Primary frequency modulation large disturbance signal peak value analysis method

Abstract

The invention relates to the technical field of signal analysis, and discloses a primary frequency modulation large disturbance signal peak value analysis method, which comprises the following steps: collecting a primary frequency modulation large disturbance signal and carrying out filtering processing of multi-band noise separation; performing multi-scale decomposition on the filtered disturbance signal, extracting a dynamic entropy of a multi-scale signal sequence, and performing dynamic entropy denoising on the disturbance signal to obtain a denoised disturbance signal; and calculating a potential peak position of the denoised disturbance signal, and carrying out multi-condition verification on the potential peak position to obtain peak information for frequency modulation disturbance analysis. Based on spectrum energy distribution, specific wavelet bases are selected for different noise frequency bands for noise reduction, the wavelet noise reduction performance is improved, the energy distribution complexity and the structure complexity of the disturbance signals on different scales are extracted, the dynamic characteristics of the disturbance signals are captured more comprehensively, the noise reduction precision is improved, and the noise reduction efficiency is improved. And the accuracy of wind power integration fault identification and equipment life monitoring in the peak value analysis process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal analysis, and particularly to a method for analyzing the peak value of a large disturbance signal in primary frequency modulation. Background Art

[0002] With the high proportion of new energy connected to the power grid, the primary frequency modulation of thermal power units needs to cope with more frequent and larger-scale non-stationary and multi-harmonic superposition characteristics. For example, the randomness of wind power grid-connected load leads to an enlarged fluctuation range of the amplitude of power grid frequency disturbance, and the disturbance signal often presents non-stationary and multi-harmonic superposition characteristics, so that the primary frequency modulation signal has large disturbance information, and it is necessary to analyze the peak value of the large disturbance signal in frequency modulation. According to the peak value analysis result, the faults of wind turbines are identified and the operating frequency is adjusted. However, the following challenges exist in the peak value analysis process: the sensor data in the wind power environment is affected by external noise, environmental interference and power grid harmonics, resulting in a large peak value extraction error. The large disturbance signal of wind power has both high-frequency short-time characteristics and low-frequency long-time changes, and it is difficult to extract key features at a single scale. Traditional peak detection methods (such as fixed time window moving average and frequency domain filtering) rely on empirical threshold setting and are prone to peak misdetection under dynamic working conditions. Summary of the Invention

[0003] In view of this, the present invention provides a method for analyzing the peak value of a large disturbance signal in primary frequency modulation, adaptively selects wavelet bases of spectra in different frequency bands according to the spectral energy distribution, selects specific wavelet bases for noise reduction for different noise frequency bands, combines multi-scale signal sequences to perform multi-condition verification on potential peak positions, obtains the peak value information of the large disturbance signal in primary frequency modulation, performs frequency modulation disturbance analysis based on the peak value information, identifies the faults of wind power grid connection, and monitors the equipment life of wind power grid connection to reduce equipment loss.

[0004] To achieve the above object, a method for analyzing the peak value of a large disturbance signal in primary frequency modulation provided by the present invention includes the following steps:

[0005] S1: Collect a large disturbance signal in primary frequency modulation and perform filtering processing for multi-band noise separation to obtain a disturbed signal after filtering processing;

[0006] S2: Perform multi-scale decomposition on the disturbed signal after filtering processing to obtain a multi-scale signal sequence of the disturbed signal, extract the dynamic entropy of the multi-scale signal sequence, perform dynamic entropy denoising on the disturbed signal to obtain a denoised disturbed signal;

[0007] S3: Calculate the potential peak position of the denoised disturbed signal;

[0008] S4: Combine the denoised disturbed signal, perform multi-condition verification on the potential peak position to obtain the peak value information of the large disturbance signal in primary frequency modulation, and perform frequency modulation disturbance analysis based on the peak value information.

