An abnormality disposal evaluation system for electric energy metering device based on multi-source information fusion

Abstract

The application relates to an electric energy metering device abnormality disposal evaluation system based on multi-source information fusion, which comprises a data acquisition module, a data processing module, a weight configuration module and an abnormality evaluation module. Through the setting, data is cleaned and unified in format, the subsequent processing of various data is guaranteed, the data abundance is improved through data processing, the data measurement dimension is unified, then the influence of different dimensions, different regions and different abnormal conditions on the evaluation result is adjusted according to actual conditions through weight configuration, so that the dimension and granularity of information analysis are guaranteed, the evaluation value is calculated in a dynamic weight and static weight combined mode, the corresponding defect elimination grade is obtained, and the sorting of abnormal work orders is realized instead of manual work.

Description

Technical Field

[0001] This invention relates to the field of power grid information processing, and more specifically, to an evaluation system for handling abnormalities in power metering devices based on multi-source information fusion. Background Technology

[0002] With the accelerating pace of China's power market reform and the deepening of the electricity spot market, the data requirements for power user information collection systems are becoming increasingly stringent. There is an urgent need to enhance the monitoring capabilities of metering devices to ensure their safe, stable, and reliable operation. This necessitates in-depth data mining, utilizing big data analytics to establish hierarchical and categorized diagnostic and handling rules for metering anomalies. This will enable rapid resolution of metering anomalies for market-oriented and critical users, preventing equipment from exceeding tolerances and comprehensively improving the health of metering devices. This will provide strong technical support for power market reform and the construction of a new power system. Statistics show that the current system experiences hundreds of thousands of clock deviation anomalies and tens of thousands of voltage anomalies (including voltage phase loss, fixed voltage values, high voltage, low voltage, and voltage imbalance) per month. However, the current system lacks a priority evaluation system for work orders, relying on manual work order resolution. Furthermore, due to the large number of work orders, some highly urgent orders cannot be processed promptly. This situation arises from an imperfect work order mechanism. In processing work orders, frontline staff face challenges such as difficulties in cross-professional and cross-system on-site workflow, an excessive volume of manually entered feedback work orders, duplicate assignments, and frequent false alarms. Furthermore, the current lack of an evaluation system for the urgency of work order assignments makes it difficult for frontline staff to accurately determine the priority of anomaly resolution. Therefore, it is necessary to improve the anomaly work order integration mechanism, implement equipment health assessment, and generate anomaly resolution urgency recommendations based on the assessment results. Currently, there are a large number of anomaly work orders, and due to the lack of priority, the large number of anomaly resolution work orders at the frontline level means that there are no corresponding personnel to handle these anomaly work orders. Simultaneously, some anomalies affect users' normal power consumption; without prioritization, this will lead to abnormal power consumption or billing for users, and also affect indicators such as transformer area line loss and line loss. To address these issues, patent announcement number CN 112527778A discloses an anomaly resolution management system and method based on incremental anomaly information database. This system provides work order identification and processing services through a primary and secondary work order filtering department, allowing for different processing methods for work orders with different situations. (Announcement number CN...) Patent 110991784A discloses a method and system for dispatching work orders in abnormally low-voltage distribution areas. By performing linear programming on historical work order data, it performs regional dispatching on unassigned work order data, achieving optimal allocation of work orders and maintenance personnel, thus improving the efficiency of maintenance personnel. However, the above two methods do not offer a good solution for work order processing strategies. Because the sequence of work orders requires consideration of a lot of information, and the types of information presentation, the access ports for information acquisition are inconsistent, and the language habits used are different, the external conditions for screening and evaluating work orders were not feasible in the past. However, with the improvement of data acquisition and the upper limit of smart metering devices, as well as the popularization of "online power grids", a large amount of information can be presented in the form of electronic data, thus providing the possibility for intelligent work order evaluation. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide an evaluation system for abnormal handling of power metering devices based on multi-source information fusion.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is: an evaluation system for abnormal handling of power metering devices based on multi-source information fusion, including a data acquisition module, a data processing module, a weight configuration module and an abnormal evaluation module;

[0005] The data acquisition module is used to acquire user electricity consumption information corresponding to abnormal work orders and to standardize the format of the acquired user electricity consumption information.

[0006] The data processing module has several preset feature items, each feature item is configured with a data quantization marker. The data processing module obtains the corresponding abnormal feature value based on the feature items that identify user electricity information and the content of the feature items through the data quantization marker.

