frequency of 50Hz, ε(t) is the step function, and the signal contains a voltage dip of 0.6pu within 0.08-0.16s, accompanied by the 5th and 14th time-varying harmonics between 0.08s-0.18s and 0.02s-0.12s, and a transient oscillation with a frequency of 29 times the power frequency within 0.11s-0.14s.

[0109] To verify the effectiveness of this method, firstly, an improved adaptive S-transform analysis is performed on the perturbation signal to obtain a two-dimensional time-frequency matrix, as shown below. Figure 2 As shown.

[0110] Depend on Figure 2 It can be seen that IAST, which uses the enhanced energy concentration optimization scheme as the parameter selection basis, has the best energy concentration performance and maintains good time-frequency resolution across the entire time-frequency plane. This is beneficial for improving the accuracy of perturbation feature extraction. The extracted fundamental frequency amplitude curve and frequency amplitude curve are shown below. Figure 3 As shown

[0111] Depend on Figure 3 It can be seen that the improved adaptive S-transform can accurately extract the amplitude transformation trend of the disturbance at the fundamental frequency and extract the precise values ​​of each frequency component.

[0112] To further verify the detection effect of this method in practical applications, a power quality disturbance detection platform was built. The disturbance signal source was implemented using the dSPACE hardware-in-the-loop simulation platform based on Matlab / Simulink. The power quality signal acquisition chip ADE9000 acquired the power quality disturbance signal in real time. The DSP used was a TMS320F28377D with a dual-core architecture. The DSP and ADE9000 communicated via SPI to exchange data in real time, acquiring the power signal from the acquisition chip and executing the detection algorithm.

[0113] Five types of composite disturbance signals, encompassing both time and frequency domains, were used to test the accuracy and speed of the algorithm's detection. Accuracy was measured by the relative error between the detected and true values ​​of each disturbance's characteristic information. The main characteristic information for each disturbance included amplitude, duration, and frequency components. The relative error value for the composite disturbance test was the average relative error of each component. The accuracy results are shown in Table 1.

[0114] Table 1. Accuracy of IAFST Detection for Different Types of Disturbances

[0115]

[0116] As shown in Table 1, IAFST can adapt to the characteristics of both single and composite disturbances, including both time and frequency domains, and thus extract the disturbance duration, amplitude, and other characteristic information more accurately.

[0117] To verify the real-time performance of various algorithms in practical applications, this paper tests the execution time of the method for different types of disturbance detection in the CCS development platform environment of DSP. The test results are shown in Table 2.

[0118] Table 2. Execution schedule for IAFST detection of different types of disturbances.

[0119]

[0120] As shown in Table 2, IAFST demonstrates good real-time performance while ensuring detection accuracy. This is due to the universality of the global adaptive Gaussian window and the fast calculation scheme based on the automatic threshold method, which further verifies the effectiveness of the proposed method.

[0121] The above description only illustrates the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and all such changes should be included within the protection scope of the present invention.

Claims

1. A power quality disturbance detection method based on improved adaptive S-transform, characterized in that, Includes the following steps: S1. Design the window width control factors for two Gaussian windows, flexibly adjust the time-domain window width and spectral characteristics of the Gaussian windows, and construct a globally adaptive Gaussian window. As an improved adaptive S-transform kernel function, it can fully adapt to the transient power quality disturbance characteristics in the power system; The structure of the globally adaptive Gaussian window is as follows: Among them, standard deviation It is a function of the signal frequency f, with the following structure: The standard deviation can be adjusted by two window width factors. and Flexible control is the foundation for window functions to adapt to various disturbance signals; parameters To control its position in the time domain, a time shift factor is used, where t is the time variable in the time domain signal and f is the frequency of the detection signal; The improved adaptive S-transform definition is as follows: = Where x(t) is any time-domain signal; S2. A parameter tuning scheme is proposed with enhancing the energy concentration (ECM) as the objective function. The upper and lower limits of the time-frequency domain resolution are used as constraints. The interior-point method is used to solve for the parameters, and the results are used as the global adaptive Gaussian window in S1. The value of the window width control factor; S21. After performing an improved adaptive S-transform on the signal, the energy concentration modulus (ECM) is defined as: The normalized IAST matrix is: The goal of the parameter tuning scheme is to increase the energy concentration of the detected signal. maximum; S22, using the trade-off between time and frequency resolution as a constraint; S23. Obtain parameter values ​​that can enhance energy concentration while taking into account the trade-off between time and frequency resolution; S3. Perform a Fast Fourier Transform on the signal and set a threshold. Points above the threshold are defined as characteristic frequency points k. i ; S4. Calculations at non-characteristic frequency points are filtered out, and only the characteristic frequency point k is considered. i We perform time-frequency transformation at the point and construct a fast algorithm for improved adaptive S-transform to obtain a time-frequency transformation matrix containing feature information at the feature frequency point. S5. Based on the time-frequency transformation matrix in S4, feature extraction is performed to obtain the final perturbation feature information value.

