Generator set active power low frequency oscillation on-line identification method and system

Abstract

The invention provides a generator set active power low-frequency oscillation online identification method and system, and relates to the field of operation safety of generator sets, and the method comprises the steps: obtaining active power data of a generator set, and carrying out the preprocessing of the active power data; setting a sliding window, scanning the preprocessed active power data through the sliding window, performing Prony decomposition on the data in the window, and obtaining a plurality of cosine components in each window; judging whether the window is a low-frequency oscillation suspected window or not according to a preset frequency threshold value, and calculating the maximum amplitude in all cosine components of the low-frequency oscillation suspected window; according to a preset continuity threshold value and an amplitude threshold value, the maximum amplitude mean value of the continuous low-frequency oscillation suspected windows is screened, and the screened continuous windows are low-frequency oscillation abnormal signals. On the basis of an existing Prony algorithm, a judgment rule of continuous window smoothness is introduced, the anti-noise performance of the algorithm is further optimized, and the accuracy and stability of low-frequency oscillation recognition are improved.

Description

Technical Field

[0001] The invention belongs to the field of operating safety of a generator set, and in particular relates to an online identification method and system for low-frequency oscillation of active power of a generator set. Background Art

[0002] The low-frequency oscillation of the active power of the generator set is an important stability problem in the power system. It usually occurs when the system frequency is close to its natural frequency and is underdamped, causing great damage to the equipment. The occurrence of low-frequency oscillation is mainly attributed to inappropriate control methods: when the system is disturbed, the electromagnetic torque and the terminal voltage have opposite requirements for regulating the excitation current, and it is difficult for the excitation regulation to meet both requirements at the same time. If the control requirement is to suppress low-frequency oscillations in the system, but a control method that has no direct relationship with the rotor speed is used, low-frequency oscillation may result.

[0003] In the analysis of low-frequency oscillation of generator sets, the Prony algorithm is a commonly used "global" analysis tool; this algorithm fits equally spaced sampling data through a linear combination of a set of exponential functions, thereby extracting key parameters such as the amplitude, frequency, phase and damping ratio of the low-frequency oscillation signal; however, the Prony algorithm is extremely sensitive to noise, and the interference of noise will significantly affect the accuracy of parameter identification, especially the impact on the damping ratio, which will lead to misjudgment of oscillation risk.

[0004] Therefore, the existing low-frequency oscillation recognition scheme based on the Prony algorithm is sensitive to noise and interfered by noise, so the recognition accuracy is unstable and the robustness is not high. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for online identification of low-frequency oscillation of active power of a generator set. On the basis of the existing Prony algorithm, a judgment rule of continuous window smoothing is introduced, and parameters such as frequency threshold, amplitude threshold, and continuity threshold are set at the same time to further optimize the anti-noise performance of the algorithm, so as to improve the accuracy and stability of identifying low-frequency oscillation of active power of the generator set.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention provides an online identification method for low-frequency oscillation of active power of a generator set.

[0008] A method for online identification of low-frequency oscillation of active power of a generator set, comprising:

[0009] Obtain active power data of the generator set and pre-process it;

[0010] Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window;

[0011] According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated;

[0012] According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0013] Furthermore, the preprocessing includes data cleaning and normalization, and the data cleaning is to eliminate outliers and noise interference.

[0014] Furthermore, the setting of the sliding window is to intercept the active power data by setting the sliding window size and step size.

[0015] Furthermore, the Prony decomposition is performed on the data in the window, which is expressed as follows:

[0016]

[0017] Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, f k is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

[0018] Furthermore, the determination of whether the window is a suspected low-frequency oscillation window is specifically as follows:

[0019] If there is a cosine component with a frequency greater than the frequency threshold among the d cosine components obtained by window decomposition, the window is marked as a suspected low-frequency oscillation window.

[0020] Furthermore, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, specifically:

[0021] Determine the number S of continuous low-frequency oscillation suspected windows;

[0022] If S is greater than the continuity threshold, then take the average of all maximum amplitudes in the S windows;

[0023] If the average of the maximum amplitude is greater than the amplitude threshold, these S windows are considered to be low-frequency oscillation abnormal signals.

