Abnormal value identification method and device of metering automation system

Abstract

The invention discloses an abnormal value identification method and device of a metering automation system, and relates to the field of abnormal value identification. According to the invention, the influence of interference factors is removed based on an interference removal mode through a dual mode of combining interference removal with the abnormal value recognition model, and the purity of data input into the abnormal value recognition model is improved; based on the mode of the abnormal value recognition model, the recognition accuracy of the abnormal value recognition model on the abnormal value is improved through the high efficiency and the high accuracy of the data recognition large model on data processing, and finally the data of the metering automation system is cleaned according to the recognized abnormal value. According to the invention, the misjudgment of the working personnel on the operation of the electric power system caused by abnormal data in the metering automation system is avoided, the safe and stable operation of the electric power system is ensured, and the safety and reliability of power supply for users are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of abnormal value recognition, and in particular to an abnormal value recognition method and device for a metering automation system. Background Art

[0002] With the popularization of power transportation equipment and the growth of power demand, the power system has become increasingly large and complex. In the power system, the application of metering automation system is extensive and in-depth, which plays a vital role in the operation, management and optimization of the power system. The metering automation system can monitor and record various operating data of power equipment in real time, including voltage, current, etc., and can also monitor and analyze the load of the power system in real time. By analyzing the data collected by the metering automation system, it is possible to analyze the actual operation of the power system and dispatch the power system.

[0003] However, the data collected by the metering automation system often have many interferences, such as electrical noise and electromagnetic interference. These interferences will cause the data collected by the metering automation system to be unstable, resulting in abnormal values, which will eventually affect the analysis and judgment of the power system. In the prior art, technicians usually manually identify and process abnormal values, but this abnormal value identification and processing method often leads to unsatisfactory identification results due to various interferences, manual experience and other factors. Therefore, there is an urgent need for a method or device for identifying abnormal values ​​in the metering automation system to improve the shortcomings of the current prior art. Summary of the invention

[0004] The present invention provides a method and a device for identifying abnormal values ​​in a metering automation system, so as to solve the technical problem that the abnormal values ​​in the metering automation system cannot be accurately identified in the prior art.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides an abnormal value identification method of a measurement automation system, comprising:

[0006] Acquire a plurality of first power data from a metering automation system, and perform interference removal on the plurality of first power data according to a preset interference removal method to obtain a plurality of second power data;

[0007] Inputting the plurality of second power data into a preset first abnormal value recognition model to obtain an abnormal value recognition set corresponding to the plurality of second power data;

[0008] According to a preset outlier marking method, in combination with the outlier identification set, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers;

[0009] According to the plurality of third power data marked with abnormal values, data in the metering automation system is cleaned.

[0010] It can be understood that, compared with the prior art, the present invention removes interference from the power data collected by the metering automation system, and recognizes abnormal values ​​through a preset abnormal value recognition model; finally, the abnormal values ​​are marked to realize the cleaning of data in the metering automation system. The present invention removes the influence of interference factors based on the interference removal method by combining the abnormal value recognition model with the dual method of interference removal, thereby improving the purity of the data input into the abnormal value recognition model; based on the method of the abnormal value recognition model, with the high efficiency and high accuracy of the data recognition large model in data processing, the recognition accuracy of the abnormal value recognition model for abnormal values ​​is improved, and finally the data of the metering automation system is cleaned according to the identified abnormal values, avoiding the staff's erroneous judgment on the operation of the power system caused by the abnormal data in the metering automation system, ensuring the safe and stable operation of the power system, and providing protection for the safety and reliability of power supply to users.

[0011] As a preferred solution, according to a preset interference removal method, interference removal is performed on the plurality of first power data to obtain a plurality of second power data, specifically:

[0012] Acquire noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method;

[0013] After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

[0014] By removing interference from power data and selecting the filtering order of low-pass filtering and sliding filtering according to the frequency ratio of noise, the interference of power data can be better removed and the purity of the data can be improved.

[0015] As a preferred solution, the inputting of the plurality of second power data into a preset first abnormal value recognition model to obtain an abnormal value recognition set corresponding to the plurality of second power data is specifically:

[0016] Acquire an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model;

[0017] A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

[0018] The outlier identification model is used to identify outliers in the data after interference removal, which improves the accuracy of outlier identification.

