Response-driven multi-energy coupling power system disturbance event detection method and system

Abstract

The invention provides a response-driven multi-energy coupling power system disturbance event detection method and system, and the method comprises the steps: S1, obtaining PMU measurement data in a multi-energy coupling power system, and carrying out the singular value calculation of the PMU measurement data in a sliding window through a singular value decomposition method; s2, determining a threshold value of a maximum singular value by using a 3sigma principle, comparing a singular value result with the threshold value of the maximum singular value, and screening initial abnormal data; s3, performing data loss detection on the initial abnormal data by using zero value monitoring to obtain non-loss abnormal data; and S4, performing disturbance event detection based on the time-space difference of the non-loss abnormal data. According to the method, the disturbance event, the data loss and the data abnormity can be effectively distinguished from the response data, so that the occurrence of the system disturbance event can be accurately detected.

Description

Technical Field

[0001] This invention belongs to the field of power system stability and control technology, specifically relating to a response-driven method and system for detecting disturbance events in multi-energy coupled power systems. Background Technology

[0002] Artificial intelligence methods mainly use intelligent algorithms to identify abnormal features in disturbed data, thereby detecting the occurrence of disturbance events.

[0003] Currently available solutions include:

[0004] A disturbance event detection method based on kernel principal component analysis (KPCA) is proposed. This method maps PMU measurement data to a high-dimensional space by constructing a nonlinear function, thereby better characterizing the dynamic characteristics of the power system and realizing event detection.

[0005] The Local Outlier Factor (LOF) method is used to detect abnormal events in power systems.

[0006] An ensemble learning method is proposed to detect data anomalies in PMU data, which can distinguish between bad data and disturbance events.

[0007] A tool for detecting and classifying anomalous data using an unsupervised stacked ensemble learning algorithm is proposed. The proposed method uses a combination of three unsupervised methods to analyze a selected data window: Isolation Forest, KMeans, and Local Outlier Probability (LoOP) tool. This method can classify data as anomalous or abnormal.

[0008] A deep generative model based on a long short-term memory variational autoencoder is proposed, which can adapt to a moderate amount of outlier data during training. Furthermore, the detection accuracy is improved by introducing a log-cosh-based cost function. However, the artificial intelligence method requires pre-training, which poses a significant challenge to its practicality.

[0009] Existing methods for detecting disturbances in power systems are mostly based on measurement signals without any data quality issues. However, actual data may have various data quality problems during transmission, such as outliers and data loss. These data quality problems may lead to false detections of disturbances. Summary of the Invention

[0010] To address the problems existing in the prior art, the present invention provides a response-driven method for detecting disturbance events in a multi-energy coupled power system, comprising:

[0011] Step S1: Obtain PMU measurement data in a multi-energy coupled power system, and use the singular value decomposition method to calculate singular values ​​of the PMU measurement data within the sliding window;

[0012] Step S2: Use the 3sigma principle to determine the threshold of the maximum singular value, compare the singular value results with the threshold of the maximum singular value, and filter the initial outlier data;

[0013] Step S3: Use zero-value monitoring to detect data loss in the initial abnormal data to obtain non-loss abnormal data;

[0014] Step S4: Detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

[0015] Optionally, in step S1, the method for calculating singular values ​​of the PMU measurement data within the sliding window using singular value decomposition specifically includes:

[0016] The PMU measurement data is reduced in dimensionality using a sliding data window method to obtain dimensionality-reduced PMU measurement data.

[0017] The difference data is obtained by performing interpolation processing on the dimensionality-reduced PMU measurement data;

[0018] A Hankel matrix is ​​constructed based on the difference data, and the singular value decomposition method is used to decompose the Hankel matrix to obtain a singular value matrix.

[0019] Optionally, the singular value matrix is ​​specifically:

[0020]

[0021] Where r is the total number of singular values, denoted as each singular value in the singular value matrix.

[0022] Optionally, in step S2, the threshold for determining the maximum singular value using the 3sigma principle is specifically as follows:

[0023] α j,crl =μ j -3σ j ,α j,cru =μ j +3σ j

[0024] Where, μ j Let σ be the mean of the maximum singular values ​​of the j-th PMU measurement data within a sliding window of s steady-state data. j Let α be the standard deviation of the maximum singular value of the j-th PMU measurement data within a sliding window of s steady-state data. j,cruand α j,crl These are the upper and lower limits of the singular value threshold for the j-th PMU, respectively.