2. The power quality disturbance detection method based on improved adaptive S-transform according to claim 1, characterized in that, The method in S22 that uses the trade-off between time and frequency resolution as a constraint is as follows: The overall constraints are: To determine the smallest resolvable time interval, for a discrete signal sampling sequence, it can be... Replace with , Let n be the sampling period of the signal, where n is an integer greater than or equal to 1. Transformed into constraint condition one: in, The minimum frequency for detection, The sampling period of the signal; This determines the upper limit of frequency resolution; similarly, it can be... Transformed into constraint condition two: in, Based on the sampling theorem, the sampling frequency determines the frequency. The sampling period of the signal.

3. The power quality disturbance detection method based on improved adaptive S-transform according to claim 2, characterized in that, The method for determining the parameter values ​​in S23 that can enhance energy concentration while also considering the trade-off between time and frequency resolution is as follows: To improve the speed of parameter optimization, it is also necessary to impose reasonable limits on the parameters themselves, which serves as constraint condition three. The overall parameter optimization scheme is as follows: The interior point method is used to solve for the parameters of the above optimization scheme, and the parameter values ​​that can enhance energy concentration and take into account the trade-off between time and frequency resolution are obtained.

4. The power quality disturbance detection method based on improved adaptive S-transform according to claim 1, characterized in that, The method for determining the characteristic frequency point of the disturbance characteristic signal in S3 includes the following steps: S31. Perform a fast Fourier transform on the signal to obtain the spectrum of the disturbance signal, preliminarily determine the frequency distribution, and determine the main energy component contained in the actual disturbance signal. S32. Set a threshold based on the characteristics of the actual disturbance signal, define the frequency components with amplitudes higher than this threshold in the spectrum as the main frequency points, and automatically filter out the remaining non-main frequency points. The number of main frequency points v determined by the automatic threshold method is much smaller than N, thereby reducing the computation time complexity to O(Nlog2N) and reducing the amount of computation.

5. The power quality disturbance detection method based on the improved adaptive S-transform according to claim 4, characterized in that, The calculation method for filtering out non-characteristic frequency points in S4 includes the following steps: S41. First, discretize it to improve the time variable in the adaptive S-transform. Discretize f and mT respectively s and k / NT s ; S42. During the two-dimensional time-frequency matrix calculation of the disturbance signal, all rows containing non-dominant frequency points are calculated as zero, and only the dominant frequency point k is calculated. i The discretized expression for the relevant part is: Where, k i The dominant frequency point of the signal is selected by automatic thresholding, where m and l have values ​​of 0, 1, 2, ..., N-1, and i has values ​​of 1, 2, ..., v, X((l+k) i ) / NT s ) is the discrete form of X(f+f) containing only the main frequency, W(l / NT) s , k i / NT s ) is the discrete Fourier transform of the adaptive Gaussian window.

6. The power quality disturbance detection method based on improved adaptive S-transform according to claim 1 is characterized in that, The method for extracting the feature information of the final disturbance based on the final time-frequency feature map in S5 includes the following steps: S51. First, extract the time-domain feature information of the disturbance signal. In the time-frequency matrix of the two, the time-domain variation curve of a certain fixed frequency component in the disturbance signal can be extracted. Calculate the difference curve of the curve to obtain the start time of the disturbance of the frequency component. Further extract the feature information of the magnitude of the disturbance signal amplitude change from the time-domain variation curve. S52. Based on the final two-dimensional time-frequency matrix, extract the frequency domain feature information such as the number of frequency components contained in the disturbance signal. Then, extract the frequency normalization value of the vertical axis in the improved adaptive S-variation two-dimensional time-frequency matrix to accurately obtain the frequency value of each disturbance signal component.

Images