[0019] As a preferred solution, the initial outlier recognition model is obtained, and the initial outlier recognition model is trained according to a preset outlier recognition model training method to obtain a trained first outlier recognition model, specifically:

[0020] According to the preset outlier recognition model construction strategy, an initial outlier recognition model is constructed, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning;

[0021] According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features;

[0022] The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

[0023] By constructing the model through a variety of outlier recognition model construction strategies and selecting features through a variety of feature types, the outlier recognition model that is finally trained can have stronger recognition capabilities and higher recognition accuracy.

[0024] As a preferred solution, according to the preset outlier marking method, in combination with the outlier identification set, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers, specifically:

[0025] According to the type of outliers in the outlier identification set, combined with a preset outlier marking method, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method.

[0026] By marking the outliers, the types of outliers can be classified, so as to better clean up the outliers in the metering automation system.

[0027] Accordingly, an embodiment of the present invention provides an outlier identification device for a metrology automation system, comprising: an interference removal module, an outlier identification module, an outlier marking module, and a data cleaning module;

[0028] The interference removal module is used to obtain a number of first power data from the metering automation system, and perform interference removal on the first power data according to a preset interference removal method to obtain a number of second power data;

[0029] The outlier recognition module is used to input the plurality of second power data into a preset first outlier recognition model to obtain an outlier recognition set corresponding to the plurality of second power data;

[0030] The outlier marking module is used to mark outliers in a plurality of second power data according to a preset outlier marking method in combination with the outlier identification set, so as to obtain a plurality of third power data marked with outliers;

[0031] The data cleaning module is used to clean the data in the metering automation system according to the third power data marked with abnormal values.

[0032] As a preferred solution, the interference removal module includes: an interference removal unit;

[0033] The interference removal unit is used to obtain noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method;

[0034] After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

[0035] As a preferred solution, the outlier identification module includes: an outlier identification unit;

[0036] The outlier recognition unit is used to obtain an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model;

[0037] A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

[0038] As a preferred solution, the outlier identification unit includes: an outlier identification model training unit;

[0039] The outlier recognition model training unit is used to construct an initial outlier recognition model according to a preset outlier recognition model construction strategy, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning;

[0040] According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features;

[0041] The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

[0042] As a preferred solution, the outlier marking module includes: an outlier marking unit;

[0043] The outlier marking unit is used to mark the outliers in a number of second power data according to the type of outliers in the outlier identification set, combined with a preset outlier marking method, to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 : A flowchart of a method for identifying abnormal values ​​in a metering automation system provided by an embodiment of the present invention;

[0045] Figure 2 : A structural schematic diagram of an abnormal value identification device for a metering automation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of 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.

[0047] In the operation and maintenance of the power system, the metering automation system plays a vital role. The metering automation system can monitor and record the operating data of various equipment in the power system, and can analyze the working status of various equipment in the power system. However, abnormal values ​​in the metering automation system will affect the monitoring and recording of data, thereby affecting the management of the power system.

[0048] Outliers in the measurement automation system refer to sample points where some values ​​in the data sample are significantly deviated from the rest of the values, which can also be called outliers. There are usually three ways to deal with outliers: set as missing values, fill in, and do not process. Set as missing values ​​means setting the outliers to Null values, fill in the mean, median, mode, random number, and number 0, and do not process means treating the outliers as normal values ​​and not processing them.

[0049] Embodiment 1

[0050] Please refer to Figure 1 , is a flowchart of a method for identifying abnormal values ​​in a metering automation system provided by an embodiment of the present invention, including steps S101-S104.

[0051] Step S101: obtaining a plurality of first power data from a metering automation system, and performing interference removal on the plurality of first power data according to a preset interference removal method to obtain a plurality of second power data.

[0052] In this embodiment, the first power data is obtained from the metering automation system, and interference is removed from the first power data according to a preset interference removal method to obtain the second power data, specifically:

[0053] Acquire noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method;

[0054] After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

[0055] It should be noted that the low-pass filtering method described in this embodiment refers to a method of filtering out interference by passing data through a low-pass filter, and selecting the corresponding pole quality Q value according to the bandwidth of the signal in the data; the sliding filtering method is derived from the first-order low-pass filtering method, and the sliding regards the sudden changes in the signal as interference and removes the interference by smoothing.

