Method and system for mining abnormal data in power grid measurement data

Abstract

The invention discloses a method and system for mining abnormal data in power grid measurement data, relates to the field of data processing, and solves the problem that the mining accuracy of the abnormal data in the existing power grid measurement data is insufficient, and the method comprises the following steps: S1, obtaining a standard data set and a comparison data set; s2, dividing the comparison data set into a normal deviation data set and a relaxed deviation data set; calculating the data difference under the normal deviation to obtain a normal difference; calculating the data difference under the relaxed deviation to obtain a relaxed difference; s3, constructing a difference model, calculating a difference value according to the difference model, and calculating according to the difference value to obtain a difference threshold value; calculating the floating value of each data; s4, acquiring power grid measurement data, and performing anomaly mining on the power grid measurement data according to the difference threshold and the floating value of each piece of data; according to the method, the accuracy of mining the abnormal data in the power grid measurement data can be effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing and relates to power grid measurement technology. Specifically, it is a method and system for mining abnormal data in power grid measurement data. Background Technology

[0002] Existing methods and systems for mining outlier data in power grid measurement data have the following specific drawbacks:

[0003] 1. Existing methods and systems for mining abnormal data in power grid measurement data usually involve acquiring high-dimensional time-series data (such as PMU measurement data) and analyzing abnormal data based on the high-dimensional time-series data. When traditional similarity measures (such as Euclidean distance) fail, this method faces difficulties in dimensionality reduction or feature selection, which affects the accuracy of anomaly detection.

[0004] 2. Existing methods are computationally inefficient when processing massive amounts of data, making it difficult to meet real-time requirements. Furthermore, they often lack adaptability to the complex and ever-changing power grid operating environment, making it difficult to accurately identify various anomalies.

[0005] To address this, we propose a method and system for mining abnormal data in power grid measurement data. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for mining abnormal data in power grid measurement data, thereby improving the accuracy of mining abnormal data in power grid measurement data.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a method for mining abnormal data in power grid measurement data, comprising:

[0008] Step S1: Introduce a power frequency signal to the power grid, measure various data of the power grid under the power frequency signal, obtain a standard dataset, obtain the frequency deviation of the power grid during operation, introduce an alternative frequency signal to the power grid based on the frequency deviation, measure various data of the power grid under the alternative frequency signal, and obtain a comparison dataset;

[0009] Step S2: Divide the frequency deviation into normal deviation and relaxed deviation. Based on the normal deviation and relaxed deviation, divide the comparison dataset into normal deviation dataset and relaxed deviation dataset. Calculate the data difference under normal deviation based on the normal deviation dataset and the standard dataset to obtain the normal difference. Calculate the data difference under relaxed deviation based on the relaxed deviation dataset and the standard dataset to obtain the relaxed difference.

[0010] Step S3: Construct a difference model using normal difference, relaxed difference, and frequency deviation. Calculate the difference values ​​under different frequency deviations based on the difference model. Calculate the threshold based on the difference values ​​to obtain the difference threshold. Calculate the difference ratio of each data point using the normal difference set and the relaxed difference set, combined with the standard dataset, to obtain the fluctuation value.

[0011] Step S4: Acquire power grid measurement data in real time, and perform anomaly detection on the power grid measurement data based on the difference threshold and the fluctuation value of each data.

[0012] Furthermore, the specific steps of step S1 are as follows:

[0013] Step S11: Measure various data of the power grid under the power frequency signal; obtain the number of measurement data items, denoted as aa, and record the measurement values ​​of various data of the power grid as bz(1) to bz(aa); save the measurement values ​​of various data of the power grid from bz(1) to bz(aa) to obtain the standard dataset;

[0014] Step S12: Introduce a different frequency signal into the power grid, measure various data of the power grid under the different frequency signal, and record them as yp(1) to yp(aa); save the measured values ​​of various data of the power grid under the different frequency signal yp(1) to yp(aa) to obtain the comparison dataset.

[0015] Furthermore, the specific steps of step S2 are as follows:

[0016] Step S21: Divide the frequency deviation, obtain the frequency deviation when the power system is running normally, and record it as normal deviation; obtain the allowable value of the deviation under the condition of an accident or a small capacity system, and record it as relaxed deviation.

[0017] Step S22: Based on the normal deviation, set the frequency signals, test multiple frequency signals under the normal deviation by passing them into the power grid, measure various data of the power grid, and obtain the normal deviation dataset; based on the relaxed deviation, set the frequency signals, test multiple frequency signals under the relaxed deviation by passing them into the power grid, measure various data of the power grid, and obtain the relaxed deviation dataset.

[0018] Step S23: Process the normal deviation dataset, calculate it in conjunction with the standard dataset to obtain the normal difference, and statistically analyze the normal difference to obtain the normal difference set; process the relaxed deviation dataset, calculate it in conjunction with the standard dataset to obtain the relaxed difference, and statistically analyze the relaxed difference to obtain the relaxed difference set.