[0007] The weight configuration module generates a dynamic weight value for each feature item based on the feature items of the user's electricity consumption information;

[0008] The anomaly evaluation module obtains anomaly evaluation values ​​based on the anomaly feature values ​​of the feature items weighted by dynamic weight values, and generates corresponding defect elimination levels based on the magnitude of the anomaly evaluation values.

[0009] Furthermore: the data processing module includes a discretization processing unit and a normalization processing unit. When the data quantization marker is a discrete marker, the feature item is processed by the discretization processing unit; when the data quantization marker is a normalization marker, the feature item is processed by the normalization processing unit.

[0010] The discretization processing unit is configured with discrete conditions corresponding to different feature items. Each discrete condition corresponds to an abnormal feature value. When the content of the feature item satisfies the discrete condition, the corresponding abnormal feature value is obtained.

[0011] The normalization processing unit is configured with normalization mapping relationships corresponding to different feature items, and processes the content of the feature items into corresponding abnormal feature values ​​according to the normalization mapping relationships.

[0012] Furthermore, the normalization mapping relationship includes a function mapping relationship, a degree mapping relationship, and an order mapping relationship. The function mapping relationship is configured with different mapping functions for different feature items. The content of the feature item is used as the input of the function to obtain the abnormal feature value. The degree mapping relationship is configured with several different degree thresholds for different feature items. The degree threshold closest to the content of the feature item is used as the corresponding abnormal feature value. The order mapping relationship is configured with several order feature values. Each order feature value corresponds to one or more feature item orders. The feature items are sorted according to the size of their content to determine the feature item order of each feature item. The order feature value corresponding to the feature item order is used as the corresponding abnormal feature value.

[0013] Further: The data processing module includes a data compensation unit, which is used to identify missing data corresponding to feature items and is configured with a missing data compensation algorithm to compensate for the missing data. The missing data compensation algorithm is... ,in, For the missing data, For the first The similarity weights corresponding to the missing relevant information For the missing inference function of feature terms corresponding to missing relevant information, For the first The missing correlation value corresponding to each missing related information, This represents the total number of missing relevant information.

[0014] Furthermore: the weight configuration module includes a big data configuration unit, a consistency configuration unit, and a correlation configuration unit. The big data configuration unit is used to generate big data weights, the consistency configuration unit is used to generate consistency weights, and the correlation configuration unit is used to generate correlation weights. For any dynamic weight value, there are... ,in For the first The dynamic weight values ​​corresponding to each feature term For the first The big data weights corresponding to each feature item For the first The relevance weights corresponding to each feature term For the first Consistency weights corresponding to each feature term.

[0015] Furthermore: the big data configuration unit is used to call a historical anomaly information database, which stores a number of historical anomaly information entries. These entries include anomaly feature values ​​corresponding to several feature items and anomaly evaluation results. The big data configuration unit is pre-configured with an evaluation result table, which stores several historical evaluation values. These historical evaluation values ​​are indexed by the anomaly evaluation results. The big data configuration unit matches the corresponding historical evaluation value based on the anomaly feature value of the feature item. ,in, For the first The big data sub-weights corresponding to each feature term For the first The abnormal feature values ​​corresponding to each feature term For the first matching relationship There are 10 abnormal feature values. For the first matching relationship The historical evaluation values ​​corresponding to each abnormal feature value The number of abnormal feature values ​​with matching relationships, if satisfying , If the preset matching difference benchmark is used, then the abnormal feature value is considered to be related to the first... The feature items have an abnormal matching relationship; the big data configuration unit calculates the weights of the big data through a weight balancing algorithm, the weight balancing algorithm being... ,in, As a preset big data trade-off value, This refers to the total number of big data weights generated by the user through the big data configuration unit.

[0016] Furthermore: the relevance configuration unit is pre-configured with several feature associations and a corresponding association weight sub-value for each feature association. When the abnormal feature values ​​corresponding to any two feature items satisfy the feature association relationship, the corresponding weight association sub-value is obtained. After filtering all feature associations, all weight association sub-values ​​for each feature item are obtained. The relevance configuration unit calculates the relevance weights using a weight balancing algorithm. The weight balancing algorithm is as follows: ,in, For the first The weight associated sub-values ​​of each feature term The pre-defined relevance trade-off value. The total number of correlation weights generated for this user through the correlation configuration unit.