[0009] As a further improvement method of the present invention:

[0010] Optionally, perform filtering processing for multi-band noise separation on the primary frequency modulation large disturbance signal, including:

[0011] Adopt a sliding window method to calculate the local trend of the primary frequency modulation large disturbance signal, perform detrending processing on the primary frequency modulation large disturbance signal, and perform spectral analysis on the detrended primary frequency modulation large disturbance signal by using the Fourier transform method to obtain the frequency domain representation of the detrended primary frequency modulation large disturbance signal;

[0012] Calculate the energy of the frequency domain representation in different frequency bands to form a spectral energy distribution. Take the spectral energy distribution as an input feature, use a wavelet basis action selection model to generate an optimal wavelet basis sequence corresponding to the input feature, and perform wavelet transform processing for multi-band noise separation on the detrended primary frequency modulation large disturbance signal to obtain a filtered disturbance signal, where the optimal wavelet basis sequence is the optimal wavelet basis of the detrended primary frequency modulation large disturbance signal in different frequency bands.

[0013] Optionally, the wavelet basis action selection model is a reinforcement learning model, including a state space, an action space, and a policy matrix. The state space includes all spectral energy distributions, the action space includes multiple wavelet bases, and the policy matrix is composed of policy values between the spectral energy distribution and the wavelet basis sequence. The wavelet basis sequence is a sequence composed of E wavelet bases. The wavelet basis action selection model takes the spectral energy distribution as an input feature, selects the wavelet basis sequence with the largest policy value between the input feature and the policy matrix as the optimal wavelet basis sequence, and the policy matrix is a model parameter to be optimized and solved;

[0014] Construct a reward function for evaluating the multi-band noise separation effect of the selected wavelet basis sequence in the policy matrix. The variables of the reward function are the spectral energy distribution and the wavelet basis sequence, and the expression of the reward function is:

[0015] ;

[0016] ;

[0017] ;

[0018] Where:

[0019] represents the reward function, represents the spectral energy distribution input to the reward function, represents the wavelet basis sequence input to the reward function, Represents the e-th wavelet basis in the wavelet basis sequence. The e-th wavelet basis has a corresponding relationship with the e-th frequency band, and the e-th frequency band is subjected to wavelet transform using the e-th wavelet basis;

[0020] Represents the signal-to-noise ratio improvement amount of the signal after wavelet transform of the corresponding signal using the wavelet basis sequence S ; Represents the time consumption of wavelet transform of the corresponding signal using the wavelet basis sequence S ; Represents the smoothness of the wavelet basis sequence S;

[0021] Successively represent the hyperparameter weights of the signal-to-noise ratio improvement amount, time consumption, and smoothness.

[0022] Optionally, based on the reward function, construct a loss function for the policy matrix:

[0023] ;

[0024] ;

[0025] Where:

[0026] Represents the loss function of the policy matrix, Represents the i-th group of spectral energy distributions in the state space, Represents the j-th group of wavelet basis sequences constructed based on the wavelet basis in the action space, Represents the number of spectral energy distributions in the state space, Represents the number of constructed wavelet basis sequences;

[0027] Represents the policy matrix, ;

[0028] Solve the policy matrix based on the loss function, and construct a wavelet basis action selection model using the solved policy matrix.

[0029] Optionally, perform multi-scale decomposition on the filtered disturbance signal to obtain scale signals of the disturbance signal at different scales, and form a multi-scale signal sequence, including:

[0030] Initialize the center frequencies of the perturbation signal at A scales and the scale signal spectrum, and initialize the Lagrange multipliers. Iterate the scale signal spectrum based on the center frequencies and the Lagrange multipliers, perform weighted calculation on the scale signal spectrum, iteratively update the center frequencies, and update the Lagrange multipliers using a dynamic step size update method until the iterative change amount of the scale signal spectrum is lower than a preset change threshold, to obtain the scale signal spectrum of the perturbation signal y at A scales, where A represents the number of scales. Perform inverse Fourier transform on the scale signal spectrum to obtain the scale signals at A scales, and form a multi-scale signal sequence.

[0031] Optionally, perform singular value decomposition on the multi-scale signal sequence to obtain the singular value sequence of the multi-scale signal sequence, and extract the energy entropy and singular value entropy of the multi-scale signal sequence respectively, including:

[0032] The calculation formula of the energy entropy is:

[0033] ;

[0034] Where:

[0035] represents the scale signal spectrum of the scale signal at the -th scale in the multi-scale signal sequence, represents the energy of the scale signal spectrum ; ;

[0036] represents the energy entropy of the multi-scale signal sequence;

[0037] The calculation formula of the singular value entropy is:

[0038] ;

[0039] Where:

[0040] represents the -th singular value in the singular value sequence, represents the singular value entropy.