[0019] Furthermore, the specific steps of step S22 are as follows:

[0020] Step S221: Based on the normal deviation, set bb different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp1(a, b), where a represents the a-th measurement data, b represents the b-th frequency signal, and yp1 represents the frequency signal under the normal deviation. Save yp1(a, b) to obtain the normal deviation dataset.

[0021] Step S222: Based on the relaxation deviation, set cc different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp2(a,c), where a represents the a-th measurement data, c represents the c-th frequency signal, and yp2 represents the frequency signal under the relaxation deviation. Save yp2(a,c) to obtain the relaxation deviation dataset.

[0022] Furthermore, the specific steps of step S23 are as follows:

[0023] Step S231: Based on the standard dataset, extract the measured values ​​bz(1) to bz(aa) of various data of the power grid in the standard dataset; extract and calculate the data in the normal deviation dataset and the standard dataset to obtain the normal difference zcy;

[0024] ;

[0025] Where: zcy(a, b) represents the normal difference between the a-th data and the b-th inter-frequency signal under normal deviation; yp1(a, b) represents the measured value of the a-th data and the b-th inter-frequency signal under normal deviation; bz(a) represents the measured value of the a-th data under power frequency signal.

[0026] Statistical analysis of normal differences yields a set of normal differences.

[0027] Step S232: Extract and calculate the data from the relaxed deviation dataset and the standard dataset to obtain the relaxed difference fcy;

[0028] ;

[0029] Where: fcy(a, b) represents the relaxation difference of the a-th data and the c-th inter-frequency signal under the relaxation deviation; yp2(a, c) represents the measured value of the a-th data and the c-th inter-frequency signal under the relaxation deviation; bz(a) represents the measured value of the a-th data under the power frequency signal.

[0030] The relaxation differences are statistically analyzed to obtain a relaxation difference set.

[0031] Furthermore, the specific steps of step S3 are as follows:

[0032] Step S31: Based on the normal difference and frequency deviation, calculate the change value of the normal difference when the frequency deviation is the normal deviation, and calculate the change weight of each data in the normal deviation dataset based on the change value of the normal deviation and the normal difference.

[0033] Based on the relaxation difference and frequency deviation, the change value of the relaxation difference under the frequency deviation is calculated, and the change weight of each data in the relaxation deviation dataset is calculated based on the change value of the relaxation deviation and the relaxation difference.

[0034] Based on the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation, a model is constructed to obtain the difference model;

[0035] Step S32: Substitute the normal deviation dataset and the relaxed deviation dataset into the difference model to calculate the difference value; calculate the first threshold based on the difference value corresponding to the normal deviation dataset; calculate the second threshold based on the difference value corresponding to the relaxed deviation dataset; the first threshold and the second threshold constitute the difference threshold.

[0036] Step S33: Based on the normal difference set and the relaxed difference set, perform difference statistics on each data to obtain the representative difference value of each data; calculate the ratio of the representative difference value of each data to each data in the standard dataset to obtain the fluctuation value of each data.

[0037] Furthermore, the specific steps of step S31 are as follows:

[0038] Step S311: Obtain the normal difference set and the different frequency value zyp of the different frequency signal under normal deviation; according to the normal difference set, obtain the normal difference zcy(a,b); according to the different frequency value zyp of the different frequency signal under normal deviation and the normal difference zcy(a,b), calculate to obtain the normal change value zbh.

[0039] ;

[0040] Where: zbh(a, b) refers to the normal variation value of the a-th data item and the b-th inter-frequency signal; zyp(b) represents the inter-frequency value of the b-th inter-frequency signal;

[0041] The mean of the bb normal variation values ​​zbh(a,b) is calculated to obtain the variation weight zbq of each data in the normal deviation data set;

[0042] ;

[0043] Where: zhq(a) represents the change weight of the a-th data item in the normal deviation dataset;

[0044] Step S312: Obtain the relaxation difference set and the different frequency value fyp of the different frequency signal under the relaxation deviation; according to the relaxation difference set, obtain the relaxation difference fcy(a,c); according to the different frequency value fyp of the different frequency signal under the relaxation deviation and the relaxation difference fcy(a,c), calculate to obtain the relaxation change value fbh.

[0045] ;

[0046] Where: fbh(a,c) refers to the relaxed change value of the c-th frequency signal in the a-th data item; fyp(c) represents the frequency value of the c-th frequency signal;

[0047] The mean of cc normal variation values ​​fbh(a,c) is calculated to obtain the variation weight fbq of each data in the relaxed deviation data set;

[0048] ;

[0049] Where: fhq(a) represents the change weight of the a-th data item in the relaxed deviation data set;

[0050] Step S313: Obtain the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation. Set two values, u and v. When the value of the frequency deviation matches the normal deviation, u = 1 and v = 0. When the value of the frequency deviation matches the relaxed deviation, u = 0 and v = 1.

