Information anomaly identification fitting method suitable for wide-area measurement data quality improvement

Abstract

The application provides an information anomaly identification fitting method suitable for wide-area measurement data quality improvement, relates to the technical field of information anomaly identification, and comprises the following steps: S1, multiple measurement nodes are arranged on a power distribution network, and collection results of wide-area measurement data are obtained; S2, comparison analysis is performed on the collection results, and an abnormal data feature library is constructed; S3, an information anomaly identification model for data improvement is constructed; S4, abnormal data such as missing data and collection failure data are screened; and S5, missing data is comprehensively fitted and filled according to historical data. Through the arrangement of multiple measurement nodes, the data obtained by each measurement node is compared, and missing data and fault data are filled by using historical measurement data and adjacent measurement node data, so that the problem that the existing measurement data quality cannot be improved and abnormal or missing measurement data cannot be filled is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information anomaly identification, and particularly relates to an information anomaly identification fitting method suitable for wide-area measurement data quality improvement. BACKGROUND

[0002] Wide Area Measurement System (WAMS) is a real-time monitoring system based on synchronous vector measurement technology and aiming at power system dynamic process detection, analysis and control. WAMS has the technical characteristics of high-precision synchronous vector measurement in different places, high-speed communication and rapid reflection, and is very suitable for large-span power grids, especially for the real-time monitoring of dynamic processes of interconnected power grids.

[0003] In the prior art, when wide-area measurement data is identified, there are deficiencies in analyzing the collection quality status of the wide-area measurement data, and the existing identification method cannot effectively identify the anomalies of the measurement data, and there are problems of being unable to accurately correct abnormal data or fill in missing data. SUMMARY

[0004] In view of the deficiencies in the prior art, the present application aims to provide an information anomaly identification fitting method suitable for wide-area measurement data quality improvement, which compares the data obtained by each measurement node by setting multiple measurement nodes, and fills in the fault data and missing data through historical measurement data and adjacent measurement node data, so as to solve the problem that the existing measurement data quality has not been improved and the abnormal or missing measurement data cannot be filled in.

[0005] In order to achieve the above-mentioned purpose, the present application is realized by the following technical scheme: an information anomaly identification fitting method suitable for wide-area measurement data quality improvement, comprising the following steps:

[0006] Step S1: multiple measurement nodes are set for a distribution network, and the collection results of wide-area measurement data are obtained through the measurement nodes, wherein the collection results include node actual voltage, node actual current, node current threshold, node voltage threshold, voltage fluctuation value and current fluctuation value;

[0007] Step S2: comparing and analyzing the collection results of the wide-area measurement data, obtaining the user characteristics of the abnormal data that have not been successfully collected, and constructing an abnormal data feature library;

[0008] Step S3: based on the abnormal data feature library, an information anomaly identification model for data improvement is constructed;

[0009] Step S4: using the information anomaly data identification model, abnormal detection is performed on the data, and abnormal data such as missing data and collection failure are screened out;

[0010] Step S5: According to the historical data of the measurement node, the missing data is comprehensively fitted and filled.

[0011] Further, the step S1 comprises the following sub-steps:

[0012] Step S101: Based on the element configuration in the power distribution network, a plurality of measurement nodes are configured, the actual current and voltage of each measurement node are acquired, and the historical node data is set;

[0013] Step S102: The historical node data of the measurement node is acquired, the historical average current of the node is obtained, the node current threshold is set, the historical average voltage is obtained, and the node voltage threshold is set;

[0014] Step S103: The preset current and preset voltage of the measurement node are acquired; the absolute value of the difference between the node current threshold and the preset current is set as the current fluctuation value, and the absolute value of the difference between the node voltage threshold and the preset voltage is set as the voltage fluctuation value.

[0015] Further, the step S2 comprises the following sub-steps:

[0016] Step S201: The acquisition result collected by the measurement node is acquired; the acquisition result is compared and analyzed to obtain a data acquisition normal or data acquisition abnormal signal;

[0017] Step S202: When the data acquisition abnormality is detected, the fault analysis of the corresponding measurement node is performed to obtain a fault node, and the data of the fault node is set as abnormal data;

[0018] Step S203: The user characteristics corresponding to the abnormal data are acquired, the user characteristics include user type and power consumption period, and an abnormal data feature library is established.