[0056] In this embodiment, the proportion of noise at each frequency level in the noise information is obtained. When the proportion of high-frequency noise exceeds a preset threshold, a low-pass filter is first used to remove the high-frequency noise and then a sliding filter is used to smooth the data and noise; otherwise, a sliding filter is first used for smoothing and then a low-pass filter is used to remove the high-frequency noise.

[0057] Step S102: inputting the plurality of second power data into a preset first abnormal value recognition model to obtain abnormal value recognition sets corresponding to the plurality of second power data.

[0058] In this embodiment, the inputting of the plurality of second power data into a preset first abnormal value recognition model to obtain an abnormal value recognition set corresponding to the plurality of second power data is specifically:

[0059] Acquire an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model;

[0060] A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

[0061] In this embodiment, the initial outlier recognition model is obtained, and the initial outlier recognition model is trained according to a preset outlier recognition model training method to obtain a trained first outlier recognition model, specifically:

[0062] According to the preset outlier recognition model construction strategy, an initial outlier recognition model is constructed, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning;

[0063] According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features;

[0064] The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

[0065] It should be noted that the outlier detection strategy based on statistical distribution described in this embodiment refers to outlier detection under the premise of assuming that the data conforms to the statistical distribution; the distance-based outlier detection strategy refers to calculating and detecting outliers based on the distance of each data point; the density-based outlier detection strategy refers to detecting outliers based on the density information of the data points; the clustering-based outlier detection strategy divides the data into clusters and detects outliers based on the clusters to which they belong; the machine learning-based outlier detection strategy refers to detecting outliers based on a machine learning algorithm.

[0066] In an optional embodiment, the preset data processing method is specifically as follows: first, after removing duplicate data in a preset model training set, the data in an invalid range is deleted and the format of the data is corrected to obtain a model training set with data cleaning completed; the missing values ​​are filled according to the type of missing values ​​in the model training set, wherein the missing values ​​are filled by any one or a combination of statistical-based filling, model-based filling and business rule-based filling.

[0067] It should be noted that the numerical features described in this embodiment include: mean, standard deviation, skewness and kurtosis; the categorical features include: frequency distribution and cross analysis; the texture features include: gray-level co-occurrence matrix generation features and Fourier transform generation features; the shape features are described by boundary descriptors; and the color features are represented by color histograms.

[0068] In an optional embodiment, the selection criteria of the basic model of the outlier recognition model include: the data type of the data to be recognized, the interference type in the data to be recognized, and the characteristics of the algorithm model; the selectable models of the basic model of the outlier recognition model include: the isolation forest model and the DBSCAN clustering model. The evaluation criteria of the outlier recognition model include: the recognition accuracy and the recall rate of the model.

[0069] In an optional embodiment, the training method of the outlier recognition model includes: a supervised learning method or an unsupervised learning method. Among them, the supervised learning method refers to training the model by annotating the training data, that is, constructing a mapping relationship between the model input and output. The unsupervised learning method includes: an isolation forest algorithm or a DBSCAN algorithm. After the training of the outlier recognition model described in this embodiment is completed, it also includes adaptively adjusting the parameters of the model according to the evaluation results of the outlier recognition model, so that the outlier recognition model can select better features.

[0070] Step S103: according to a preset outlier marking method, in combination with the outlier identification set, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers.

[0071] In this embodiment, according to the preset outlier marking method, in combination with the outlier identification set, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers, specifically:

[0072] According to the type of outliers in the outlier identification set, combined with a preset outlier marking method, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method.

[0073] It should be noted that the image-based marking method described in this embodiment refers to drawing the data into a chart or a curve chart to mark outliers; the statistical distribution-based marking method refers to marking outliers according to the statistical distribution characteristics of the data; the rule-based marking method refers to marking outliers based on preset rules and thresholds; the automatic-based marking method refers to marking outliers through statistical methods or machine learning algorithms.

[0074] Step S104: Cleaning up the data in the metering automation system according to the plurality of third power data marked with abnormal values.

[0075] This embodiment removes interference from the power data collected from the metering automation system, and recognizes outliers through a preset outlier recognition model; finally, outliers are marked to clean up the data in the metering automation system. The present invention uses a dual method of removing interference and combining it with an outlier recognition model, and removes the influence of interference factors based on the interference removal method to improve the purity of the data input into the outlier recognition model; based on the outlier recognition model method, with the high efficiency and high accuracy of the data recognition large model in data processing, the outlier recognition model improves the recognition accuracy of outliers, and finally cleans up the data of the metering automation system according to the identified outliers, avoiding the staff's erroneous judgment of the operation of the power system caused by the abnormal data in the metering automation system, ensuring the safe and stable operation of the power system, and providing protection for the safety and reliability of power supply to users.