[0025] Optionally, in step S4, the specific content of perturbation event detection based on the spatiotemporal differences of the non-lost abnormal data is as follows:

[0026] The non-lost abnormal data is marked to obtain marked abnormal data;

[0027] The perturbation ratio is calculated based on the marked abnormal data, and perturbation event detection is achieved based on the perturbation ratio result.

[0028] Optionally, the specific content of the marked abnormal data is as follows:

[0029]

[0030] Where, β j This represents the label value of the j-th PMU data. If the j-th PMU data shows abnormal characteristics under the i-th data window, it is labeled as 1; otherwise, it is labeled as 0.

[0031] Optionally, the specific method for calculating the percentage of disturbance is as follows:

[0032]

[0033] Where K is the number of PMUs that detected anomalous features; η is the proportion of PMUs that detected anomalous features; and M is β. j The total number of PMUs marked as 0 and 1.

[0034] This invention also discloses a response-driven disturbance event detection system for a multi-energy coupled power system, comprising:

[0035] The singular value calculation module is used to acquire PMU measurement data in a multi-energy coupled power system and use the singular value decomposition method to calculate the singular values ​​of the PMU measurement data within the sliding window.

[0036] The initial anomaly filtering module is used to determine the threshold of the maximum singular value using the 3sigma principle, compare the singular value results with the threshold of the maximum singular value, and filter the initial anomaly data.

[0037] The zero-value monitoring and filtering module is used to perform data loss detection on the initial abnormal data using zero-value monitoring to obtain non-loss abnormal data.

[0038] The disturbance event detection module is used to detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] This invention provides a method for identifying system disturbance events from system response data. First, the maximum singular value of the data within a window is calculated in real time based on the SVD-3sigma method. When the calculated maximum singular value exceeds a threshold, an abnormal characteristic is detected in the data. Then, it monitors whether zero values ​​appear in the data within the window to determine if the abnormal characteristic is caused by data loss. If no zero values ​​appear, the abnormal characteristic is determined to be caused by a disturbance event or anomalies. Next, the number of PMUs exhibiting abnormal characteristics in the current data window is calculated. If the number exceeds 50%, a disturbance event is determined to have occurred in the system. This method can effectively distinguish between disturbance events, data loss, and data anomalies from the response data, thereby accurately detecting the occurrence of system disturbance events. Attached Figure Description

[0041] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 Flowchart of the disturbance event detection method according to an embodiment of the present invention;

[0043] Figure 2 Wiring diagram of an IEEE 10-machine 39-node system according to an embodiment of the present invention;

[0044] Figure 3 Frequency response data of an IEEE 10-machine, 39-node system according to an embodiment of the present invention;

[0045] Figure 4 Maximum singularity curve of frequency response data of IEEE 10-machine 39-node system in this embodiment of the invention;

[0046] Figure 5 Measured system frequency response data from embodiments of the present invention;

[0047] Figure 6 Maximum singular value curve of measured frequency response data in an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] Example 1

[0051] like Figure 1 As shown, a response-driven method for detecting disturbance events in a multi-energy coupled power system includes:

[0052] Step S1: Obtain PMU measurement data in a multi-energy coupled power system, and use the singular value decomposition method to calculate singular values ​​of the PMU measurement data within the sliding window.

[0053] This embodiment employs a sliding data window method for real-time computation, which enables dimensionality reduction of long-term data series. Assuming the i-th sliding data window has N sampling points and a total of M PMU measurement data, and using the sliding data window for rolling computation, the j-th PMU measurement data under the i-th window can be represented as:

[0054]

[0055] In the formula: It is a vector composed of measurement data from the j-th PMU under the i-th window.

[0056] Perform difference processing on the data in equation (1):

[0057]

[0058] In the formula: It is a vector composed of the difference data in the j-th PMU under the i-th window.