[0019] Further, the comparison and analysis in the step S201 comprises:

[0020] Step S20101: The node current threshold and the current fluctuation value are acquired, the sum of the node current threshold and the current fluctuation value is set as a first node current, and the difference between the node current threshold and the current fluctuation value is set as a second node current;

[0021] Step S20102: The node real-time current is acquired, when the node real-time current is less than or equal to the first node current and greater than or equal to the second node current, it is judged that the data acquisition is normal;

[0022] When the node real-time current is greater than the first node current or the node real-time current is less than the second node current, it is judged that the data acquisition is abnormal;

[0023] Step S20103: Obtain a node voltage threshold and a voltage fluctuation value, set the sum of the node voltage threshold and the voltage fluctuation value as a first node voltage, and set the difference between the node voltage threshold and the voltage fluctuation value as a second node voltage;

[0024] Step S20104: Obtain a node real-time voltage, and when the node real-time voltage is less than or equal to the first node voltage and greater than or equal to the second node voltage, determine that data acquisition is normal.

[0025] When the node real-time voltage is greater than the first node voltage or the node real-time voltage is less than the second node voltage, it is determined that data acquisition is abnormal.

[0026] Further, the fault analysis in the step S202 includes:

[0027] When data acquisition is detected to be abnormal, a fault analysis is performed on the corresponding measurement node to obtain a fault node, and data of the fault node is set as abnormal data.

[0028] Step S20201: Obtain a measurement node in which data appears to be abnormal, set it as a to-be-detected node, and obtain two adjacent measurement nodes of the to-be-detected node, set them as standard nodes.

[0029] Step S20202: Compare and analyze the collection results of the standard nodes.

[0030] Step S20203: When the collection results of the two standard nodes are both normal, the to-be-detected node is determined to be a fault node.

[0031] When the collection results of the two standard nodes are both abnormal, the to-be-detected node is determined to be a normal node.

[0032] Further, the step S4 includes: using an information abnormal data identification model to perform abnormal detection on the real-time obtained node data, when a fault node is detected, the voltage data and the current data of the node are emptied and set as abnormal empty data,

[0033] The step S4 further includes: setting all missing data as empty, setting it as missing empty data, and obtaining a measurement node corresponding to the missing empty data.

[0034] Further, the step S5 includes the following sub-steps:

[0035] Step S501: Obtain abnormal empty data and a corresponding fault node, obtain two adjacent measurement nodes of the fault node, and when the two adjacent measurement nodes of the fault node are both normal nodes, set the two adjacent measurement nodes as complementary empty nodes.

[0036] Step S502: When the fault node is adjacent to two measurement nodes, take one measurement node outward from the fault node again until all the measurement nodes are normal nodes, and set the two measurement nodes as empty nodes;

[0037] Step S503: Obtain the node current threshold and node voltage threshold of the two empty nodes, obtain the average value of the node current threshold of the two empty nodes and set it as the empty current, obtain the average value of the node voltage threshold of the two empty nodes and set it as the empty voltage, and fill the empty current and the empty voltage into the abnormal empty data;

[0038] Step S504: Obtain the missing empty data and the corresponding measurement node, obtain the historical data of the measurement node, and obtain the node current threshold and the node voltage threshold of the measurement node;

[0039] Step S505: Fill the node current threshold and the node voltage threshold into the missing empty data.

[0040] Further, the step S3 comprises the following sub-steps:

[0041] Step S301: Construct a node topology network according to the topology relationship of the measurement nodes;

[0042] Step S302: Perform K-MEAN clustering analysis on all abnormal data to obtain a plurality of abnormal data clusters;

[0043] Step S303: Analyze the correlation between each abnormal data cluster and the fault result, and configure the correlation priority between the abnormal data cluster and the abnormal result according to the correlation;

[0044] Step S304: Sort the abnormal data clusters of each abnormal result according to the size of the correlation priority to obtain an abnormal sequence;

[0045] Step S305: Extract the corresponding weight correction value according to the user feature extraction, and mark the corresponding weight correction value in the abnormal sequence;

[0046] Step S306: Bring the abnormal sequence into the node topology network to generate an abnormal recognition path;

[0047] Step S307: Extract the first abnormal recognition item in the abnormal recognition path as a recognition trigger item, and the information abnormal recognition model is a set of abnormal recognition paths.