[0051] Based on the change weights zhq(a) of each data point in the normal deviation dataset, the change weights fhq(a) of each data point in the relaxed deviation dataset, and the standard dataset and established values ​​u and v, construct a difference model cmx;

[0052] ;

[0053] Where: CL(a) is the model parameter, which refers to the actual measured value of the a-th data item; bz(a) is the measured value of the a-th data item under power frequency signal.

[0054] Furthermore, the specific steps of step S33 are as follows:

[0055] Step S331: Obtain the normal difference set and the relaxed difference set, perform difference statistics on each data in the normal difference set and the relaxed difference set, calculate the normal difference zcy(a,b) in the normal difference set and the relaxed difference fcy(a,c) in the relaxed difference set, and obtain the difference representative value cdb.

[0056] ;

[0057] Where: cdb(a) represents the difference representative value of the a-th data item;

[0058] Step S332: Extract each data item from the standard data to obtain bz(1) to bz(aa), and calculate the ratio of bz(1) to bz(aa) with the difference representative value cdb(a) to obtain the floating value fdz of each data item;

[0059] ;

[0060] Where: fdz(a) represents the floating value of the a-th data item, and bz(a) represents the a-th standard data item.

[0061] Furthermore, the specific steps of step S4 are as follows:

[0062] Step S41: Acquire power grid measurement data in real time, substitute the power grid measurement data into the difference model for calculation, and obtain the real-time difference value scy of the power grid; obtain the first threshold dyy and the second threshold dey according to the difference threshold; compare the real-time difference value of the power grid with the first threshold and the second threshold to judge the overall data of the power grid.

[0063] If scy ≤ dyy, it indicates that the overall data of the power grid is normal.

[0064] If dyy<scy≤dey, it indicates that the overall data of the power grid is abnormal, but it is within a controllable range, and an early warning needs to be issued to maintenance personnel.

[0065] If scy > dey, the power grid data is abnormal and needs to be processed as soon as possible.

[0066] Step S42: Obtain the standard dataset, calculate the ratio between the power grid measurement data and each data item in the standard dataset, obtain the ratio value, and combine it with the fluctuation value of each data item to perform specific data anomaly checks on each data item of the power grid measurement data. If the ratio value of each data item is greater than the fluctuation value of each data item, then the data item is abnormal.

[0067] A system for mining abnormal data in power grid measurement data, the processing system includes:

[0068] Data acquisition module: used to input power frequency signals into the power grid, measure various data of the power grid under the power frequency signals, obtain a standard dataset, acquire the frequency deviation of the power grid during operation, input different frequency signals into the power grid based on the frequency deviation, measure various data of the power grid under the different frequency signals, and obtain a comparison dataset;

[0069] Data processing module: used to divide frequency deviations into normal deviations and relaxed deviations; based on normal deviations and relaxed deviations, the comparison dataset is divided into normal deviation datasets and relaxed deviation datasets; based on the normal deviation dataset and the standard dataset, the data difference under normal deviation is calculated to obtain the normal difference; based on the relaxed deviation dataset and the standard dataset, the data difference under relaxed deviation is calculated to obtain the relaxed difference.

[0070] The data analysis module is used to construct a difference model based on normal difference, relaxed difference, and frequency deviation. It calculates the difference value under different frequency deviations based on the difference model, and calculates the threshold based on the difference value to obtain the difference threshold. It also calculates the difference ratio of each data point by combining the normal difference set and the relaxed difference set with the standard dataset to obtain the fluctuation value.

[0071] Anomaly detection module: Used to acquire power grid measurement data in real time, and to detect anomalies in the power grid measurement data based on the difference threshold and the fluctuation value of each data.

[0072] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0073] 1. This invention tests the circuit by changing the frequency of the test signal, that is, by passing a signal of a certain frequency close to the power frequency to the circuit under test. By changing the frequency of the signal, more test data can be obtained and the data calculation can be made more accurate.

[0074] 2. Extract data such as voltage and current from the circuit under test, build a model based on multiple data points, and perform anomaly detection on the data; enhance the readability of the data by extracting and calculating the basic information of the test circuit, and make the results of anomaly detection easier to interpret.

[0075] 3. A model is built based on various basic data. By detecting data on the line in real time, the data is substituted into the model for calculation, simplifying the calculation process and enhancing the real-time performance and accuracy of anomaly detection. At the same time, the basic data has wide applicability, enabling the model to be widely used. Attached Figure Description

[0076] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0077] Figure 1 This is a schematic diagram of the method of the present invention;

[0078] Figure 2 This is a schematic diagram of the data processing of the present invention;

[0079] Figure 3 This is a schematic diagram of the system of the present invention; Detailed Implementation

[0080] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0081] Example 1

[0082] Please see Figure 1 This invention provides a technical solution: a method for mining abnormal data in power grid measurement data, comprising:

[0083] Step S1: Introduce a power frequency signal to the power grid, measure various data of the power grid under the power frequency signal, obtain a standard dataset, obtain the frequency deviation of the power grid during operation, introduce an alternative frequency signal to the power grid based on the frequency deviation, measure various data of the power grid under the alternative frequency signal, and obtain a comparison dataset;

[0084] It should be noted that power frequency refers to the rated operating frequency used by power generation, transmission, transformation, and power consumption equipment in a power system, measured in Hertz (Hz). It is a benchmark parameter for the stable operation of the power grid and directly affects the design, manufacture, and use of equipment.