[0041] Optionally, based on the modal entropy and the singular value entropy, extract the dynamic entropy of the multi-scale signal sequence, and perform dynamic entropy denoising on the perturbation signal to obtain the denoised perturbation signal, including:

[0042] The calculation formula of the dynamic entropy is:

[0043] ;

[0044] Where:

[0045] represents the dynamic entropy;

[0046] Use the dynamic entropy to perform dynamic denoising on the perturbation signal, where the dynamic denoising formula is:

[0047] ;

[0048] ;

[0049] where:

[0050] represents the inverse Fourier transform;

[0051] represents a preset control parameter, represents the scale signal spectrum corresponding transformation weight;

[0052] represents the denoised perturbation signal after dynamic denoising.

[0053] Optionally, calculate the potential peak position of the denoised perturbation signal, including:

[0054] Perform sliding window processing on the denoised perturbation signal using a peak detection sliding window, and calculate the mean value of the signal values within the sliding window and the standard deviation std of the signal values to generate an adaptive threshold for the sliding window :

[0055] ;

[0056] where:

[0057] represents a preset adaptive control parameter;

[0058] Take the signal positions where the signal values within the sliding window are higher than the adaptive threshold as the potential peak positions.

[0059] Optionally, combine the denoised perturbation signal and perform multi-condition verification on the potential peak positions to obtain the peak information of the primary frequency modulation large perturbation signal, including:

[0060] The multi-condition verification method includes: in the denoised perturbation signal, if the signal values before and after the potential peak position maintain a monotonic trend, then the first condition verification of the potential peak position passes; if the potential peak position is also a potential peak position in at least scale signals, then the second condition verification of the potential peak position passes;

[0061] Take the potential peak positions after the first condition verification is passed and the second condition verification is passed as peaks; extract the peak information of the peaks, where the peak information is the signal value of the denoised disturbance signal at the peaks.

[0062] To solve the above problems, the present invention provides an electronic device, which includes:

[0063] A memory that stores at least one instruction;

[0064] A communication interface that enables communication of the electronic device; and

[0065] A processor that executes the instructions stored in the memory to implement the above-mentioned peak analysis method for a large disturbance signal of primary frequency regulation.

[0066] To solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned peak analysis method for a large disturbance signal of primary frequency regulation.

[0067] Compared with the prior art, the present invention proposes a peak analysis method for a large disturbance signal of primary frequency regulation, and this technology has the following advantages:

[0068] First of all, this solution proposes a multi-band adaptive noise reduction method. A wavelet basis action selection model is constructed by using reinforcement learning. A reward function is constructed based on the smoothness of the wavelet basis sequence, the improvement of the signal-to-noise ratio, and the time consumption of wavelet processing. The policy matrix in the wavelet basis action selection model is optimized and solved. Different wavelet bases of the spectrum in different frequency bands are adaptively selected according to the spectrum energy distribution. Specific wavelet bases are selected for noise reduction for different noise frequency bands. A smooth wavelet basis sequence is set based on the reward function, and the denoising amplitude is reduced near the peak signal to avoid important information being misfiltered and improve the filtering and noise reduction performance.

[0069] At the same time, this solution proposes a dynamic feature capture and peak analysis method. By combining the energy entropy and singular value entropy of the spectrum, the dynamic entropy of the disturbance signal is calculated. The energy entropy can measure the complexity of the energy distribution of the disturbance signal at different scales, and the singular value entropy is used to measure the structural complexity of the disturbance signal. It can effectively detect abnormal noise. By combining the two to construct an adaptive threshold, the dynamic features of the disturbance signal can be captured more comprehensively, and the denoising accuracy can be improved. According to the distribution characteristics of the denoised disturbance signal in the sliding window, the potential peak positions are identified, and the potential peak positions are verified by multiple conditions in combination with the multi-scale signal sequence to obtain the peak information of the large disturbance signal of primary frequency regulation. Based on the peak information, the frequency modulation disturbance analysis is carried out to identify the wind power grid connection fault and the equipment life monitoring of the wind power grid connection, and reduce the equipment loss. Description of the Drawings

[0070] Figure 1 The flowchart of a method for analyzing the peak value of a primary frequency modulation large disturbance signal provided by an embodiment of the present invention.