[0059] Based on the difference data obtained from equation (2), the Hankel matrix is ​​constructed, and the Hankel matrix is ​​decomposed by SVD:

[0060]

[0061] In the formula: The Hankel matrix is ​​composed of the difference data in the j-th PMU under the i-th window; and Let X be the singular value matrix, left singular vector, and right singular vector of the Hankel matrix formed by the j-th PMU under the i-th window.

[0062] Among them, singular value matrix This can be further expressed as:

[0063]

[0064] In the formula: r represents the total number of singular values, Let each singular value in the singular value matrix represent a singular value, and let them have a relation.

[0065] Step S2: Use the 3sigma principle to determine the threshold of the maximum singular value, compare the singular value results with the threshold of the maximum singular value, and filter the initial abnormal data.

[0066] In singular value decomposition (SVD), most of the energy in the data is concentrated in the larger singular values. This means that larger singular values ​​contain most of the information in the data, while smaller singular values ​​reflect more noise. Therefore, to extract the information that best represents the anomalous features in the data, this section selects the largest singular value. This is used as the output of a sliding data window. Upper and lower singular value thresholds are set to detect data anomalies.

[0067]

[0068] In the formula, α j,cru and α j,crl These are the upper and lower limits of the singular value threshold for the j-th PMU, respectively.

[0069] After calculating the real-time singular values ​​of different PMU data using sliding window computation, the key is to determine the singular value triggering threshold α. j,cru and α j,crl To determine whether there are any abnormalities in the data, the 3sigma principle is used to determine the outlier threshold.

[0070] Let the mean of a set of test data be μ and the standard deviation be σ. Then the 3 sigma principle applies: the probability that the data falls within the interval (μ-σ, μ+σ) is 0.6826; the probability that the data falls within the interval (μ-2σ, μ+2σ) is 0.9545; and the probability that the data falls within the interval (μ-3σ, μ+3σ) is 0.9973. It can be assumed that almost all data values ​​are concentrated within the interval (μ-3σ, μ+3σ), with the probability of falling outside this range being less than 0.3%. Therefore, it can be considered that any data outside the interval (μ-3σ, μ+3σ) exhibits data anomalies.

[0071] Based on the above 3sigma principle, the upper and lower limits of the singular value threshold of the j-th PMU can be expressed as:

[0072] α j,crl =μ j -3σ j ,α j,cru =μ j +3σ j (6)

[0073] Where: μ j σ represents the mean of the maximum singular values ​​of the j-th PMU under a sliding window of s steady-state data; j Let represent the standard deviation of the maximum singular value of the j-th PMU under the s-th steady-state data sliding window.

[0074] in,

[0075]

[0076] In the formula: s represents the number of sliding windows in the steady-state data.

[0077] For interconnected power grids, the transient response process of frequency data in different regions exhibits spatiotemporal distribution characteristics, but the steady-state frequency changes show strong consistency. Therefore, the singular value calculation results of steady-state measurement data from different PMUs are slightly different but very close. Thus, the singular value thresholds calculated from multiple PMU data can be integrated to form a unified singular value threshold for data anomaly detection. To prevent missed detections, this embodiment selects the smallest singular value threshold calculated by different PMUs as the overall threshold.

[0078] α crl =min(α) j,crl ),α cru =min(α) j,cru (8)

[0079] In the formula: α crl and α cru These are the upper and lower limits of the system singularity thresholds under the detection data, respectively.

[0080] Step S3: Use zero-value monitoring to detect data loss in the initial abnormal data to obtain non-loss abnormal data.

[0081] The purpose of this step is to determine whether the abnormal characteristics of the tested data are caused by data loss. In data centers, data loss is usually manifested as a zero value. Therefore, zero-value monitoring is performed on the data within the window to observe whether there is any data with zero values. If so, it is determined that data loss has occurred; if not, no data loss has occurred.

[0082]

[0083] Step S4: Detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

[0084] For different PMU measurement data, an outlier in one PMU at a certain time does not necessarily mean that other PMU data will also show outliers at that time. This is the spatiotemporal variability between PMU measurement data, meaning the probability of the same measurement data from different PMUs simultaneously showing outliers is very low. However, for disturbance events, almost all PMU measurement data in the system will simultaneously exhibit anomalous characteristics. Utilizing this characteristic, based on the previous step, the detected different PMU data are labeled as follows:

[0085]

[0086] Where: β j This represents the label value of the j-th PMU data. If the j-th PMU data shows abnormal characteristics under the i-th data window, it is labeled as 1; otherwise, it is labeled as 0.