[0085] According to the "Permissible Frequency Deviation of Power Quality Power System" (GB / T 15945-2008), the power frequency in my country is 50Hz, and the frequency deviation during normal operation shall not exceed ±0.2Hz; in the case of an accident or a small-capacity system, the deviation can be relaxed to ±0.5Hz.

[0086] Step S11: Measure various data of the power grid under the power frequency signal; obtain the number of measurement data items, denoted as aa, and record the measurement values ​​of various data of the power grid as bz(1), bz(2), ..., bz(aa); save the measurement values ​​of various data of the power grid bz(1), bz(2), ..., bz(aa) to obtain the standard dataset;

[0087] Step S12: Introduce a different frequency signal into the power grid and measure various data of the power grid under the different frequency signal, denoted as yp(1), yp(2), ... yp(aa); save the measured values ​​of various data of the power grid under the different frequency signal yp(1), yp(2), ... yp(aa) to obtain the comparison dataset.

[0088] Step S2: Divide the frequency deviation into normal deviation and relaxed deviation. Based on the normal deviation and relaxed deviation, divide the comparison dataset into normal deviation dataset and relaxed deviation dataset. Calculate the data difference under normal deviation based on the normal deviation dataset and the standard dataset to obtain the normal difference. Calculate the data difference under relaxed deviation based on the relaxed deviation dataset and the standard dataset to obtain the relaxed difference.

[0089] Step S21: According to the "Power Quality Power System Frequency Allowable Deviation", the frequency deviation is divided, the frequency deviation during normal operation of the power system is obtained and recorded as the normal deviation, and the allowable value of the deviation under the condition of accident or small capacity system is obtained and recorded as the relaxed deviation.

[0090] It should be noted that the relaxed deviation does not include the value of the normal deviation. For example, the frequency deviation should not exceed ±0.2Hz during normal operation; in the case of an accident or a small-capacity system, the deviation can be relaxed to ±0.5Hz; then the normal deviation is [-0.2, 0.2], and the relaxed deviation is [-0.5, -0.2) ∪ (0.2, 0.5];

[0091] Step S22: Based on the normal deviation, set the frequency signals, test multiple frequency signals under the normal deviation by passing them into the power grid, measure various data of the power grid, and obtain the normal deviation dataset; based on the relaxed deviation, set the frequency signals, test multiple frequency signals under the relaxed deviation by passing them into the power grid, measure various data of the power grid, and obtain the relaxed deviation dataset.

[0092] Step S221: Based on the normal deviation, set bb different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp1(a, b), where a represents the a-th measurement data, b represents the b-th frequency signal, and yp1 represents the frequency signal under the normal deviation. Save yp1(a, b) to obtain the normal deviation dataset.

[0093] Step S222: Based on the relaxation deviation, set cc different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp2(a,c), where a represents the a-th measurement data, c represents the c-th frequency signal, and yp2 represents the frequency signal under the relaxation deviation; save yp2(a,c) to obtain the relaxation deviation dataset.

[0094] Step S23: Process the normal deviation dataset, calculate it in conjunction with the standard dataset to obtain the normal difference, and statistically analyze the normal difference to obtain the normal difference set; process the relaxed deviation dataset, calculate it in conjunction with the standard dataset to obtain the relaxed difference, and statistically analyze the relaxed difference to obtain the relaxed difference set.

[0095] Step S231: Based on the standard dataset, extract the measured values ​​bz(1), bz(2), ..., bz(aa) of various data of the power grid in the standard dataset; extract and calculate the data in the normal deviation dataset and the standard dataset to obtain the normal difference zcy;

[0096] ;

[0097] Where: zcy(a, b) represents the normal difference between the a-th data and the b-th inter-frequency signal under normal deviation; yp1(a, b) represents the measured value of the a-th data and the b-th inter-frequency signal under normal deviation; bz(a) represents the measured value of the a-th data under power frequency signal.

[0098] Statistical analysis of normal differences yields a set of normal differences.

[0099] Step S232: Extract and calculate the data from the relaxed deviation dataset and the standard dataset to obtain the relaxed difference fcy;

[0100] ;

[0101] Where: fcy(a, b) represents the relaxed difference of the a-th data and the c-th inter-frequency signal under relaxed deviation; yp2(a, c) represents the measured value of the a-th data and the c-th inter-frequency signal under relaxed deviation; bz(a) represents the measured value of the a-th data under power frequency signal.

[0102] The relaxation differences are statistically analyzed to obtain a set of relaxation differences.