[0071] The realization, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Specific embodiments

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] An embodiment of the present application provides a method for analyzing the peak value of a primary frequency modulation large disturbance signal. The execution subject of the method for analyzing the peak value of the primary frequency modulation large disturbance signal includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for analyzing the peak value of the primary frequency modulation large disturbance signal can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0074] Refer to Figure 1 , Embodiment 1 of the present invention is:

[0075] A method for analyzing the peak value of a primary frequency modulation large disturbance signal includes the following steps:

[0076] S1: Collect a primary frequency modulation large disturbance signal and perform a filtering process for multi-band noise separation to obtain a disturbed signal after the filtering process.

[0077] Performing a filtering process for multi-band noise separation on the primary frequency modulation large disturbance signal includes:

[0078] Using a sliding window method to calculate the local trend of the primary frequency modulation large disturbance signal, perform a detrending process on the primary frequency modulation large disturbance signal, and perform a spectrum analysis on the detrended primary frequency modulation large disturbance signal by means of Fourier transform to obtain the frequency domain representation of the detrended primary frequency modulation large disturbance signal; As an embodiment of the present invention, the representation form of the primary frequency modulation large disturbance signal is :

[0079] ;

[0080] Wherein:

[0081] are N signal values of the primary frequency modulation large disturbance signal x, represents the nth signal value in the primary frequency modulation large disturbance signal x, , N represents the length of the primary frequency regulation large disturbance signal x;

[0082] The detrending processing formula for the primary frequency regulation large disturbance signal x is:

[0083] ;

[0084] ;

[0085] Where:

[0086] represents the detrended result of the signal value of, is the sliding window length, and the negative information in the detrending processing formula is the local trend of the primary frequency regulation large disturbance signal;

[0087] represents the frequency regulation large disturbance signal after detrending processing;

[0088] Calculate the energy of the frequency domain representation in different frequency bands to form a spectrum energy distribution. Using the spectrum energy distribution as input features, use the wavelet basis action selection model to generate the optimal wavelet basis sequence corresponding to the input features, and perform wavelet transform processing for multi-band noise separation on the frequency regulation large disturbance signal after detrending processing to obtain the filtered disturbance signal, where the optimal wavelet basis sequence is the optimal wavelet basis of the frequency regulation large disturbance signal after detrending processing in different frequency bands. Specifically, the calculation formula for the spectrum energy distribution is:

[0089] ;

[0090] ;

[0091] Where:

[0092] represents the frequency regulation large disturbance signal after detrending processing of, f represents frequency information; represents the frequency domain representation of the spectrum energy distribution, represents the frequency domain representation in the e-th frequency band, , E represents the total number of frequency bands, where the frequency range of the e-th frequency band is , is the frequency upper limit of the e-th frequency band;

[0093] represents the differential of the frequency information f.

[0094] The wavelet basis action selection model is a reinforcement learning model, including a state space, an action space, and a policy matrix. The state space includes all spectral energy distributions, the action space includes multiple wavelet bases, and the policy matrix is composed of policy values between the spectral energy distribution and the wavelet basis sequence. The wavelet basis sequence is a sequence composed of E wavelet bases. The wavelet basis action selection model uses the spectral energy distribution as the input feature, and selects the wavelet basis sequence with the largest policy value between the input feature and the policy matrix as the optimal wavelet basis sequence. The policy matrix is the model parameter to be optimized and solved.