[0024] Furthermore, it also includes outputting the identified low-frequency oscillation abnormal signal, including the start time, end time, maximum amplitude and frequency of the oscillation.

[0025] A second aspect of the present invention provides an online identification system for low-frequency oscillation of active power of a generator set.

[0026] An online identification system for low-frequency oscillation of active power of a generator set includes a data acquisition module, a data decomposition module and a frequency judgment module:

[0027] The data acquisition module is configured to: acquire active power data of the generator set and pre-process it;

[0028] The data decomposition module is configured to: set a sliding window, scan the pre-processed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain a number of cosine components in each window;

[0029] The frequency judgment module is configured to: judge whether the window is a low-frequency oscillation suspected window according to a preset frequency threshold, and calculate the maximum amplitude of all cosine components of the low-frequency oscillation suspected window;

[0030] The abnormality identification module is configured to screen the maximum amplitude mean of the continuous low-frequency oscillation suspected window according to the preset continuity threshold and amplitude threshold, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0031] Furthermore, the preprocessing includes data cleaning and normalization, and the data cleaning is to eliminate outliers and noise interference.

[0032] Furthermore, the setting of the sliding window is to intercept the active power data by setting the sliding window size and step size.

[0033] Furthermore, the Prony decomposition is performed on the data in the window, which is expressed as follows:

[0034]

[0035] Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, f k is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

[0036] Furthermore, the determination of whether the window is a suspected low-frequency oscillation window is specifically as follows:

[0037] If there is a cosine component with a frequency greater than the frequency threshold among the d cosine components obtained by window decomposition, the window is marked as a suspected low-frequency oscillation window.

[0038] Furthermore, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, specifically:

[0039] Determine the number S of continuous low-frequency oscillation suspected windows;

[0040] If S is greater than the continuity threshold, then take the average of all maximum amplitudes in the S windows;

[0041] If the average of the maximum amplitude is greater than the amplitude threshold, these S windows are considered to be low-frequency oscillation abnormal signals.

[0042] Furthermore, it also includes outputting the identified low-frequency oscillation abnormal signal, including the start time, end time, maximum amplitude and frequency of the oscillation.

[0043] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the following steps when executed by a processor:

[0044] Obtain active power data of the generator set and pre-process it;

[0045] Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window;

[0046] According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated;

[0047] According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0048] A fourth aspect of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the program:

[0049] Obtain active power data of the generator set and pre-process it;

[0050] Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window;

[0051] According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated;

[0052] According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0053] One or more of the above technical solutions have the following beneficial effects:

[0054] Based on the existing Prony algorithm, the present invention introduces the judgment rule of continuous window smoothing, and sets parameters such as frequency threshold, amplitude threshold, continuity threshold, etc., to further optimize the anti-noise performance of the algorithm, so as to improve the accuracy and stability of the identification of low-frequency oscillation of the active power of the generator set, including the following two points:

[0055] The present invention performs segmented Prony algorithm on data through a sliding window, thereby preventing interference of noise accumulation and amplification on fitting results, weakening the influence of noise interference on Prony fitting, and improving the accuracy of parameter identification.

[0056] The present invention compensates for the influence of segmentation Prony on data integrity by setting parameters such as frequency threshold, amplitude threshold, continuity threshold, etc. while performing sliding window segmentation, thereby improving the accuracy and reliability of online identification of low-frequency oscillation of active power of generator sets and improving robustness.

[0057] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0059] Figure 1 This is a flow chart of the method of the first embodiment. DETAILED DESCRIPTION

[0060] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0062] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0063] First, let's briefly explain Prony oscillation:

[0064] Prony oscillation refers to low-frequency oscillation in power systems, which is usually analyzed and processed by the Prony algorithm. Low-frequency oscillation is the main manifestation of stability problems in modern power systems. It is mainly manifested as oscillations with a frequency lower than 1 Hz after a small disturbance, which may cause system collapse in severe cases12.