[0076] Embodiment 2

[0077] Please refer to Figure 2 , which is a structural schematic diagram of an outlier identification device for a metrology automation system provided in an embodiment of the present invention, comprising: an interference removal module 201, an outlier identification module 202, an outlier marking module 203 and a data cleaning module 204.

[0078] The interference removal module 201 is used to obtain a number of first power data from the metering automation system, and perform interference removal on the first power data according to a preset interference removal method to obtain a number of second power data.

[0079] In this embodiment, the interference removal module includes: an interference removal unit;

[0080] The interference removal unit is used to obtain noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method;

[0081] After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

[0082] The outlier identification module 202 is used to input the plurality of second power data into a preset first outlier identification model to obtain an outlier identification set corresponding to the plurality of second power data.

[0083] In this embodiment, the outlier identification module includes: an outlier identification unit;

[0084] The outlier recognition unit is used to obtain an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model;

[0085] A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

[0086] In this embodiment, the outlier identification unit includes: an outlier identification model training unit;

[0087] The outlier recognition model training unit is used to construct an initial outlier recognition model according to a preset outlier recognition model construction strategy, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning;

[0088] According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features;

[0089] The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

[0090] The outlier marking module 203 is used to mark outliers in a number of second power data according to a preset outlier marking method in combination with the outlier identification set, so as to obtain a number of third power data marked with outliers.

[0091] In this embodiment, the outlier marking module includes: an outlier marking unit;

[0092] The outlier marking unit is used to mark the outliers in a number of second power data according to the type of outliers in the outlier identification set, combined with a preset outlier marking method, to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method.

[0093] The data cleaning module 204 is used to clean the data in the metering automation system according to the third power data marked with abnormal values.

[0094] This embodiment removes interference from the power data collected from the metering automation system, and recognizes outliers through a preset outlier recognition model; finally, outliers are marked to clean up the data in the metering automation system. The present invention uses a dual method of removing interference and combining it with an outlier recognition model, and removes the influence of interference factors based on the interference removal method to improve the purity of the data input into the outlier recognition model; based on the outlier recognition model method, with the high efficiency and high accuracy of the data recognition large model in data processing, the outlier recognition model improves the recognition accuracy of outliers, and finally cleans up the data of the metering automation system according to the identified outliers, avoiding the staff's erroneous judgment of the operation of the power system caused by the abnormal data in the metering automation system, ensuring the safe and stable operation of the power system, and providing protection for the safety and reliability of power supply to users.

[0095] In summary, the embodiment of the present invention removes interference from the power data collected by the metering automation system, and recognizes abnormal values ​​through a preset abnormal value recognition model; finally, the abnormal values ​​are marked to realize the cleaning of data in the metering automation system. The present invention uses a dual method of removing interference and combining an abnormal value recognition model to remove the influence of interference factors based on the interference removal method, thereby improving the purity of the data input into the abnormal value recognition model; based on the method of the abnormal value recognition model, with the high efficiency and high accuracy of the data recognition large model in data processing, the recognition accuracy of the abnormal value recognition model for abnormal values ​​is improved, and finally the data of the metering automation system is cleaned according to the identified abnormal values, avoiding the staff's erroneous judgment on the operation of the power system caused by the abnormal data in the metering automation system, ensuring the safe and stable operation of the power system, and providing protection for the safety and reliability of power supply to users.

[0096] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying abnormal values ​​in a measurement automation system, characterized in that: include: Acquire a plurality of first power data from a metering automation system, and perform interference removal on the plurality of first power data according to a preset interference removal method to obtain a plurality of second power data; Inputting the plurality of second power data into a preset first abnormal value recognition model to obtain an abnormal value recognition set corresponding to the plurality of second power data; According to a preset outlier marking method, in combination with the outlier identification set, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers; According to the plurality of third power data marked with abnormal values, data in the metering automation system is cleaned.