[0048] Further, the step S306 comprises:

[0049] Step S30601: Take the first abnormal recognition item in the abnormal sequence as the predecessor node of the abnormal recognition path;

[0050] Step S30602, calculate the recognition cost of other abnormal recognition items in the abnormal sequence by a sequence recognition algorithm P re = β1(r f - Δb n ) + β2Δb re + β3[R f - D(re, n) * d re ], wherein P re is the recognition cost, β1is a preset current priority weight, β2is a preset correlation priority weight, β3is a preset cost priority weight, and β1+ β2+ β3= 1, r f is a preset reference recognition value, Δb n is a correlation priority value of the previous abnormal recognition item, Δb re is a correlation priority value of the current abnormal recognition item, R f is a preset reference recognition cost, D(re, n) is a node mapping function reflecting the mapping relationship between the sequence number of the previous abnormal recognition item in the abnormal sequence and the retrieval cost value corresponding to the sequence number of the current abnormal recognition item, and d re is the effective distance between the measurement node corresponding to the previous abnormal recognition item and the measurement node corresponding to the current abnormal recognition item.

[0051] Step S30603, determine the abnormal recognition item with the highest recognition cost as the next abnormal recognition item of the abnormal recognition path, and return to step S30602 until the abnormal recognition path is completed.

[0052] Further, step S4 includes:

[0053] Step S401, identify the abnormal data of the measurement node according to the trigger recognition item;

[0054] Step S402, each abnormal recognition path is configured with an index value, an index range is generated according to the index value, and a recognition random number is generated, and the next abnormal recognition item is determined according to the index range in which the recognition random number falls;

[0055] Step S403, obtain the recognition content of the abnormal recognition item corresponding to the next abnormal recognition item and obtain the corresponding measurement data to obtain the corresponding abnormal sub-value by comparison;

[0056] Step S404, calculate the abnormal total value by a preset abnormal weighting algorithm P , wherein L d is the abnormal total value, u is a preset abnormal trigger reference value, K is the total number of the current abnormal recognition items, and h mData abnormal value of the mth anomaly identification item, reflecting the abnormal degree of the measurement data corresponding to the anomaly identification item, α m The preset identification item weight of the mth anomaly identification item;

[0057] Step S405, if the total anomaly value is greater than the preset trigger upper limit value, output the anomaly result and enter step S4061, if the total anomaly value is less than the preset index lower limit value, enter step S4062, and if the total anomaly value is between the trigger upper limit value and the index lower limit value, return to step S402;

[0058] Step S4061, a gain factor is configured to increase the index value of the corresponding anomaly identification item, and the gain factor decreases over time;

[0059] Step S4062, a decay factor is configured to decrease the index value of the corresponding anomaly identification item, and the decay factor decreases over time.

[0060] Advantages of the present application: the present application firstly sets multiple measurement nodes on the power distribution network, acquires the collection results of wide-area measurement data through the measurement nodes, compares and analyzes the collection results, screens out abnormal data, and then analyzes the fault of the abnormal data and the data of the adjacent measurement nodes, so as to realize the screening of the fault cause and provide basic support for filling the data;

[0061] According to the abnormal data of the fault node, an abnormal data feature library is established, and an information anomaly identification model is constructed based on the abnormal data feature library; the information anomaly data identification model can be used for abnormal detection of data, accurate screening of abnormal data such as missing and collection failure, and accurate positioning for subsequent data filling.

[0062] The advantages of the additional aspects of the present application will be partially given in the following description of the specific embodiments, partially become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0063] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the following drawings:

[0064] Figure 1 A step flow chart of the method of the present application;

[0065] Figure 2 A judgment schematic diagram of the fault node of the present application;

[0066] Figure 3 A judgment schematic diagram of the normal node of the present application. DETAILED DESCRIPTION

[0067] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0068] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the application.

[0069] In the case of no conflict, the embodiments in the application and the features in the embodiments can be combined with each other.

[0070] Please refer to Figure 1 As shown in the drawings, the application provides an information anomaly identification fitting method suitable for wide-area measurement data quality improvement. By setting multiple measurement nodes, the data obtained by each measurement node is compared, and the historical measurement data and adjacent measurement node data are used to fill in the fault data and missing data, so that the existing measurement data quality is not improved, and the problem of abnormal or missing measurement data cannot be filled in.