[0065] The basic principle and application of Prony algorithm:

[0066] The Prony algorithm is a mathematical method for signal processing. It fits equally spaced sampled data through an attenuated linear combination of exponential functions, and can effectively obtain information such as the signal's frequency, attenuation factor, amplitude, and phase. This method assumes that the signal is composed of a series of exponential functions with arbitrary amplitude, phase, frequency, and attenuation factor, and is suitable for low-frequency oscillation analysis in power systems.

[0067] Application effects and challenges of Prony algorithm in power system:

[0068] The Prony algorithm performs well in the analysis of low-frequency oscillations in power systems and can effectively identify the characteristic parameters of the system; however, this method is sensitive to noise, especially when the electromechanical oscillation trajectory is highly nonlinear or there is noise, the results have large deviations and are difficult to meet the requirements of engineering applications.

[0069] In summary, the Prony algorithm has important application value in the analysis of low-frequency oscillations in power systems, but its sensitivity to noise limits its application in practical engineering. It is necessary to improve its robustness and accuracy by improving the algorithm.

[0070] Embodiment 1

[0071] In one embodiment of the present disclosure, a method for online identification of low-frequency oscillation of active power of a generator set is provided, comprising the following steps:

[0072] Step S1: Obtain active power data of the generator set and preprocess it.

[0073] Furthermore, the preprocessing includes data cleaning and normalization, and the data cleaning is to eliminate outliers and noise interference.

[0074] Step S2: Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain a number of cosine components in each window.

[0075] Furthermore, the setting of the sliding window is to intercept the active power data by setting the sliding window size and step size.

[0076] Furthermore, the Prony decomposition is performed on the data in the window, which is expressed as follows:

[0077]

[0078] Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, f k is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

[0079] Step S3: judging whether the window is a low-frequency oscillation suspected window according to a preset frequency threshold, and calculating the maximum amplitude of all cosine components of the low-frequency oscillation suspected window.

[0080] Furthermore, the determination of whether the window is a suspected low-frequency oscillation window is specifically as follows:

[0081] If there is a cosine component with a frequency greater than the frequency threshold among the d cosine components obtained by window decomposition, the window is marked as a suspected low-frequency oscillation window.

[0082] Step S4: According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0083] Furthermore, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, specifically:

[0084] Determine the number S of continuous low-frequency oscillation suspected windows;

[0085] If S is greater than the continuity threshold, then take the average of all maximum amplitudes in the S windows;

[0086] If the average of the maximum amplitude is greater than the amplitude threshold, these S windows are considered to be low-frequency oscillation abnormal signals.

[0087] Furthermore, it also includes outputting the identified low-frequency oscillation abnormal signal, including the start time, end time, maximum amplitude and frequency of the oscillation.

[0088] The following is a detailed description of the implementation process of a method for online identification of low-frequency oscillation of active power of a generator set according to this embodiment.

[0089] In order to improve the reliability of generator set operation, based on the existing Prony algorithm, the judgment rule of continuous window smoothing is introduced, and parameters such as frequency threshold, amplitude threshold, continuity threshold, etc. are set at the same time, so as to further optimize the anti-noise performance of the algorithm and improve the accuracy and stability of low-frequency oscillation identification.

[0090] To provide a more specific example, Figure 1 As shown, the specific steps are:

[0091] 1. Collect the active power data of the generator set and preprocess it, including data cleaning and normalization, which is used to clean the original data, remove outliers and noise interference, and normalize the data to ensure that the data is in a unified dimension and scale to facilitate subsequent algorithm processing.

[0092] Specifically, data cleaning can identify and eliminate outliers in the data set through statistical analysis methods, while normalization can scale the data to the range of [0,1] through mathematical transformation. The formula is expressed as:

[0093]

[0094] Among them, x is the original data of active power, x norm is the normalized data, max(x) and min(x) are the maximum and minimum values ​​of the original data respectively.