[0103] Step S3: Construct a difference model using normal difference, relaxed difference, and frequency deviation. Calculate the difference values ​​under different frequency deviations based on the difference model. Calculate the threshold based on the difference values ​​to obtain the difference threshold. Calculate the difference ratio of each data point using the normal difference set and the relaxed difference set, combined with the standard dataset, to obtain the fluctuation value.

[0104] Step S31: Based on the normal difference and frequency deviation, calculate the change value of the normal difference when the frequency deviation is the normal deviation, and calculate the change weight of each data in the normal deviation dataset based on the change value of the normal deviation and the normal difference.

[0105] Based on the relaxation difference and frequency deviation, the change value of the relaxation difference under the frequency deviation is calculated, and the change weight of each data in the relaxation deviation dataset is calculated based on the change value of the relaxation deviation and the relaxation difference.

[0106] Based on the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation, a model is constructed to obtain the difference model;

[0107] Step S311: Obtain the normal difference set and the different frequency value zyp of the different frequency signal under normal deviation; according to the normal difference set, obtain the normal difference zcy(a,b); according to the different frequency value zyp of the different frequency signal under normal deviation and the normal difference zcy(a,b), calculate to obtain the normal change value zbh.

[0108] ;

[0109] Where: zbh(a, b) refers to the normal variation value of the a-th data item and the b-th inter-frequency signal; zyp(b) represents the inter-frequency value of the b-th inter-frequency signal;

[0110] The mean of the bb normal variation values ​​zbh(a,b) is calculated to obtain the variation weight zbq of each data in the normal deviation data set;

[0111] ;

[0112] Where: zhq(a) represents the change weight of the a-th data item in the normal deviation dataset;

[0113] Step S312: Obtain the relaxation difference set and the different frequency value fyp of the different frequency signal under the relaxation deviation; according to the relaxation difference set, obtain the relaxation difference fcy(a,c); according to the different frequency value fyp of the different frequency signal under the relaxation deviation and the relaxation difference fcy(a,c), calculate to obtain the relaxation change value fbh.

[0114] ;

[0115] Where: fbh(a,c) refers to the relaxed change value of the a-th data item and the c-th inter-frequency signal; fyp(c) represents the inter-frequency value of the c-th inter-frequency signal.

[0116] The mean of cc normal variation values ​​fbh(a,c) is calculated to obtain the variation weight fbq of each data in the relaxed deviation data set;

[0117] ;

[0118] Where: fhq(a) represents the change weight of the a-th data item in the relaxed deviation data set;

[0119] Step S313: Obtain the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation. Set two values, u and v. When the value of the frequency deviation matches the normal deviation, u = 1 and v = 0. When the value of the frequency deviation matches the relaxed deviation, u = 0 and v = 1.

[0120] Based on the change weights zhq(a) of each data point in the normal deviation dataset, the change weights fhq(a) of each data point in the relaxed deviation dataset, and the standard dataset and established values ​​u and v, construct a difference model cmx;

[0121] ;

[0122] Where: CL(a) is the model parameter, which refers to the actual measured value of the a-th data item; bz(a) is the measured value of the a-th data item under power frequency signal.

[0123] Step S32: Substitute the normal deviation dataset and the relaxed deviation dataset into the difference model to calculate the difference value; calculate the first threshold based on the difference value corresponding to the normal deviation dataset; calculate the second threshold based on the difference value corresponding to the relaxed deviation dataset; the first threshold and the second threshold constitute the difference threshold.

[0124] Step S321: Substitute the normal deviation dataset into the difference model for calculation to obtain the difference value zcmx(b) corresponding to the normal deviation dataset. Calculate the mean of zcmx(b) to obtain the first threshold dyy.

[0125] ;

[0126] Step S322: Substitute the relaxed deviation dataset into the difference model for calculation to obtain the difference value fcmx(c) corresponding to the relaxed deviation dataset. Calculate the mean of fcmx(c) to obtain the first threshold dey.

[0127] ;

[0128] The first threshold and the second threshold are saved to obtain the difference threshold.

[0129] Step S33: Based on the normal difference set and the relaxed difference set, perform difference statistics on each data point to obtain the representative difference value of each data point; calculate the ratio of the representative difference value of each data point to each data point in the standard dataset to obtain the fluctuation value of each data point.

[0130] Please see Figure 2Step S331: Obtain the normal difference set and the relaxed difference set, perform difference statistics on each data in the normal difference set and the relaxed difference set, calculate the normal difference zcy(a,b) in the normal difference set and the relaxed difference fcy(a,c) in the relaxed difference set, and obtain the difference representative value cdb.

[0131] ;

[0132] Where: cdb(a) represents the difference representative value of the a-th data item;

[0133] Step S332: Extract each data item from the standard data to obtain bz(1), bz(2), ..., bz(aa). Calculate the ratio of bz(1), bz(2), ..., bz(aa) with the difference representative value cdb(a) to obtain the floating value fdz of each data item.

[0134] ;

[0135] Where: fdz(a) represents the floating value of the a-th data item, and bz(a) represents the a-th standard data item.