[0071] Specifically, the information anomaly identification fitting method suitable for wide-area measurement data quality improvement comprises the following steps:

[0072] Step S1: multiple measurement nodes are set for the distribution network, and the acquisition results of wide-area measurement data are obtained through the measurement nodes. The acquisition results include node actual voltage, node actual current, node current threshold, node voltage threshold, voltage fluctuation value and current fluctuation value. In the specific implementation process, multiple measurement nodes are set, which can avoid the problem of insufficient measurement depth caused by too few measurement nodes, resulting in that the finally obtained data is not comprehensive;

[0073] Step S1 comprises the following substeps:

[0074] Step S101: based on the elements in the distribution network, multiple measurement nodes are configured, and the actual current and voltage of each measurement node are obtained, which are set as historical node data;

[0075] Step S102: the historical node data of the measurement node is obtained, the historical average current of the node is calculated, and the node current threshold is set, the historical average voltage is calculated, and the node voltage threshold is set;

[0076] Step S103: the preset current and the preset voltage of the measurement node are obtained; the absolute value of the difference between the node current threshold and the preset current is set as the current fluctuation value, and the absolute value of the difference between the node voltage threshold and the preset voltage is set as the voltage fluctuation value;

[0077] Step S2: comparing and analyzing the collection results of the wide-area measurement data, obtaining user features of the abnormal data that are not collected successfully, and constructing an abnormal data feature library;

[0078] Step S2 includes the following sub-steps:

[0079] Step S201: obtaining the collection results collected by the measurement node; comparing and analyzing the collection results to obtain a data collection normal or data collection abnormal signal;

[0080] The comparison and analysis includes:

[0081] Step S20101: obtaining a node current threshold and a current fluctuation value, setting the sum of the node current threshold and the current fluctuation value as a first node current, and setting the difference between the node current threshold and the current fluctuation value as a second node current; in the specific implementation process, the first node current is the upper limit of the normal current fluctuation of the measurement node, and the second node current is the lower limit of the normal current fluctuation of the measurement node;

[0082] Step S20102: obtaining a node real-time current, when the node real-time current is less than or equal to the first node current and greater than or equal to the second node current, determining that the data collection is normal;

[0083] When the node real-time current is greater than the first node current or the node real-time current is less than the second node current, it is determined that the data collection is abnormal;

[0084] Step S20103: obtaining a node voltage threshold and a voltage fluctuation value, setting the sum of the node voltage threshold and the voltage fluctuation value as a first node voltage, and setting the difference between the node voltage threshold and the voltage fluctuation value as a second node voltage; in the specific implementation process, the first node voltage is the upper limit of the normal voltage fluctuation of the measurement node, and the second node voltage is the lower limit of the normal voltage fluctuation of the measurement node;

[0085] Step S20104: obtaining a node real-time voltage, when the node real-time voltage is less than or equal to the first node voltage and greater than or equal to the second node voltage, determining that the data collection is normal;

[0086] When the node real-time voltage is greater than the first node voltage or the node real-time voltage is less than the second node voltage, it is determined that the data collection is abnormal;

[0087] Step S202: when the data collection is detected to be abnormal, performing fault analysis on the corresponding measurement node to obtain a fault node, and setting the data of the fault node as abnormal data;

[0088] Please refer to Figure 2 and Figure 3 , the fault analysis includes:

[0089] When the data acquisition anomaly is detected, the fault analysis is performed on the corresponding measurement node to obtain a fault node, and the data of the fault node is set as abnormal data;

[0090] Step S20201: obtaining the measurement node with data acquisition anomaly, setting it as a to-be-detected node, obtaining two adjacent measurement nodes of the to-be-detected node, and setting them as standard nodes;

[0091] Step S20202: performing comparison analysis on the collection results of the standard nodes;

[0092] Step S20203: when the collection results of the two standard nodes are both normal data acquisition, the to-be-detected node is judged as a fault node;

[0093] When the collection results of the two standard nodes are both data acquisition anomaly, the to-be-detected node is judged as a normal node; in the specific implementation process, when a power grid line fails, the data acquisition of the measurement nodes on the power grid line is judged to be abnormal, and the data acquisition of all measurement nodes is abnormal, which also proves that the to-be-detected node is a normal node;

[0094] When a power grid line does not fail, the data acquisition of the measurement nodes on the power grid line is normal, and only part of or only the to-be-detected node is judged as data acquisition anomaly, and the to-be-detected node is a fault node;

[0095] Step S203: obtaining the user features corresponding to the abnormal data, the user features including a user type and a power consumption period, and establishing an abnormal data feature library;

[0096] Step S3: constructing an information anomaly recognition model for data improvement based on the abnormal data feature library; the step S3 includes the following sub-steps:

[0097] Step S301: constructing a node topology network according to the topology relationship of the measurement nodes; the topology network represents the topology relationship between the measurement nodes.