[0017] Further: The consistency configuration unit is configured with consistency conditions, and each consistency condition corresponds to a consistency benchmark value for different feature items. The consistency conditions are matched to all feature items in the user's electricity consumption information, and the consistency deviation of each feature item is calculated based on the consistency benchmark value corresponding to the consistency condition. A weighted consistency sub-value is obtained for each feature item based on the consistency deviation. The weighted consistency sub-value is positively correlated with the consistency deviation. The consistency configuration unit calculates the consistency weight value using a weight balancing algorithm. The weight balancing algorithm is... ,in, For the first The weights of each feature term are consistent sub-values. The preset consistency trade-off value, This represents the total number of consistency weights generated by the user through the consistency configuration unit.

[0018] Furthermore: the consistency configuration unit is configured with a discrete mean algorithm for calculating the weighted consistency sub-value, the discrete mean algorithm being... ,in, For consistency deviation, The mean deviation is the average of the consistency deviations of this feature item in the electricity consumption information of different users. The standard deviation is the standard deviation of the consistency deviation of this feature item of electricity consumption information of different users.

[0019] Furthermore: the anomaly evaluation module is configured with an anomaly evaluation algorithm for calculating anomaly evaluation values, the anomaly evaluation algorithm being... ,in, This is an abnormal evaluation value. For the first The abnormal feature values ​​corresponding to each feature term For the preset first The solid weight values ​​corresponding to each feature term This represents the number of feature items in the user's electricity consumption information.

[0020] The main technical effects of this invention are reflected in the following aspects: By setting it up in this way, the data is cleaned and formatted in a unified manner, ensuring the subsequent processing of various types of data. At the same time, through data processing, the data abundance is improved and the data measurement dimensions are unified. Then, through weight configuration, the influence of different dimensions, different regions, and different anomalies on the evaluation results is adjusted according to the actual situation, thereby ensuring the dimension and granularity of information analysis. At the same time, the evaluation value is calculated by combining dynamic weights and static weights, thereby obtaining the corresponding defect elimination level, thus replacing manual sorting of abnormal work orders. Attached Figure Description

[0021] Figure 1: System architecture diagram of this invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, so that the technical solution of the present invention can be more easily understood and mastered.

[0023] An evaluation system for handling anomalies in power metering devices based on multi-source information fusion includes a data acquisition module, a data processing module, a weight configuration module, and an anomaly evaluation module. By integrating work orders and evaluating equipment anomaly dispatch, the system prioritizes and categorizes abnormal work orders, thereby improving the efficiency of work order troubleshooting, the work efficiency of frontline staff, and the quality of customer service.

[0024] The data acquisition module is used to obtain user electricity consumption information corresponding to abnormal work orders and to standardize the format of the obtained user electricity consumption information. First, data acquisition includes the following: the quality of basic work order data is a prerequisite for the calculation of the "work order fusion and equipment abnormal dispatch evaluation model". A basic data acquisition module is developed, and data cleaning rules are established according to the characteristics of various data types such as archive data, collected data, and abnormal data. The basic work order data obtained through the interface is processed, and data that does not meet the basic data quality requirements is removed to ensure data accuracy. The basic data table includes: daily electricity consumption data of low-voltage users, daily line loss data of transformer areas, low-voltage abnormal work order dataset, public transformer abnormal work order dataset, dedicated transformer abnormal work order dataset, electricity user files, dedicated transformer measurement point files, low-voltage measurement point files, etc. Then, the information that can be obtained includes several feature items specifically for each work order formed based on the above content, from the perspective of the user participating in the calculation: collection point number, user number, table asset number, abnormality type, abnormality score, importance score, electricity consumption score, and comprehensive score. From the meter granular calculation dimension: meter asset number, collection point number, user number, transformer area number, anomaly type, anomaly score, whether there is a public transformer, electricity consumption percentage score, and overall score. From the transformer area granular calculation dimension: transformer area number, anomaly type, anomaly score, number of anomaly meters, collection success rate, line loss rate, electricity consumption percentage score, whether there is a public transformer, and overall score. These different dimensions generate different feature items for the same form. Some feature items are numbers, some are text, some are proportions, and some are numerical values. Therefore, for ease of measurement, they are uniformly processed using numerical values, requiring the intervention of a data processing module.

[0025] The data processing module has several preset feature items, each feature item is configured with a data quantization marker. The data processing module obtains the corresponding abnormal feature value based on the feature items that identify user electricity information and the content of the feature items through the data quantization marker.