[0136] Step S4: Acquire power grid measurement data in real time, and perform anomaly detection on the power grid measurement data based on the difference threshold and the fluctuation value of each data point;

[0137] Step S41: Acquire power grid measurement data in real time, substitute the power grid measurement data into the difference model for calculation, and obtain the real-time difference value scy of the power grid; obtain the first threshold dyy and the second threshold dey according to the difference threshold; compare the real-time difference value of the power grid with the first threshold and the second threshold to judge the overall data of the power grid.

[0138] If scy ≤ dyy, it indicates that the overall data of the power grid is normal.

[0139] If dyy<scy≤dey, it indicates that the overall data of the power grid is abnormal, but it is within a controllable range, and an early warning needs to be issued to maintenance personnel.

[0140] If scy > dey, the power grid data is abnormal and needs to be processed as soon as possible.

[0141] Step S42: Obtain the standard dataset, calculate the ratio between the power grid measurement data and each data item in the standard dataset, obtain the ratio value, and combine it with the fluctuation value of each data item to perform specific data anomaly checks on each data item of the power grid measurement data. If the ratio value of each data item is greater than the fluctuation value of each data item, then the data item is abnormal.

[0142] Example 2

[0143] Please see Figure 3 A system for mining abnormal data in power grid measurement data includes: a data acquisition module, a data processing module, a data analysis module, an anomaly verification module, and a server. The data acquisition module, data processing module, data analysis module, and anomaly verification module are respectively connected to the server, and the server controls the data acquisition module, data processing module, data analysis module, and anomaly verification module respectively.

[0144] Data acquisition module: used to input power frequency signals into the power grid, measure various data of the power grid under the power frequency signals, obtain a standard dataset, acquire the frequency deviation of the power grid during operation, input different frequency signals into the power grid based on the frequency deviation, measure various data of the power grid under the different frequency signals, and obtain a comparison dataset;

[0145] Data processing module: used to divide frequency deviations into normal deviations and relaxed deviations; based on normal deviations and relaxed deviations, the comparison dataset is divided into normal deviation datasets and relaxed deviation datasets; based on the normal deviation dataset and the standard dataset, the data difference under normal deviation is calculated to obtain the normal difference; based on the relaxed deviation dataset and the standard dataset, the data difference under relaxed deviation is calculated to obtain the relaxed difference.

[0146] The data analysis module is used to construct a difference model based on normal difference, relaxed difference, and frequency deviation. It calculates the difference value under different frequency deviations based on the difference model, and calculates the threshold based on the difference value to obtain the difference threshold. It also calculates the difference ratio of each data point by combining the normal difference set and the relaxed difference set with the standard dataset to obtain the fluctuation value.

[0147] Anomaly detection module: Used to acquire power grid measurement data in real time, and to detect anomalies in the power grid measurement data based on the difference threshold and the fluctuation value of each data.

[0148] In this application, if a corresponding calculation formula appears, the above calculation formula is a dimensionless calculation. The weighting coefficient, proportional coefficient and other coefficients in the formula are set to quantify each parameter to obtain a result value. The size of the weighting coefficient and proportional coefficient is only required to not affect the proportional relationship between the parameter and the result value.

[0149] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for mining abnormal data in power grid measurement data, characterized in that, include: Step S1: Introduce a power frequency signal to the power grid, measure various data of the power grid under the power frequency signal, obtain a standard dataset, obtain the frequency deviation of the power grid during operation, introduce an alternative frequency signal to the power grid based on the frequency deviation, measure various data of the power grid under the alternative frequency signal, and obtain a comparison dataset; Step S2: Divide the frequency deviation into normal deviation and relaxed deviation. Based on the normal deviation and relaxed deviation, divide the comparison dataset into normal deviation dataset and relaxed deviation dataset. Calculate the data difference under normal deviation based on the normal deviation dataset and the standard dataset to obtain the normal difference. Calculate the data difference under relaxed deviation based on the relaxed deviation dataset and the standard dataset to obtain the relaxed difference. Step S3: Construct a difference model using normal difference, relaxed difference, and frequency deviation. Calculate the difference values ​​under different frequency deviations based on the difference model. Calculate the threshold based on the difference values ​​to obtain the difference threshold. Calculate the difference ratio of each data point using the normal difference set and the relaxed difference set, combined with the standard dataset, to obtain the fluctuation value. Step S4: Acquire power grid measurement data in real time, and perform anomaly detection on the power grid measurement data based on the difference threshold and the fluctuation value of each data.

2. The method for mining abnormal data in power grid measurement data according to claim 1, characterized in that, The specific steps of step S1 are as follows: Step S11: Measure various data of the power grid under the power frequency signal; obtain the number of measurement data items, denoted as aa, and record the measurement values ​​of various data of the power grid as bz(1) to bz(aa); save the measurement values ​​of various data of the power grid from bz(1) to bz(aa) to obtain the standard dataset; Step S12: Introduce a different frequency signal into the power grid, measure various data of the power grid under the different frequency signal, and record them as yp(1) to yp(aa); save the measured values ​​of various data of the power grid under the different frequency signal yp(1) to yp(aa) to obtain the comparison dataset.