[0098] Step S302: performing K-MEAN clustering analysis on all abnormal data to obtain a plurality of abnormal data clusters; through the clustering analysis algorithm, similar abnormal conditions can be clustered to obtain abnormal recognition knowledge corresponding to the data.

[0099] Step S303: analyzing the correlation between each abnormal data cluster and the fault result, and configuring the correlation priority between the abnormal data cluster and the abnormal result according to the correlation; the correlation analysis between the clustering result and the abnormal result can obtain whether the abnormal data cluster is a factor causing the abnormal result, and analyze the corresponding correlation degree, preferably through a person correlation analysis algorithm.

[0100] Step S304: Sort the abnormal data clusters of each abnormal result according to the size of the relevance priority value to obtain the abnormal sequence; the purpose of relevance sorting is that since the amount of data is very large, it is necessary to start with the data that is more obvious to the abnormal situation and abnormal result, so that it is easier to configure the corresponding identification method and data monitoring method, which can effectively reduce the amount of data.

[0101] Step S305: Extract the corresponding weight correction value based on the user characteristics, and mark the corresponding weight correction value on the abnormal sequence; through the weight correction value, the possibility of the abnormal event recurring can be judged based on whether there are the same user characteristics during identification, which can also quickly identify and improve identification efficiency.

[0102] Step S306: Introduce the abnormal sequence into the node topology network to generate an anomaly identification path. Since the abnormal sequence includes several abnormal data clusters, each abnormal data cluster can be identified by querying a pre-built anomaly identification table through data feature analysis to obtain an anomaly identification item. The anomaly identification item is to identify whether the change in the data matches the content of the corresponding anomaly identification item. For example, to identify anomalies in the data at point A, it may be necessary to obtain data at point B. This task of obtaining and comparing the data can be considered as an anomaly identification item. Therefore, the order of the sequence cannot directly represent the order of anomaly identification, and the difficulty of retrieving the data must also be considered. Step S306 includes:

[0103] Step S30601: Take the first anomaly identification item in the anomaly sequence as the preceding node of the anomaly identification path;

[0104] Step S30602: Calculate the recognition cost of other anomaly identification items in the anomaly sequence using a sequence recognition algorithm, wherein the sequence recognition algorithm is P. re =β1(r f -Δb n )+β2Δb re +β3[R f -D(re,n)*d re ], where P re To identify the cost, β1 is the preset current priority weight, β2 is the preset relevance priority weight, and β3 is the preset cost priority weight, with β1 + β2 + β3 = 1, r f Δb is the preset benchmark identification value. n Δb is the relevance priority value of the previous anomaly identification item. re R is the relevance priority value of the current anomaly identification item. f Let D(re, n) be the preset baseline identification cost, and let D(re, n) be the node mapping function, reflecting the mapping relationship between the index of the previous anomaly identification item in the anomaly sequence and the replacement value corresponding to the index of the current anomaly identification item in the anomaly sequence.re The effective distance between the measurement node corresponding to the last abnormality identification item and the measurement node corresponding to the current abnormality identification item is determined; the identification cost is the highest, indicating that the calling difficulty is the lowest, so the identification cost of each abnormality identification item is calculated in turn, and the identification path is constructed.

[0105] Step S30603, determine the abnormality identification item with the highest identification cost as the next abnormality identification item of the abnormality identification path, and return to step S30602 until the abnormality identification path is constructed.

[0106] Step S307, extract the first abnormality identification item in the abnormality identification path as the identification trigger item, and the information abnormality identification model is a set of abnormality identification paths.

[0107] Step S4: using the information abnormality data identification model, the abnormal data is detected and screened, such as missing, collection failure and the like; step S4 includes:

[0108] Step S401, identify the abnormal data of the measurement node according to the trigger identification item; the principle is to quickly screen through the corresponding content of the trigger identification item. Since one abnormal result may correspond to multiple trigger identification items, generally, such setting can increase efficiency, but can also reduce the possibility of missing abnormal events.

[0109] Step S402, each abnormality identification path is configured with an index value, and the corresponding index range is generated according to the index value, and an identification random number is generated, and the next abnormality identification item is determined according to the index range where the identification random number falls; because multiple abnormality identification paths may share abnormality identification items, after identification, the position of the next identification item in each path is determined by the random number generated, and the index value is determined by the gain factor, the attenuation factor and the identification cost of the corresponding abnormality identification item.