[0026] The data processing module includes a discretization processing unit and a normalization processing unit. When the data quantization marker is a discrete marker, the feature item is processed by the discretization processing unit; when the data quantization marker is a normalized marker, the feature item is processed by the normalization processing unit. The normalization processing unit is configured with normalization mapping relationships corresponding to different feature items, and processes the content of the feature item into the corresponding abnormal feature value according to the normalization mapping relationship. Different features have different value ranges. In some algorithms, such as linear models or distance-related models like clustering models and KNN models, the value range of features will have a significant impact on the final result. For example, the value range of binary features is [0, 1], while the value of distance features may be [0, positive infinity). In practical use, the distance is truncated, for example, [0, 3000000]. However, because these two features have inconsistent value ranges, the model may be more biased towards the feature with the larger value range. In order to balance the features with inconsistent value ranges, it is necessary to perform normalization processing to normalize the feature values ​​to the interval [0, 1]. The normalization mapping relationships include function mapping relationships, gradation mapping relationships, and ranking mapping relationships. The function mapping relationship has different mapping functions configured for different feature items. The content of the feature item is used as the input to the function to obtain the abnormal feature value. For example, function normalization maps feature values ​​to the [0, 1] interval through the mapping function. For instance, the maximum-minimum normalization method is a linear mapping. There are also mappings using non-linear functions, such as the log function. The gradation mapping relationship has several different gradation thresholds configured for different feature items. The gradation threshold closest to the feature item's content is used as the corresponding abnormal feature value. For example, dimensionality normalization can use the maximum-minimum normalization method, but the maximum and minimum values ​​selected are the maximum and minimum values ​​of the corresponding category, i.e., local maximum and minimum values ​​are used, not global maximum and minimum values. The ranking mapping relationship has several ranking feature values. Each ranking feature value corresponds to one or more feature item orders. The feature item order is determined by sorting the feature items according to their content size, and the ranking feature value corresponding to the feature item order is used as the corresponding abnormal feature value. For example, sorting normalization sorts features by size regardless of their original values ​​and assigns a new value based on the order of the features.

[0027] The discretization processing unit is configured with discrete conditions corresponding to different feature terms. Each discrete condition corresponds to an abnormal feature value. When the content of the feature term satisfies the discrete condition, the corresponding abnormal feature value is obtained. For example, as mentioned above, the value space of continuous values ​​may be infinite. In order to facilitate representation and processing in the model, continuous value features need to be discretized. Commonly used discretization methods include equal-value partitioning and equal-quantity partitioning. Equal-value partitioning divides the feature equally according to the value range, and the values ​​within each segment are treated equally. For example, if the value range of a certain feature is [0, 10], we can divide it into 10 segments: [0, 1), [1, 2), ..., [9, 10). Equal-quantity partitioning divides the feature equally according to the total number of samples, and each segment is divided into 1 segment with an equal number of samples. For example, the distance feature has a value range of [0, 3000000], and now it needs to be divided into 10 segments. If it is divided proportionally, we will find that most of the samples are in the first segment. Using equal partitioning avoids this problem. The possible final partitions are [0, 100), [100, 300), [300, 500), ..., [10000, 3000000]. The first intervals are denser, while the latter are sparser. The discrete conditions can be based on whether the value falls within the above range, or on the characteristics of a specific number or text. On the other hand, the data processing module can also quantify the data format by assigning values ​​to text information. For example, for anomaly types such as no data on the concentrator meter, no meter reading for several consecutive days, meter running backwards, meter stopped, meter running out of power, uneven reading, reverse power abnormality, reverse reading, low voltage, severe low voltage, voltage phase loss, battery undervoltage, terminal clock error, energy meter clock error, power differential abnormality, meter disconnection, etc., the module can then pre-assign anomaly values ​​to the text information based on the corresponding data values ​​to obtain anomaly characteristic values.