3. The method for mining abnormal data in power grid measurement data according to claim 1, characterized in that, The specific steps of step S2 are as follows: Step S21: Divide the frequency deviation, obtain the frequency deviation when the power system is running normally, and record it as normal deviation; obtain the allowable value of the deviation under the condition of an accident or a small capacity system, and record it as relaxed deviation. Step S22: Based on the normal deviation, set the frequency signals, test multiple frequency signals under the normal deviation by passing them into the power grid, measure various data of the power grid, and obtain the normal deviation dataset; based on the relaxed deviation, set the frequency signals, test multiple frequency signals under the relaxed deviation by passing them into the power grid, measure various data of the power grid, and obtain the relaxed deviation dataset. Step S23: Process the normal deviation dataset, combine it with the standard dataset to calculate the normal difference, and perform statistics on the normal difference to obtain the normal difference set; The relaxed deviation dataset is processed and combined with the standard dataset to calculate the relaxed difference. The relaxed difference is then statistically analyzed to obtain the relaxed difference set.

4. The method for mining abnormal data in power grid measurement data according to claim 3, characterized in that, The specific steps of step S22 are as follows: Step S221: Based on the normal deviation, set bb different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp1(a, b), where a represents the a-th measurement data, b represents the b-th frequency signal, and yp1 represents the frequency signal under the normal deviation. Save yp1(a, b) to obtain the normal deviation dataset. Step S222: Based on the relaxation deviation, set cc different frequency signals, and test each frequency signal by passing it into the power grid. Based on the test results, obtain the measured values ​​of various data of the power grid under different frequency signals, denoted as yp2(a,c), where a represents the a-th measurement data, c represents the c-th frequency signal, and yp2 represents the frequency signal under the relaxation deviation. Save yp2(a,c) to obtain the relaxation deviation dataset.

5. The method for mining abnormal data in power grid measurement data according to claim 3, characterized in that, The specific steps of step S23 are as follows: Step S231: Based on the standard dataset, extract the measured values ​​bz(1) to bz(aa) of various data of the power grid in the standard dataset; extract and calculate the data in the normal deviation dataset and the standard dataset to obtain the normal difference zcy; ； Where: zcy(a, b) represents the normal difference between the a-th data and the b-th inter-frequency signal under normal deviation; yp1(a, b) represents the measured value of the a-th data and the b-th inter-frequency signal under normal deviation; bz(a) represents the measured value of the a-th data under power frequency signal. Statistical analysis of normal differences yields a set of normal differences. Step S232: Extract and calculate the data from the relaxed deviation dataset and the standard dataset to obtain the relaxed difference fcy; ； Where: fcy(a, b) represents the relaxation difference of the a-th data and the c-th inter-frequency signal under the relaxation deviation; yp2(a, c) represents the measured value of the a-th data and the c-th inter-frequency signal under the relaxation deviation; bz(a) represents the measured value of the a-th data under the power frequency signal. The relaxation differences are statistically analyzed to obtain a relaxation difference set.

6. The method for mining abnormal data in power grid measurement data according to claim 1, characterized in that, The specific steps of step S3 are as follows: Step S31: Based on the normal difference and frequency deviation, calculate the change value of the normal difference when the frequency deviation is the normal deviation, and calculate the change weight of each data in the normal deviation dataset based on the change value of the normal deviation and the normal difference. Based on the relaxation difference and frequency deviation, the change value of the relaxation difference under the frequency deviation is calculated, and the change weight of each data in the relaxation deviation dataset is calculated based on the change value of the relaxation deviation and the relaxation difference. Based on the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation, a model is constructed to obtain the difference model; Step S32: Substitute the normal deviation dataset and the relaxed deviation dataset into the difference model to calculate the difference value; calculate the first threshold based on the difference value corresponding to the normal deviation dataset; calculate the second threshold based on the difference value corresponding to the relaxed deviation dataset; the first threshold and the second threshold constitute the difference threshold. Step S33: Based on the normal difference set and the relaxed difference set, perform difference statistics on each data to obtain the representative difference value of each data; calculate the ratio of the representative difference value of each data to each data in the standard dataset to obtain the fluctuation value of each data.