[0095] Construct a reward function for evaluating the multi-band noise separation effect of the selected wavelet basis sequence in the policy matrix. The variables of the reward function are the spectral energy distribution and the wavelet basis sequence. The expression of the reward function is:

[0096] ;

[0097] ;

[0098] ;

[0099] Where:

[0100] represents the reward function, represents the spectral energy distribution input to the reward function, represents the wavelet basis sequence input to the reward function, represents the e-th wavelet basis in the wavelet basis sequence. The e-th wavelet basis has a corresponding relationship with the e-th frequency band, and the e-th frequency band is subjected to wavelet transform using the e-th wavelet basis;

[0101] represents the signal-to-noise ratio improvement of the signal after wavelet transform of using the wavelet basis sequence S, represents the time consumption of wavelet transform of using the wavelet basis sequence S for the corresponding signal, represents the smoothness of the wavelet basis sequence S;

[0102] represent the hyperparameter weights of the signal-to-noise ratio improvement, time consumption, and smoothness in sequence; specifically, .

[0103] Based on the reward function, construct the loss function of the policy matrix:

[0104] ;

[0105] ;

[0106] Wherein:

[0107] represents the loss function of the said policy matrix, represents the i-th group of spectrum energy distribution in the state space, represents the j-th group of wavelet basis sequences constructed based on the wavelet basis in the action space, represents the number of spectrum energy distributions in the state space, represents the number of constructed wavelet basis sequences;

[0108] represents the policy matrix, ;

[0109] Solve the said policy matrix based on the loss function, and construct a wavelet basis action selection model by using the solved policy matrix. Specifically, the solution method of the said policy matrix is the Adam optimization algorithm.

[0110] S2: Perform multi-scale decomposition on the filtered perturbation signal to obtain a multi-scale signal sequence of the perturbation signal, extract the dynamic entropy of the multi-scale signal sequence, and perform dynamic entropy denoising on the perturbation signal to obtain the denoised perturbation signal.

[0111] Perform multi-scale decomposition on the said filtered perturbation signal to obtain scale signals of the perturbation signal at different scales, and form a multi-scale signal sequence, including:

[0112] The representation form of the said perturbation signal is:

[0113] ;

[0114] Wherein:

[0115] represents the filtered perturbation signal, represents N signal values of the perturbation signal y, represents the n-th signal value of the perturbation signal y;

[0116] Initialize the center frequencies and scale signal spectra of the said perturbation signal at A scales, and initialize the Lagrange multipliers. Iterate the scale signal spectra based on the center frequencies and Lagrange multipliers, perform weighted calculation on the scale signal spectra, iteratively update the center frequencies, and update the Lagrange multipliers by using a dynamic step size update method until the iterative change amount of the scale signal spectra is lower than a preset change threshold to obtain the scale signal spectra of the said perturbation signal y at A scales, where A represents the number of scales. Perform inverse Fourier transform on the scale signal spectra to obtain scale signals at A scales, and form a multi-scale signal sequence. Among them, the The spectral iteration formula for the scale signal at a certain scale is as follows:

[0117] ;

[0118] where:

[0119] represents the spectral of the scale signal at the th scale obtained from the t-th iteration, represents the center frequency of the t-th iteration result of the scale signal at the th scale, represents the preset maximum number of iterations, ;

[0120] represents the t-th iteration result of the Lagrange multiplier, represents 's spectrum;

[0121] represents the average value of the center frequencies of the t-th iteration result,

[0122] ;

[0123] The iteration variation is: , where represents the L1 norm, represents the square of the L1 norm;

[0124] The iteration formula for the center frequency is:

[0125] ;

[0126] where:

[0127] represents the center frequency information, represents the differential of the center frequency information;

[0128] The iteration formula for the Lagrange multiplier is:

[0129] ;

[0130] ;

[0131] where:

[0132] represents the scale signal corresponding to the spectral of the scale signal ;

[0133] denotes the Lagrange multiplier and the iteration step size, denotes the preset initial step size.

[0134] Performing singular value decomposition on the multi-scale signal sequence to obtain the singular value sequence of the multi-scale signal sequence, and respectively extracting the energy entropy and singular value entropy of the multi-scale signal sequence, including:

[0135] The calculation formula for the energy entropy is:

[0136] ;

[0137] where:

[0138] denotes the scale signal spectrum of the scale signal of the th scale in the multi-scale signal sequence, denotes the scale signal spectrum energy, ;

[0139] denotes the energy entropy of the multi-scale signal sequence;

[0140] The calculation formula for the singular value entropy is:

[0141] ;

[0142] where:

[0143] denotes the th singular value in the singular value sequence, denotes the singular value entropy.