[0028] The data processing module includes a data compensation unit, which is used to identify missing data corresponding to feature items and is configured with a missing data compensation algorithm to compensate for the missing data. The missing data compensation algorithm is as follows: ,in, For the missing data, For the first The similarity weights corresponding to each missing relevant information For the missing inference function of feature terms corresponding to missing relevant information, For the first The missing correlation value corresponding to each missing related information, This represents the total number of missing relevant information. Data cleaning is performed based on missing data by filling in missing values, smoothing noisy data, identifying or deleting outliers, and resolving inconsistencies. The main objectives are: format standardization, removal of abnormal data, error correction, and removal of duplicate data. For example, if user electricity consumption data is missing, it can be filled with the mean and median based on historical user electricity consumption; if line loss data collection for transformer areas fails, it can be filled with the mean and median of historical line loss rates. This could also include missing metering data, which can be corrected by comparing historical data. Similarity relationships and corresponding missing data inference functions are calculated. For example, the electricity consumption of a transformer area can be inferred from the electricity consumption of each user (considering line loss). The similarity weight can represent the similarity between the current situation and past situations. For example, if the time characteristics or waveform characteristics are similar, the corresponding similarity weight can be determined based on this degree of approximation, and then the theoretical missing value can be obtained based on the graph function.

[0029] The weight configuration module generates a dynamic weight value for each feature item based on the user's electricity consumption information. The weight configuration module includes a big data configuration unit, a consistency configuration unit, and a correlation configuration unit. The big data configuration unit generates big data weights and calls a historical anomaly information database. This database stores several historical anomaly information items, including anomaly feature values ​​corresponding to several feature items and anomaly evaluation results. The big data configuration unit pre-configures an evaluation result table, which stores several historical evaluation values ​​indexed by the anomaly evaluation results. The big data configuration unit matches the corresponding historical evaluation value based on the anomaly feature value of the feature item. ,in, For the first The big data sub-weights corresponding to each feature term For the first The abnormal feature values ​​corresponding to each feature term For the first matching relationship There are 10 abnormal feature values. For the first matching relationship The historical evaluation values ​​corresponding to each abnormal feature value The number of abnormal feature values ​​with matching relationships, if satisfying , If the preset matching difference benchmark is used, then the abnormal feature value is considered to be related to the first... The feature items have an abnormal matching relationship; the big data configuration unit calculates the weights of the big data through a weight balancing algorithm, the weight balancing algorithm being... ,in, As a preset big data trade-off value, This refers to the total number of big data weights generated by the big data configuration unit for this user. The risk of this feature is assessed based on the results of similar features from other users. Historically, if a certain content appears in an abnormal work order, it often results in complete machine failure. Therefore, the corresponding difference is determined using the above algorithm, and then the weighting influence of historical data on this data is calculated.

[0030] The consistency configuration unit is used to generate consistency weights. The consistency configuration unit is configured with consistency conditions, and each consistency condition corresponds to a consistency benchmark value for different feature items. All feature items in the user's electricity consumption information are matched against the corresponding consistency conditions, and the consistency deviation of each feature item is calculated based on the consistency benchmark value corresponding to the consistency condition. A weighted consistency sub-value for each feature item is obtained based on the consistency deviation. The weighted consistency sub-value is positively correlated with the consistency deviation. The consistency configuration unit calculates the consistency weights using a weight balancing algorithm. The weight balancing algorithm is... ,in, For the first The weights of each feature term are consistent sub-values. The preset consistency trade-off value, This refers to the total number of consistency weights generated by the user through the consistency configuration unit. The consistency configuration unit is configured with a discrete mean algorithm to calculate the weighted consistency sub-values. The discrete mean algorithm is... ,in, For consistency deviation, The mean deviation is the average of the consistency deviations of this feature item in the electricity consumption information of different users. The standard deviation is the standard deviation of the consistency deviation of this feature item in the electricity consumption information of different users. The consistency deviation reflects the consistency of multiple different feature items. If the consistency deviation is large, it means that the deviation of multiple different feature items of the same user is large, so the anomaly is more serious. Conversely, the anomaly is more predictable. By calculating the dispersion and deviation, the consistency sub-value of each weight can be calculated, thereby calculating the corresponding weight.

[0031] The correlation configuration unit is used to generate correlation weights. The unit pre-configures several feature associations and corresponding correlation weight sub-values ​​for each feature association. When any two feature items' corresponding abnormal feature values ​​satisfy a feature association relationship, the corresponding weighted correlation sub-values ​​are obtained. After filtering all feature associations, all weighted correlation sub-values ​​for each feature item are obtained. The correlation configuration unit calculates the correlation weights using a weight balancing algorithm. ,in, For the first The weight associated sub-values ​​of each feature term The pre-defined relevance trade-off value. This represents the total number of correlation weights generated for this user through the correlation configuration unit. Correlation indicates the relationship between anomalous features. If two feature items simultaneously show corresponding anomalous data, the weight configuration needs to be increased, affecting the calculation result of the defect elimination level.