7. The method for mining abnormal data in power grid measurement data according to claim 6, characterized in that, The specific steps of step S31 are as follows: Step S311: Obtain the normal difference set and the different frequency value zyp of the different frequency signal under normal deviation; according to the normal difference set, obtain the normal difference zcy(a,b); according to the different frequency value zyp of the different frequency signal under normal deviation and the normal difference zcy(a,b), calculate to obtain the normal change value zbh. ； Where: zbh(a, b) refers to the normal variation value of the a-th data item and the b-th inter-frequency signal; zyp(b) represents the inter-frequency value of the b-th inter-frequency signal; The mean of the bb normal variation values ​​zbh(a,b) is calculated to obtain the variation weight zbq of each data in the normal deviation data set; ； Where: zhq(a) represents the change weight of the a-th data item in the normal deviation dataset; Step S312: Obtain the relaxation difference set and the different frequency value fyp of the different frequency signal under the relaxation deviation; according to the relaxation difference set, obtain the relaxation difference fcy(a,c); according to the different frequency value fyp of the different frequency signal under the relaxation deviation and the relaxation difference fcy(a,c), calculate to obtain the relaxation change value fbh. ； Where: fbh(a,c) refers to the relaxed change value of the c-th frequency signal in the a-th data item; fyp(c) represents the frequency value of the c-th frequency signal; The mean of cc normal variation values ​​fbh(a,c) is calculated to obtain the variation weight fbq of each data in the relaxed deviation data set; ； Where: fhq(a) represents the change weight of the a-th data item in the relaxed deviation data set; Step S313: Obtain the change weights of each data point in the normal deviation dataset, the change weights of each data point in the relaxed deviation dataset, and the frequency deviation. Set two values, u and v. When the value of the frequency deviation matches the normal deviation, u = 1 and v = 0. When the value of the frequency deviation matches the relaxed deviation, u = 0 and v = 1. Based on the change weights zhq(a) of each data point in the normal deviation dataset, the change weights fhq(a) of each data point in the relaxed deviation dataset, and the standard dataset and established values ​​u and v, construct a difference model cmx; ； Where: CL(a) is the model parameter, which refers to the actual measured value of the a-th data item; bz(a) is the measured value of the a-th data item under power frequency signal.

8. The method for mining abnormal data in power grid measurement data according to claim 6, characterized in that, The specific steps of step S33 are as follows: Step S331: Obtain the normal difference set and the relaxed difference set, perform difference statistics on each data in the normal difference set and the relaxed difference set, calculate the normal difference zcy(a,b) in the normal difference set and the relaxed difference fcy(a,c) in the relaxed difference set, and obtain the difference representative value cdb. ； Where: cdb(a) represents the difference representative value of the a-th data item; Step S332: Extract each data item from the standard data to obtain bz(1) to bz(aa), and calculate the ratio of bz(1) to bz(aa) with the difference representative value cdb(a) to obtain the floating value fdz of each data item; ； Where: fdz(a) represents the floating value of the a-th data item, and bz(a) represents the a-th standard data item.

9. The method for mining abnormal data in power grid measurement data according to claim 1, characterized in that, The specific steps of step S4 are as follows: Step S41: Acquire power grid measurement data in real time, substitute the power grid measurement data into the difference model for calculation, and obtain the real-time difference value scy of the power grid; obtain the first threshold dyy and the second threshold dey according to the difference threshold; compare the real-time difference value of the power grid with the first threshold and the second threshold to judge the overall data of the power grid. If scy ≤ dyy, it indicates that the overall data of the power grid is normal. If dyy<scy≤dey, it indicates that the overall data of the power grid is abnormal, but it is within a controllable range, and an early warning needs to be issued to maintenance personnel. If scy > dey, the power grid data is abnormal and needs to be processed as soon as possible. Step S42: Obtain the standard dataset, calculate the ratio between the power grid measurement data and each data item in the standard dataset, obtain the ratio value, and combine it with the fluctuation value of each data item to perform specific data anomaly checks on each data item of the power grid measurement data. If the ratio value of each data item is greater than the fluctuation value of each data item, then the data item is abnormal.

10. A system for mining abnormal data in power grid measurement data, applicable to the method for mining abnormal data in power grid measurement data as described in any one of claims 1-9, characterized in that, The processing system includes: Data acquisition module: used to input power frequency signals into the power grid, measure various data of the power grid under the power frequency signals, obtain a standard dataset, acquire the frequency deviation of the power grid during operation, input different frequency signals into the power grid based on the frequency deviation, measure various data of the power grid under the different frequency signals, and obtain a comparison dataset; Data processing module: used to divide frequency deviations into normal deviations and relaxed deviations; based on normal deviations and relaxed deviations, the comparison dataset is divided into normal deviation datasets and relaxed deviation datasets; based on the normal deviation dataset and the standard dataset, the data difference under normal deviation is calculated to obtain the normal difference; based on the relaxed deviation dataset and the standard dataset, the data difference under relaxed deviation is calculated to obtain the relaxed difference. The data analysis module is used to construct a difference model based on normal difference, relaxed difference, and frequency deviation. It calculates the difference value under different frequency deviations based on the difference model, and calculates the threshold based on the difference value to obtain the difference threshold. It also calculates the difference ratio of each data point by combining the normal difference set and the relaxed difference set with the standard dataset to obtain the fluctuation value. Anomaly detection module: Used to acquire power grid measurement data in real time, and to detect anomalies in the power grid measurement data based on the difference threshold and the fluctuation value of each data.

Images