[0095] 2. Set a sliding window, scan the preprocessed active power data through the sliding window, and obtain m sliding windows.

[0096] Specifically, let the sliding window size be S, the step size be L, and the nth sliding window be W j (j=0,1,…,m), the total number of sliding windows m is:

[0097]

[0098] Where n is the total length of the preprocessed active power data, and m is the total number of sliding windows.

[0099] 3. Use the Prony algorithm to decompose the active power data in each sliding window to obtain d cosine components.

[0100] The Prony algorithm is a mathematical model that uses complex exponential decay linear combinations to fit equally spaced sampling data. After appropriate expansion, it can directly estimate the frequency, attenuation factor, amplitude and phase of a given signal, and can more comprehensively express the characteristics of transient signals than traditional Fourier analysis methods.

[0101] In Fourier analysis, it is assumed that the signal consists of a series of unattenuated sinusoidal frequency components, while the Prony algorithm assumes that the model is a combination of a series of exponential functions with arbitrary amplitude, phase, frequency and attenuation factor, that is, it consists of some attenuated sine / cosine signal components.

[0102] Specifically, the Prony algorithm is used to decompose the data in the window. i The active power data in is represented as X = {x1, x2, ..., x n}, Prony algorithm is expressed as:

[0103]

[0104] Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, f k is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

[0105] The above formula is fitted by the least square method to determine the frequency, amplitude, damping ratio and phase of the cosine component.

[0106] 4. Filter the cosine component according to the frequency threshold f to determine the suspected low-frequency oscillation window. The specific judgment rules are as follows:

[0107] If the jth window W j There is a frequency f in the d cosine components of the Prony decomposition k >f, the window is marked as a potential low-frequency oscillation suspected window, and the window W is recorded at the same time. j The maximum amplitude of all cosine components of

[0108] 5. Further screen the low-frequency oscillation data through continuity threshold and amplitude threshold to ensure the accuracy of the recognition results.

[0109] Specifically, for the window W that continuously satisfies condition 4j Count, recorded as S. If S is greater than the continuity threshold l * , then for all S windows Take the mean value and record it as if Also meets the amplitude threshold The continuous window is considered to be a low-frequency oscillation abnormal signal.

[0110] 6. Output the filtered low-frequency oscillation data, including the start time, end time, maximum amplitude and frequency of the oscillation.

[0111] Embodiment 2

[0112] In one embodiment of the present disclosure, a system for online identification of low-frequency oscillation of active power of a generator set is provided, including a data acquisition module, a data decomposition module and a frequency judgment module:

[0113] The data acquisition module is configured to: acquire active power data of the generator set and pre-process it;

[0114] The data decomposition module is configured to: set a sliding window, scan the pre-processed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain a number of cosine components in each window;

[0115] The frequency judgment module is configured to: judge whether the window is a low-frequency oscillation suspected window according to a preset frequency threshold, and calculate the maximum amplitude of all cosine components of the low-frequency oscillation suspected window;

[0116] The abnormality identification module is configured to screen the maximum amplitude mean of the continuous low-frequency oscillation suspected window according to the preset continuity threshold and amplitude threshold, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0117] Furthermore, the preprocessing includes data cleaning and normalization, and the data cleaning is to eliminate outliers and noise interference.

[0118] Furthermore, the setting of the sliding window is to intercept the active power data by setting the sliding window size and step size.

[0119] Furthermore, the Prony decomposition is performed on the data in the window, which is expressed as follows:

[0120]

[0121] Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, fk is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

[0122] Furthermore, the determination of whether the window is a suspected low-frequency oscillation window is specifically as follows:

[0123] If there is a cosine component with a frequency greater than the frequency threshold among the d cosine components obtained by window decomposition, the window is marked as a low-frequency oscillation suspected window.