[0087] Furthermore, within the same sliding data window, the proportion of PMU data exhibiting anomalous characteristics is used to determine whether the anomalous characteristics are caused by disturbance events or outliers:

[0088]

[0089] In the formula: K is the number of PMUs that detected abnormal features; η is the proportion of PMUs that detected abnormal features; and M is β. j The total number of PMUs marked as 0 and 1. When this proportion is greater than 50%, it is considered that most PMUs in the system exhibit abnormal characteristics within this data window, indicating a disturbance event in the system. When this proportion is less than 50%, it is considered that a minority of PMUs in the system exhibit abnormal characteristics within this data window, indicating that the abnormal characteristics are caused by outliers.

[0090] Example 2

[0091] The present invention also provides a response-driven multi-energy coupled power system disturbance event detection system for implementing the method of embodiment one, comprising:

[0092] The singular value calculation module is used to acquire PMU measurement data in a multi-energy coupled power system and to perform singular value calculation on the PMU measurement data within a sliding window using the singular value decomposition method.

[0093] This embodiment employs a sliding data window method for real-time computation, which enables dimensionality reduction processing of long-term data series in real time. Assuming the i-th sliding data window has N sampling points and a total of M PMU measurement data, and using the sliding data window for rolling computation, the j-th PMU measurement data under the i-th window can be represented as:

[0094]

[0095] In the formula: It is a vector composed of measurement data from the j-th PMU under the i-th window.

[0096] Perform difference processing on the data in equation (12):

[0097]

[0098] In the formula: It is a vector composed of the difference data in the j-th PMU under the i-th window.

[0099] Based on the difference data obtained from equation (13), the Hankel matrix is ​​constructed, and the Hankel matrix is ​​decomposed by SVD:

[0100]

[0101] In the formula: The Hankel matrix is ​​composed of the difference data in the j-th PMU under the i-th window; and Let X be the singular value matrix, left singular vector, and right singular vector of the Hankel matrix formed by the j-th PMU under the i-th window.

[0102] Among them, singular value matrix This can be further expressed as:

[0103]

[0104] In the formula: r represents the total number of singular values, Let each singular value in the singular value matrix represent a singular value, and let them have a relation.

[0105] The initial anomaly filtering module is used to determine the threshold of the maximum singular value using the 3sigma principle, compare the singular value results with the threshold of the maximum singular value, and filter the initial anomaly data.

[0106] In singular value decomposition (SVD), most of the energy in the data is concentrated in the larger singular values. This means that larger singular values ​​contain most of the information in the data, while smaller singular values ​​reflect more noise. Therefore, to extract the information that best represents the anomalous features in the data, this section selects the largest singular value. This is used as the output of a sliding data window. Upper and lower singular value thresholds are set to detect data anomalies.

[0107]

[0108] In the formula, α j,cru and αj,crl These are the upper and lower limits of the singular value threshold for the j-th PMU, respectively.

[0109] After calculating the real-time singular values ​​of different PMU data using sliding window computation, the key is to determine the singular value triggering threshold α. j,cru and α j,crl To determine whether there are any abnormalities in the data, the 3sigma principle is used to determine the outlier threshold.

[0110] Let the mean of a set of test data be μ and the standard deviation be σ. Then the 3 sigma principle applies: the probability that the data falls within the interval (μ-σ, μ+σ) is 0.6826; the probability that the data falls within the interval (μ-2σ, μ+2σ) is 0.9545; and the probability that the data falls within the interval (μ-3σ, μ+3σ) is 0.9973. It can be assumed that almost all data values ​​are concentrated within the interval (μ-3σ, μ+3σ), with the probability of falling outside this range being less than 0.3%. Therefore, it can be considered that any data outside the interval (μ-3σ, μ+3σ) exhibits data anomalies.

[0111] Based on the above 3sigma principle, the upper and lower limits of the singular value threshold of the j-th PMU can be expressed as:

[0112] α j,crl =μ j -3σ j ,α j,cru =μ j +3σ j (17)

[0113] Where: μ j σ represents the mean of the maximum singular values ​​of the j-th PMU under a sliding window of s steady-state data; j Let represent the standard deviation of the maximum singular value of the j-th PMU under the s-th steady-state data sliding window.