2. The method for identifying abnormal values ​​in a measurement automation system according to claim 1, characterized in that: The method of obtaining a plurality of first power data from the metering automation system and removing interference from the plurality of first power data according to a preset interference removal method to obtain a plurality of second power data is specifically as follows: Acquire noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method; After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

3. The method for identifying abnormal values ​​in a measurement automation system according to claim 1, characterized in that: The inputting of the plurality of second power data into a preset first abnormal value recognition model to obtain an abnormal value recognition set corresponding to the plurality of second power data is specifically: Acquire an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model; A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

4. The method for identifying abnormal values ​​in a measurement automation system according to claim 3, characterized in that: The initial outlier recognition model is obtained, and the initial outlier recognition model is trained according to a preset outlier recognition model training method to obtain a trained first outlier recognition model, specifically: According to the preset outlier recognition model construction strategy, an initial outlier recognition model is constructed, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning; According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features; The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

5. The method for identifying abnormal values ​​in a measurement automation system according to claim 1, characterized in that: According to the preset outlier marking method, in combination with the outlier identification set, outliers in a plurality of second power data are marked to obtain a plurality of third power data marked with outliers, specifically: According to the type of outliers in the outlier identification set, combined with a preset outlier marking method, outliers in a number of second power data are marked to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method.

6. An abnormal value identification device for a measurement automation system, characterized in that: include: Interference removal module, outlier identification module, outlier marking module and data cleaning module; The interference removal module is used to obtain a number of first power data from the metering automation system, and perform interference removal on the first power data according to a preset interference removal method to obtain a number of second power data; The outlier recognition module is used to input the plurality of second power data into a preset first outlier recognition model to obtain an outlier recognition set corresponding to the plurality of second power data; The outlier marking module is used to mark outliers in a plurality of second power data according to a preset outlier marking method in combination with the outlier identification set, so as to obtain a plurality of third power data marked with outliers; The data cleaning module is used to clean the data in the metering automation system according to the third power data marked with abnormal values.

7. The abnormal value identification device of the measurement automation system according to claim 6, characterized in that: The interference removal module comprises: an interference removal unit; The interference removal unit is used to obtain noise information and a preset interference removal method in a plurality of first power data, wherein the noise information includes: noise of a plurality of frequency levels and a proportion of noise of each frequency level in the noise information; the preset interference removal method includes: a low-pass filtering method and a sliding filtering method; After determining the order of using the low-pass filtering method and the sliding filtering method according to the proportion of the noise of each frequency level in the noise information, the plurality of first power data are filtered to obtain the plurality of second power data.

8. The abnormal value identification device of a measurement automation system according to claim 6, characterized in that: The outlier identification module includes: an outlier identification unit; The outlier recognition unit is used to obtain an initial outlier recognition model, and train the initial outlier recognition model according to a preset outlier recognition model training method to obtain a trained first outlier recognition model; A plurality of second power data are input into the trained first outlier recognition model to obtain outlier recognition sets corresponding to the plurality of second power data.

9. The abnormal value identification device of the measurement automation system according to claim 8, characterized in that: The outlier identification unit includes: an outlier identification model training unit; The outlier recognition model training unit is used to construct an initial outlier recognition model according to a preset outlier recognition model construction strategy, wherein the preset outlier recognition model construction strategy includes: an outlier detection strategy based on statistical distribution, an outlier detection strategy based on distance, an outlier detection strategy based on density, an outlier detection strategy based on clustering, and an outlier detection strategy based on machine learning; According to the preset data processing method and feature selection method, the preset model training set is processed and feature selected in turn to obtain the first model training set; wherein the data processing method includes: data cleaning and missing value filling; the features include: statistical features, numerical features, classification features, image features, texture features, shape features and color features; The first model training set is input into the initial outlier recognition model, and the initial outlier recognition model is trained according to a preset outlier recognition model training method until the recognition result output by the initial outlier recognition model meets the preset standard, and the training of the initial outlier recognition model is completed to obtain a trained first outlier recognition model.

10. The abnormal value identification device of a measurement automation system according to claim 6, characterized in that: The outlier marking module comprises: an outlier marking unit; The outlier marking unit is used to mark the outliers in a number of second power data according to the type of outliers in the outlier identification set, combined with a preset outlier marking method, to obtain a number of third power data marked with outliers; wherein the outlier marking method includes: any one or a combination of a graphic-based marking method, a statistical distribution-based marking method, a rule-based marking method and an automatic-based marking method.

Images