[0110] Step S403, obtain the identification content of the abnormality identification item corresponding to the next abnormality identification item and obtain the corresponding measurement data, to obtain the corresponding abnormal subvalue;

[0111] Step S404, calculate the total abnormal value by a preset abnormal weighting algorithm, the abnormal weighting algorithm is Where L d is the total abnormal value, u is a preset abnormal trigger reference value, the abnormal trigger reference value is obtained by looking up the table according to the weight correction value, K is the total number of the current abnormality identification item, h m is the data abnormal value of the mth abnormality identification item, reflecting the abnormality degree of the measurement data corresponding to the abnormality identification item, a m is the preset identification item weight of the mth abnormality identification item;

[0112] Step S405, if the abnormal total value is greater than the preset trigger upper limit value, an abnormal result is output and step S4061 is entered, if the abnormal total value is less than the preset index lower limit value, step S4062 is entered, and if the abnormal total value is between the trigger upper limit value and the index lower limit value, step S402 is returned;

[0113] Step S4061, a gain factor is configured to increase the index value of the corresponding abnormal identification item, and the gain factor decreases over time;

[0114] Step S4062, a decay factor is configured to decrease the index value of the corresponding abnormal identification item, and the decay factor decreases over time. In this way, the corresponding optimal identification path can be learned, and dynamic gain or decay can ensure that the learning result will not be solidified.

[0115] Step S4 includes: using the information abnormal data identification model to perform abnormal detection on the real-time acquired node data, when a fault node is detected, the voltage data and the current data of the node are nullified and set as abnormal null data,

[0116] The step S4 further includes: setting all missing data as null, setting as missing null data, and acquiring the measurement node corresponding to the missing null data;

[0117] Step S5: according to the historical data of the measurement node, the missing data is comprehensively fitted and filled;

[0118] Step S5 includes the following sub-steps:

[0119] Step S501: acquiring the abnormal null data and the corresponding fault node; acquiring two measurement nodes adjacent to the fault node; when both of the two measurement nodes adjacent to the fault node are normal nodes, the two adjacent measurement nodes are set as the null nodes; in the specific implementation process, the historical data of the fault node may be inaccurate, and if the historical data of the node is used to fill the abnormal null data, the final result may not be accurate enough; the data detected by the adjacent measurement nodes has similarity; the calculation and filling of the adjacent measurement nodes are preferred for filling the abnormal null data, which can well reduce the error;

[0120] Step S502: when there is a fault node among the two measurement nodes adjacent to the fault node, one measurement node is further taken outwards from the fault node, until all the acquired measurement nodes are normal nodes, and the two acquired measurement nodes are set as the null nodes; in the specific implementation process, if multiple adjacent nodes fail at the same time, the normal nodes are sequentially taken outwards from the fault node until the normal node data is acquired;

[0121] Step S503: obtaining the node current threshold and the node voltage threshold of the two empty nodes; obtaining the average of the node current threshold of the two empty nodes and setting it as the empty current; obtaining the average of the node voltage threshold of the two empty nodes and setting it as the empty voltage; filling the empty current and the empty voltage into the abnormal empty data; in the specific implementation process, the voltage and the current among three adjacent nodes are basically consistent under normal circumstances, so the average of the node current threshold and the average of the node voltage threshold of the two empty nodes are less different from the normal data of the fault node;

[0122] Step S504: obtaining the missing empty data and the corresponding measurement node; obtaining the historical data of the measurement node; obtaining the node current threshold and the node voltage threshold of the measurement node; in the specific implementation process, the measurement node corresponding to the missing empty data is a normal node, when the missing empty data appears, it may be due to data loss, rather than the fault of the measurement node, the historical data of the measurement node still has certain reference value, and the missing empty data can be filled according to the historical data of the measurement node;

[0123] Step S505: filling the node current threshold and the node voltage threshold into the missing empty data.

[0124] Those skilled in the art will understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage media can be realized by any type of volatile or non-volatile storage devices or their combinations, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk. These computer program instructions can also be stored in a computer readable storage medium which can guide the computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction devices, which realize the flowFigure 1 one or more processes and / or functions specified in one or more blocks. Figure 1 one or more blocks.