[0114] in,

[0115]

[0116] In the formula: s represents the number of sliding windows in the steady-state data.

[0117] For interconnected power grids, the transient response process of frequency data in different regions exhibits spatiotemporal distribution characteristics, but the steady-state frequency changes show strong consistency. Therefore, the singular value calculation results of steady-state measurement data from different PMUs are slightly different but very close. Thus, the singular value thresholds calculated from multiple PMU data can be integrated to form a unified singular value threshold for data anomaly detection. To prevent missed detections, this embodiment selects the smallest singular value threshold calculated by different PMUs as the overall threshold.

[0118] α crl =min(α) j,crl ),α cru =min(α) j,cru (19)

[0119] In the formula: α crl and α cru These are the upper and lower limits of the system singularity thresholds under the detection data, respectively.

[0120] The zero-value monitoring and filtering module is used to perform data loss detection on the initial abnormal data using zero-value monitoring to obtain non-loss abnormal data.

[0121] The purpose of this step is to determine whether the abnormal characteristics of the tested data are caused by data loss. In data centers, data loss is usually manifested as a zero value. Therefore, zero-value monitoring is performed on the data within the window to observe whether there is any data with zero values. If so, it is determined that data loss has occurred; if not, no data loss has occurred.

[0122]

[0123] The disturbance event detection module is used to detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

[0124] For different PMU measurement data, an outlier in one PMU at a certain time does not necessarily mean that other PMU data will also show outliers at that time. This is the spatiotemporal variability between PMU measurement data, meaning the probability of the same measurement data from different PMUs simultaneously showing outliers is very low. However, for disturbance events, almost all PMU measurement data in the system will simultaneously exhibit anomalous characteristics. Utilizing this characteristic, based on the previous step, the detected different PMU data are labeled as follows:

[0125]

[0126] Where: β j This represents the label value of the j-th PMU data. If the j-th PMU data shows abnormal characteristics under the i-th data window, it is labeled as 1; otherwise, it is labeled as 0.

[0127] Furthermore, within the same sliding data window, the proportion of PMU data exhibiting anomalous characteristics is used to determine whether the anomalous characteristics are caused by disturbance events or outliers:

[0128]

[0129] In the formula: K is the number of PMUs that detected abnormal features; η is the proportion of PMUs that detected abnormal features. When this proportion is greater than 50%, it is considered that most PMUs in the system have abnormal features in this data window, and it can be determined that the system has a disturbance event. When this proportion is less than 50%, it is considered that a minority of PMUs in the system have abnormal features in this data window, and it can be considered that the abnormal features in the data are caused by outliers.

[0130] Example 3

[0131] This embodiment verifies the accuracy of the proposed disturbance event detection method using response data from an IEEE 10-machine 39-node system and a measured system.

[0132] The specific model of the IEEE 10-machine 39-node system is as follows: Figure 2 As shown. A three-phase fault was simulated on bus 17 at t=5s. The frequency was extracted as the data to be measured, and some outliers were randomly added. The PMU frequency data under the three-phase fault condition is shown below. Figure 3 As shown, and combining the maximum singular value results of the sliding data window with the singular value threshold calculation results based on the 3sigma principle, as shown... Figure 4 As shown, when a disturbance occurs in the system or an outlier appears in the data, the maximum singular value exhibits a significant spike. When the system is in steady state, when an outlier or disturbance suddenly appears within the data window, the maximum singular value always exceeds the threshold. All outliers and disturbances can be accurately detected, and within the data window where the disturbance occurs, the proportion of PMUs with singular values ​​exceeding the threshold reaches 100%, enabling accurate identification of disturbances. However, due to... Figure 4 It can also be observed that during the transient changes following a disturbance event, the maximum singular value fluctuates significantly, potentially exceeding or remaining within the threshold. Therefore, the calculated maximum singular value during the transient process offers no guidance for identifying data anomalies. Furthermore, the system's operating point may change after a disturbance event, altering the singular value threshold used for steady-state calculations. Consequently, the singular value threshold needs to be recalculated based on sliding a certain number of data windows across the steady-state data after the disturbance event.