[0124] Furthermore, the maximum amplitude mean of the suspected continuous low-frequency oscillation window is screened, specifically:

[0125] Determine the number S of continuous low-frequency oscillation suspected windows;

[0126] If S is greater than the continuity threshold, then take the average of all maximum amplitudes in the S windows;

[0127] If the average of the maximum amplitude is greater than the amplitude threshold, these S windows are considered to be low-frequency oscillation abnormal signals.

[0128] Furthermore, it also includes outputting the identified low-frequency oscillation abnormal signal, including the start time, end time, maximum amplitude and frequency of the oscillation.

[0129] Embodiment 3

[0130] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the following steps are implemented:

[0131] Obtain active power data of the generator set and pre-process it;

[0132] Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window;

[0133] According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated;

[0134] According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0135] It should be noted here that the various steps executed in the processor are the same as the specific implementation process of each module in Example 1, and will not be repeated here.

[0136] Embodiment 4

[0137] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the program:

[0138] Obtain active power data of the generator set and pre-process it;

[0139] Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window;

[0140] According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated;

[0141] According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

[0142] It should be noted here that the various steps executed in the processor are the same as the specific implementation process of each module in Example 1, and will not be repeated here.

[0143] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0144] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for online identification of low-frequency oscillation of active power of a generator set, characterized in that: include: Obtain active power data of the generator set and pre-process it; Set a sliding window, scan the preprocessed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain several cosine components in each window; According to the preset frequency threshold, whether the window is a low-frequency oscillation suspected window is judged, and the maximum amplitude of all cosine components of the low-frequency oscillation suspected window is calculated; According to the preset continuity threshold and amplitude threshold, the maximum amplitude mean of the continuous low-frequency oscillation suspected window is screened, and the screened continuous window is the low-frequency oscillation abnormal signal.

2. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: The preprocessing includes data cleaning and normalization, and the data cleaning is to eliminate outliers and noise interference.

3. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: The setting of the sliding window is to intercept the active power data by setting the sliding window size and step size.

4. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: The Prony decomposition is performed on the data in the window, which is expressed by the formula: Among them, x i is the i-th data in the window, A k and φ k are the amplitude and phase of the kth cosine component, f k is the frequency of the kth cosine component, τ is the phase shift, e i is the error term, and d represents the total number of cosine components of the signal decomposition.

5. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: The determination of whether the window is a suspected low-frequency oscillation window is specifically as follows: If there is a cosine component with a frequency greater than the frequency threshold among the d cosine components obtained by window decomposition, the window is marked as a suspected low-frequency oscillation window.

6. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: The screening of the maximum amplitude mean of the suspected continuous low-frequency oscillation window is specifically as follows: Determine the number S of continuous low-frequency oscillation suspected windows; If S is greater than the continuity threshold, then take the average of all maximum amplitudes in the S windows; If the average of the maximum amplitude is greater than the amplitude threshold, these S windows are considered to be low-frequency oscillation abnormal signals.

7. A method for online identification of low-frequency oscillation of active power of a generator set as claimed in claim 1, characterized in that: It also includes outputting the identified low-frequency oscillation abnormal signal, including the start time, end time, maximum amplitude and frequency of the oscillation.

8. An online identification system for low-frequency oscillation of active power of a generator set, characterized in that: Including data acquisition, data decomposition module and frequency judgment module: The data acquisition module is configured to: acquire active power data of the generator set and pre-process it; The data decomposition module is configured to: set a sliding window, scan the pre-processed active power data through the sliding window, perform Prony decomposition on the data in the window, and obtain a number of cosine components in each window; The frequency judgment module is configured to: judge whether the window is a low-frequency oscillation suspected window according to a preset frequency threshold, and calculate the maximum amplitude of all cosine components of the low-frequency oscillation suspected window; The abnormality identification module is configured to screen the maximum amplitude mean of the continuous low-frequency oscillation suspected window according to the preset continuity threshold and amplitude threshold, and the screened continuous window is the low-frequency oscillation abnormal signal.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; a processor for executing the computer-readable instructions; Wherein, when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitorily stored, wherein when the computer-readable instructions are executed by a computer, the method of any one of claims 1 to 7 is performed.

Images