[0125] The above-described embodiments are merely specific implementations of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the same. The protection scope of the present application is not limited to this. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that any modification or easy-to-think change or equivalent replacement of some technical features of the technical solutions recorded in the foregoing embodiments can be made within the technical range disclosed by the present application by any person skilled in the art. The modification, change or replacement does not make the corresponding technical solution deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An information anomaly identification fitting method suitable for wide-area measurement data quality improvement, characterized in that, Comprising the following steps: Step S1: A plurality of measurement nodes are set for the power distribution network, and the collection results of wide-area measurement data are obtained through the measurement nodes, the collection results including actual node voltage, actual node current, node current threshold, node voltage threshold, voltage fluctuation value and current fluctuation value; Step S2: The collection results of wide-area measurement data are compared and analyzed to obtain user features of abnormal data that are not successfully collected, and an abnormal data feature library is constructed; Step S3: Based on the abnormal data feature library, an information anomaly identification model for data promotion is constructed; Step S4: The information anomaly data identification model is used to detect abnormalities in the data, and missing or failed collection abnormal data is screened out; Step S5: The missing data is comprehensively fitted and filled according to the historical data of the measurement nodes; The step S3 comprises the following sub-steps: Step S301: A node topology network is constructed according to the topological relationship of the measurement nodes; Step S302: K-MEAN clustering analysis is performed on all abnormal data to obtain a plurality of abnormal data clusters; Step S303: The correlation between each abnormal data cluster and the fault result is analyzed, and the correlation priority between the abnormal data cluster and the abnormal result is configured according to the correlation; Step S304: The abnormal data clusters of each abnormal result are sorted according to the size of the correlation priority to obtain an abnormal sequence; Step S305: The corresponding weight correction value is extracted according to the user feature extraction, and the corresponding weight correction value is marked in the abnormal sequence; through the weight correction value, whether the same user feature exists is judged to determine the possibility of occurrence of abnormal events again; Step S306: The abnormal sequence is brought into the node topology network to generate an abnormal identification path; Step S307: The first abnormal identification item in the abnormal identification path is extracted as an identification trigger item, and the information anomaly identification model is a set of abnormal identification paths.

2. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 1, characterized in that, The step S1 comprises the following sub-steps: Step S101: A plurality of measurement nodes are configured based on the elements in the power distribution network, and the actual current and voltage of each measurement node are obtained, which are set as historical node data; Step S102: The historical node data of the measurement nodes are obtained, the historical average current of the nodes is calculated, and the node current threshold is set, the historical average voltage is calculated, and the node voltage threshold is set; Step S103: The preset current and voltage of the measurement nodes are obtained; The absolute value of the difference between the node current threshold and the preset current is set as the current fluctuation value, and the absolute value of the difference between the node voltage threshold and the preset voltage is set as the voltage fluctuation value.

3. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 2, characterized in that, The step S2 comprises the following sub-steps: Step S201: The collection results collected by the measurement nodes are obtained; the collection results are compared and analyzed to obtain data acquisition normal or data acquisition abnormal signals; Step S202: When data acquisition abnormality is detected, fault analysis is performed on the corresponding measurement nodes to obtain fault nodes, and the data of the fault nodes are set as abnormal data; Step S203: The user features corresponding to the abnormal data are obtained, the user features including user type and power consumption period, and an abnormal data feature library is established.

4. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 3, characterized in that, The comparison and analysis in step S201 comprises: Step S20101: Obtain a node current threshold value and a current fluctuation value, set a sum of the node current threshold value and the current fluctuation value as a first node current, and set a difference between the node current threshold value and the current fluctuation value as a second node current; Step S20102: Obtain a node real-time current, and determine that data acquisition is normal when the node real-time current is less than or equal to the first node current and greater than or equal to the second node current; Determine that data acquisition is abnormal when the node real-time current is greater than the first node current or the node real-time current is less than the second node current; Step S20103: Obtain a node voltage threshold value and a voltage fluctuation value, set a sum of the node voltage threshold value and the voltage fluctuation value as a first node voltage, and set a difference between the node voltage threshold value and the voltage fluctuation value as a second node voltage; Step S20104: Obtain a node real-time voltage, and determine that data acquisition is normal when the node real-time voltage is less than or equal to the first node voltage and greater than or equal to the second node voltage; Determine that data acquisition is abnormal when the node real-time voltage is greater than the first node voltage or the node real-time voltage is less than the second node voltage.

5. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 4, characterized in that, The fault analysis in the step S202 includes: When data acquisition is detected to be abnormal, performing fault analysis on a corresponding measurement node to obtain a fault node, and setting data of the fault node as abnormal data; Step S20201: Obtain a measurement node in which data is abnormal, set the measurement node as a to-be-detected node, and obtain two adjacent measurement nodes of the to-be-detected node, set the two adjacent measurement nodes as standard nodes; Step S20202: Compare and analyze collection results of the standard nodes; Step S20203: When collection results of the two standard nodes are both data acquisition abnormal, determine that the to-be-detected node is a normal node; when the collection results of the two standard nodes are both data acquisition normal, determine that the to-be-detected node is a fault node.

6. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 5, characterized in that, The step S4 includes: using an information abnormal data identification model to perform abnormal detection on real-time obtained node data, when a fault node is detected, setting voltage data and current data of the node as empty, and setting the voltage data and the current data as abnormal empty data, The step S4 further includes: setting all missing data as empty, setting the missing data as missing empty data, and obtaining measurement nodes corresponding to the missing empty data.

7. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 6, characterized in that, The step S5 includes the following sub-steps: Step S501: Obtain abnormal empty data and a corresponding fault node, obtain two adjacent measurement nodes of the fault node, and set the two adjacent measurement nodes as empty nodes when the two adjacent measurement nodes of the fault node are both normal nodes; Step S502: When there is a fault node in the two adjacent measurement nodes of the fault node, take one measurement node outward from the fault node again, until all obtained measurement nodes are normal nodes, and set the two obtained measurement nodes as empty nodes; Step S503: Obtain node current threshold values and node voltage threshold values of the two empty nodes, obtain an average value of the node current threshold values of the two empty nodes and set the average value as empty current, obtain an average value of the node voltage threshold values of the two empty nodes and set the average value as empty voltage, and fill the empty current and the empty voltage into the abnormal empty data. Step S504: acquire the missing null data and the corresponding measurement node; acquire the historical data of the measurement node; obtain the node current threshold and the node voltage threshold of the measurement node; Step S505: fill the node current threshold and the node voltage threshold into the missing null data.

8. The information anomaly identification fitting method suitable for wide-area measurement data quality improvement according to claim 1, characterized in that, The step S306 includes: Step S30601, taking the first abnormality identification item in the abnormality sequence as the pre-sequence node of the abnormality identification path; Step S30602, calculating the identification cost of other abnormal identification items in the abnormal sequence by a sequence identification algorithm P re = β1(r f - Δb n ) + β2Δb re + β3[R f - D(re, n) * d re ], wherein P re is the identification cost, β1is a preset current priority weight, β2is a preset correlation priority weight, β3is a preset cost priority weight, and β1+ β2+ β3= 1, r f is a preset reference identification value, Δb n is a correlation priority value of a previous abnormal identification item, Δb re is a correlation priority value of a current abnormal identification item, R f is a preset reference identification cost, D(re, n) is a node mapping function, reflecting a mapping relationship between a sequence number of the previous abnormal identification item in the abnormal sequence and a retrieval cost value corresponding to a sequence number of the current abnormal identification item in the abnormal sequence, and d re is an effective distance between a measurement node corresponding to the previous abnormal identification item and a measurement node corresponding to the current abnormal identification item. Step S30603, determining the abnormality identification item with the highest identification cost as the next abnormality identification item of the abnormality identification path, and returning to step S30602 until the abnormality identification path is constructed.

9. The information anomaly identification fitting method suitable for wide area measurement data quality improvement according to claim 8, characterized in that, Step S4 includes: Step S401, identifying the abnormal data of the measurement node according to the trigger identification item; Step S402, each abnormality identification path is configured with an index value, an index range is generated according to the index value, and an identification random number is generated, and the next abnormality identification item is determined according to the index range where the identification random number falls; Step S403, acquiring the identification content of the abnormality identification item corresponding to the next abnormality identification item and acquiring the corresponding measurement data to obtain the corresponding abnormality sub-value; Step S404, calculating an abnormal total value by a preset abnormal weighting algorithm, the abnormal weighting algorithm being wherein L d is the abnormal total value, u is a preset abnormal trigger reference value, K is a total number of current abnormal identification items, h m is a data abnormal value of the mth abnormal identification item, reflecting an abnormal degree of the measurement data corresponding to the abnormal identification item, a m is a preset identification item weight of the mth abnormal identification item; Step S405, if the total abnormality value is greater than the preset trigger upper limit value, output the abnormality result and enter step S4061, if the total abnormality value is less than the preset index lower limit value, enter step S4062, and if the total abnormality value is between the trigger upper limit value and the index lower limit value, return to step S402; Step S4061, configured with a gain factor to increase the index value of the corresponding abnormality identification item, and the gain factor decreases with time; Step S4062, configured with a decay factor to decrease the index value of the corresponding abnormality identification item, and the decay factor decreases with time.

Images