[0133] Example 4

[0134] The accuracy of the proposed disturbance event detection method was verified using measured system response data. Measured frequency data are as follows: Figure 5 As shown, outliers appear in the frequency curve at t=20s and t=57s, and a disturbance event occurs in the system at t=176s. The maximum value of the frequency singularity calculated by the method of this invention is as follows: Figure 6As shown, when an outlier or disturbance suddenly appears in the data window, the maximum singular value exceeds the threshold. All outliers and disturbances can be accurately detected. In the same data window where the disturbance occurs, the proportion of PMUs with singular values ​​exceeding the threshold reaches 100%, which can accurately identify the disturbance.

[0135] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A response-driven method for detecting disturbance events in a multi-energy coupled power system, characterized in that, include: Step S1: Obtain PMU measurement data in a multi-energy coupled power system, and use the singular value decomposition method to calculate singular values ​​of the PMU measurement data within the sliding window; Step S2: Use the 3sigma principle to determine the threshold of the maximum singular value, compare the singular value results with the threshold of the maximum singular value, and filter the initial outlier data; Step S3: Use zero-value monitoring to detect data loss in the initial abnormal data to obtain non-loss abnormal data; Step S4: Detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

2. The method according to claim 1, characterized in that, In step S1, the method for calculating singular values ​​of the PMU measurement data within the sliding window using singular value decomposition specifically includes: The PMU measurement data is reduced in dimensionality using a sliding data window method to obtain dimensionality-reduced PMU measurement data. The difference data is obtained by performing interpolation processing on the dimensionality-reduced PMU measurement data; A Hankel matrix is ​​constructed based on the difference data, and the singular value decomposition method is used to decompose the Hankel matrix to obtain a singular value matrix.

3. The method according to claim 2, characterized in that, The singular value matrix is ​​specifically: Where r is the total number of singular values, denoted as each singular value in the singular value matrix.

4. The method according to claim 1, characterized in that, In step S2, the threshold for determining the maximum singular value using the 3sigma principle is specifically as follows: a j,crl =μ j -3s j ,a j,cru =μ j +3s j Where, μ j Let σ be the mean of the maximum singular values ​​of the j-th PMU measurement data within a sliding window of s steady-state data. j Let αj,cru and αj,crl be the standard deviation of the maximum singular value of the j-th PMU measurement data under the sliding window of s steady-state data, and let αj,crl be the upper and lower limits of the singular value threshold of the j-th PMU measurement data, respectively.

5. The method according to claim 1, characterized in that, In step S4, the specific content of perturbation event detection based on the spatiotemporal differences of the non-lost abnormal data is as follows: The non-lost abnormal data is marked to obtain marked abnormal data; The perturbation ratio is calculated based on the marked abnormal data, and perturbation event detection is achieved based on the perturbation ratio result.

6. The method according to claim 5, characterized in that, The specific content of the marked abnormal data is as follows: Where, β j This represents the label value of the j-th PMU data point. If an anomaly is observed in the j-th PMU data point within the i-th data window, it is labeled as 1; otherwise, it is labeled as 0. α crl and α cru These are the upper and lower limits of the singularity threshold for PMU measurement data, respectively.

7. The method according to claim 1, characterized in that, The specific method for calculating the proportion of disturbances is as follows: Where K is the number of PMUs that detected anomalous features; η is the proportion of PMUs that detected anomalous features; and M is β. j The total number of PMUs marked as 0 and 1.

8. A response-driven disturbance event detection system for a multi-energy coupled power system, used to implement the method described in any one of claims 1-7, characterized in that, include: The singular value calculation module is used to acquire PMU measurement data in a multi-energy coupled power system and use the singular value decomposition method to calculate the singular values ​​of the PMU measurement data within the sliding window. The initial anomaly filtering module is used to determine the threshold of the maximum singular value using the 3sigma principle, compare the singular value results with the threshold of the maximum singular value, and filter the initial anomaly data. The zero-value monitoring and filtering module is used to perform data loss detection on the initial abnormal data using zero-value monitoring to obtain non-loss abnormal data. The disturbance event detection module is used to detect disturbance events based on the spatiotemporal differences of the non-lost abnormal data.

Images