[0144] Specifically, converting the multi-scale signal sequence into a multi-scale signal matrix, where the multi-scale signal matrix is a matrix with A rows, and each row corresponds to a set of scale signals, and performing singular value decomposition on the multi-scale signal matrix to obtain A singular values.

[0145] Based on the modal entropy and the singular value entropy, extracting the dynamic entropy of the multi-scale signal sequence, and performing dynamic entropy denoising on the perturbation signal to obtain the denoised perturbation signal, including:

[0146] The calculation formula for the dynamic entropy is:

[0147] ;

[0148] where:

[0149] denotes the dynamic entropy;

[0150] Perform dynamic denoising on the perturbation signal using the dynamic entropy, where the dynamic denoising formula is:

[0151] ;

[0152] ;

[0153] Where:

[0154] represents the inverse Fourier transform;

[0155] represents a preset control parameter, represents the scale signal spectrum The corresponding transformation weight;

[0156] represents the denoised perturbation signal after dynamic denoising, , represents the denoised perturbation signal The nth signal value.

[0157] S3: Calculate the potential peak position of the denoised perturbation signal.

[0158] Calculating the potential peak position of the denoised perturbation signal includes:

[0159] Perform sliding window processing on the denoised perturbation signal using a peak detection sliding window, and calculate the mean value of the signal values within the sliding window And the standard deviation std of the signal values to generate an adaptive threshold for the sliding window :

[0160] ;

[0161] Where:

[0162] represents a preset adaptive control parameter;

[0163] Take the signal positions within the sliding window where the signal values are higher than the adaptive threshold As the potential peak positions.

[0164] S4: Combine the denoised perturbation signal, perform multi-condition verification on the potential peak positions, obtain the peak information of the primary frequency modulation large perturbation signal, and perform frequency modulation perturbation analysis based on the peak information.

[0165] The combination of the denoised perturbation signal, performing multi-condition verification on the potential peak positions, and obtaining the peak information of the primary frequency modulation large perturbation signal includes:

[0166] The multi - condition verification method includes: in the perturbed signal after denoising, if the signal values before and after the potential peak position maintain a monotonic trend, the first - condition verification of the potential peak position passes; if the potential peak position is also a potential peak position in at least scale signals, the second - condition verification of the potential peak position passes;

[0167] Take the potential peak positions that pass the first - condition verification and the second - condition verification as peaks; extract the peak information of the peaks, where the peak information is the signal value of the perturbed signal after denoising at the peak.

[0168] As an embodiment of the present invention, the frequency - modulation perturbation analysis includes: whether the peak information exceeds a preset peak threshold, the number of peaks exceeding the preset peak threshold. If there is peak information exceeding 10% of the preset peak threshold, it indicates that there is a wind - power grid - connection fault. If there is peak information exceeding 30% of the preset peak threshold and it appears frequently, it may cause gear - box fatigue failure and key components need to be replaced in advance. If there are multiple peak information exceeding the preset peak threshold in a short period of time, it can be judged that strong wind or a storm is coming, and the wind - turbine speed should be reduced in advance to reduce mechanical losses.

[0169] Embodiment 2:

[0170] This solution conducts a comparative experiment on the peak - analysis method for the primary frequency - modulation large - perturbation signal, the fixed - threshold peak - analysis method, the fixed - threshold peak - analysis method based on wavelet denoising, and the fixed - threshold peak - analysis method based on empirical mode decomposition. After the signal is denoised by wavelet denoising or empirical mode decomposition in the fixed - threshold peak - analysis method based on wavelet denoising and the fixed - threshold peak - analysis method based on empirical mode decomposition, the fixed - threshold peak - analysis method is used for peak detection. The experimental signal is the large - perturbation signal of the wind - turbine power. The comparative - experiment results are shown in Table 1:

[0171] Table 1

[0172] Experimental method Accuracy rate Recall rate F1-score Analysis method for peak value of large disturbance signal in primary frequency regulation 98.2% 97.6% 97.9% Fixed threshold peak value analysis method 85.2% 75.2% 79.9% Fixed threshold peak value analysis method based on wavelet denoising 91.3% 85.5% 88.3% Fixed threshold peak value analysis method based on empirical mode decomposition 92.8% 88.6% 90.7%

[0173] As shown in Table 1, the peak - analysis method for the primary frequency - modulation large - perturbation signal has better performance results and better noise - reduction performance for the perturbed signal. The fixed - threshold peak - analysis method is sensitive to peak drift, resulting in a low recall rate.

[0174] It should be understood that the above - mentioned embodiments are only for illustration purposes and are not limited by this structure in the scope of the patent application.

[0175] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments. And the term "including", "comprising" or any other variant thereof in this article is intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such a process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, device, article or method including such element.

[0176] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0177] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A peak analysis method for a primary frequency modulation large disturbance signal, characterized in that: The method comprises: S1: collect a large frequency-modulated disturbance signal and perform filtering processing for multi-band noise separation to obtain a disturbance signal after filtering; The primary frequency modulation large disturbance signal is a frequency change signal generated by the wind turbine generator set during the primary frequency modulation process when a power disturbance occurs in the power grid; S2: Perform multi-scale decomposition on the disturbance signal after filtering to obtain a multi-scale signal sequence of the disturbance signal, extract the dynamic entropy of the multi-scale signal sequence, perform dynamic entropy denoising on the disturbance signal, and obtain the denoised disturbance signal; S3: Calculate the potential peak position of the disturbance signal after denoising; S4: Combined with the denoised disturbance signal, multi-condition verification is performed on the potential peak position to obtain the peak information of a large frequency modulation disturbance signal. Frequency modulation disturbance analysis is performed based on the peak information, where the frequency modulation disturbance analysis includes wind power grid-connected fault identification and equipment life monitoring.

2. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 1, characterized in that: The filtering process for multi-band noise separation is performed on the primary frequency modulation large disturbance signal, including: The local trend of the primary frequency modulation large disturbance signal is calculated by a sliding window method, the primary frequency modulation large disturbance signal is detrended, and the spectrum analysis of the detrended frequency modulation large disturbance signal is performed by a Fourier transform method to obtain a frequency domain representation of the detrended frequency modulation large disturbance signal; The energy of the frequency domain representation in different frequency bands is calculated to form a spectrum energy distribution. The spectrum energy distribution is used as an input feature, and an optimal wavelet basis sequence corresponding to the input feature is generated by using a wavelet basis action selection model. The frequency modulated large disturbance signal after detrending is subjected to wavelet transform processing for multi-band noise separation to obtain a disturbance signal after filtering, wherein the optimal wavelet basis sequence is the optimal wavelet basis of the frequency modulated large disturbance signal after detrending in different frequency bands.

3. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 2, characterized in that: The wavelet-based action selection model is a reinforcement learning model, including a state space, an action space and a strategy matrix, wherein the state space includes all spectrum energy distributions, the action space includes multiple wavelet bases, the strategy matrix is ​​composed of strategy values ​​between spectrum energy distributions and wavelet basis sequences, wherein the wavelet basis sequence is a sequence composed of E wavelet bases, the wavelet-based action selection model uses spectrum energy distribution as an input feature, selects a wavelet basis sequence with the largest strategy value between the input feature and the strategy matrix as the optimal wavelet basis sequence, and the strategy matrix is ​​a model parameter to be optimized and solved; A reward function is constructed to evaluate the multi-band noise separation effect of the wavelet basis sequence selected in the strategy matrix. The variables of the reward function are the spectrum energy distribution and the wavelet basis sequence. The expression of the reward function is: ； ； ； in: represents the reward function, represents the spectral energy distribution of the input reward function, The wavelet basis sequence representing the input reward function, represents the e-th wavelet basis in the wavelet basis sequence, the e-th wavelet basis corresponds to the e-th frequency band, and the e-th frequency band is subjected to wavelet transform using the e-th wavelet basis; Represents the use of wavelet basis sequence S After the corresponding signal is transformed by wavelet, the signal-to-noise ratio of the signal is improved. Represents the use of wavelet basis sequence S The time consumption of wavelet transform of the corresponding signal, Indicates the smoothness of the wavelet basis sequence S; The hyperparameter weights represent the improvement in signal-to-noise ratio, time consumption, and degree of smoothing, respectively.

4. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 3, characterized in that: Based on the reward function, the loss function of the strategy matrix is ​​constructed: ； ； in: represents the loss function of the policy matrix, represents the ith group of spectrum energy distribution in the state space, Represents the jth group of wavelet basis sequences constructed based on the wavelet basis in the action space, represents the number of spectral energy distributions in the state space, Represents the number of constructed wavelet basis sequences; represents the strategy matrix, ; The strategy matrix is ​​solved based on the loss function, and a wavelet-based action selection model is constructed using the solved strategy matrix.

5. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 1, characterized in that: The disturbance signal after filtering is subjected to multi-scale decomposition to obtain scale signals of the disturbance signal at different scales to form a multi-scale signal sequence, including: Initialize the center frequency and scale signal spectrum of the disturbance signal at A scales, and initialize the Lagrange multiplier. Iterate the scale signal spectrum based on the center frequency and the Lagrange multiplier, perform weighted calculation on the scale signal spectrum, iteratively update the center frequency, and update the Lagrange multiplier using a dynamic step update method until the iterative change of the scale signal spectrum is lower than a preset change threshold, so as to obtain the scale signal spectrum of the disturbance signal y at A scales, wherein A represents the number of scales, perform inverse Fourier transform on the scale signal spectrum, obtain the scale signal at A scales, and form a multi-scale signal sequence.

6. A method for analyzing the peak value of a primary frequency modulation large disturbance signal as claimed in claim 5, characterized in that: Performing singular value decomposition on the multi-scale signal sequence to obtain a singular value sequence of the multi-scale signal sequence, and extracting energy entropy and singular value entropy of the multi-scale signal sequence respectively, including: The calculation formula of the energy entropy is: ； in: represents the first The scaled signal spectrum of the scaled signal of the scale, Represents the scale signal spectrum The energy ; represents the energy entropy of the multi-scale signal sequence; The calculation formula of the singular value entropy is: ； in: represents the first singular values, represents the singular value entropy.

7. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 6, characterized in that: Based on the modal entropy and the singular value entropy, the dynamic entropy of the multi-scale signal sequence is extracted, and the disturbance signal is subjected to dynamic entropy denoising to obtain the denoised disturbance signal, including: The calculation formula of the dynamic entropy is: ； in: represents dynamic entropy; The dynamic entropy is used to dynamically denoise the disturbance signal, wherein the dynamic denoising formula is: ； ； in: represents inverse Fourier transform; Indicates the preset control parameters, Represents the scale signal spectrum The corresponding transformation weights; represents the denoised disturbance signal after dynamic denoising.

8. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 7, characterized in that: Calculating the potential peak position of the denoised disturbance signal, including: The peak detection sliding window is used to perform sliding window processing on the disturbance signal after denoising, and the mean value of the signal value in the sliding window is calculated. And the signal value standard deviation std, generate the adaptive threshold of the sliding window : ； in: represents the preset adaptive control parameters; The signal value in the sliding window is higher than the adaptive threshold The signal position is taken as the potential peak position.

9. A method for analyzing a peak value of a primary frequency modulation large disturbance signal as claimed in claim 1, characterized in that: The method combines the disturbance signal after denoising and performs multi-condition verification on the potential peak position to obtain the peak information of the primary frequency modulation large disturbance signal, including: The multi-condition verification method includes: before and after the potential peak position in the disturbance signal after denoising If the signal value maintains a monotonic trend, the first condition of the potential peak position is verified; the potential peak position is at least If the scale signal is also a potential peak position, then the second condition verification of the potential peak position is passed; The potential peak position after the first condition is verified and the second condition is verified is taken as the peak; the peak information of the peak is extracted, and the peak information is the signal value of the disturbance signal at the peak after denoising.

Images