[0032] For any dynamic weight value, we have ,in For the first The dynamic weight values ​​corresponding to each feature term For the first The big data weights corresponding to each feature item For the first The relevance weights corresponding to each feature term For the first Each feature item has a corresponding consistency weight. By setting dynamic weight values, values ​​can be dynamically assigned based on the actual content of the feature items and historically generated data, allowing the evaluation results to be corrected according to the actual situation, thereby improving accuracy and judgment ability.

[0033] The anomaly evaluation module obtains anomaly evaluation values ​​based on the anomaly feature values ​​weighted by dynamic weight values, and generates corresponding defect elimination levels based on the magnitude of the anomaly evaluation values. The anomaly evaluation module is configured with an anomaly evaluation algorithm for calculating the anomaly evaluation values. The anomaly evaluation algorithm is as follows: ,in, This is an abnormal evaluation value. For the first The abnormal feature values ​​corresponding to each feature term For the preset first The solid weight values ​​corresponding to each feature term This represents the number of feature items in the user's electricity consumption information. By calculating the anomaly evaluation value, a corresponding defect elimination level is generated. Based on the defect elimination level, the urgency of defect elimination in the work order is determined, thus replacing manual judgment.

[0034] Of course, the above are just typical examples of the present invention. In addition, the present invention may have many other specific embodiments. All technical solutions formed by equivalent substitution or equivalent transformation fall within the scope of protection claimed by the present invention.

Claims

1. A fault handling and evaluation system for power metering devices based on multi-source information fusion, characterized in that: It includes a data acquisition module, a data processing module, a weight configuration module, and an anomaly evaluation module; The data acquisition module is used to acquire user electricity consumption information corresponding to abnormal work orders and to standardize the format of the acquired user electricity consumption information. The data processing module has several preset feature items, each feature item is configured with a data quantization marker. The data processing module obtains the corresponding abnormal feature value based on the feature items that identify user electricity information and the content of the feature items through the data quantization marker. The weight configuration module generates a dynamic weight value for each feature item based on the feature items of the user's electricity consumption information; The anomaly evaluation module obtains anomaly evaluation values ​​based on the anomaly feature values ​​of the feature items weighted by dynamic weight values, and generates corresponding defect elimination levels based on the magnitude of the anomaly evaluation values. The data processing module includes a data compensation unit, which is used to identify missing data corresponding to feature items and is configured with a missing data compensation algorithm to compensate for the missing data. The missing data compensation algorithm is as follows: ,in, For the missing data, For the first The similarity weights corresponding to each missing relevant information For the missing inference function of feature terms corresponding to missing relevant information, For the first The missing correlation value corresponding to each missing related information, This represents the total number of missing relevant information.

2. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 1, characterized in that: The data processing module includes a discretization processing unit and a normalization processing unit. When the data quantization marker is a discrete marker, the feature item is processed by the discretization processing unit; when the data quantization marker is a normalization marker, the feature item is processed by the normalization processing unit. The discretization processing unit is configured with discrete conditions corresponding to different feature items. Each discrete condition corresponds to an abnormal feature value. When the content of the feature item satisfies the discrete condition, the corresponding abnormal feature value is obtained. The normalization processing unit is configured with normalization mapping relationships corresponding to different feature items, and processes the content of the feature items into corresponding abnormal feature values ​​according to the normalization mapping relationships.

3. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 2, characterized in that: The normalization mapping relationship includes a function mapping relationship, a gradation mapping relationship, and a sorting mapping relationship. The function mapping relationship is configured with different mapping functions for different feature items. The content of the feature item is used as the input of the function to obtain the abnormal feature value. The gradation mapping relationship is configured with several different gradation thresholds for different feature items. The gradation threshold closest to the content of the feature item is used as the corresponding abnormal feature value. The sorting mapping relationship is configured with several sorting feature values. Each sorting feature value corresponds to one or more feature item orders. The feature item order is determined by sorting according to the size of the feature item content, and the sorting feature value corresponding to the feature item order is used as the corresponding abnormal feature value.

4. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 1, characterized in that: The weight configuration module includes a big data configuration unit, a consistency configuration unit, and a correlation configuration unit. The big data configuration unit generates big data weights, the consistency configuration unit generates consistency weights, and the correlation configuration unit generates correlation weights. For any dynamic weight value, there is... ,in For the first The dynamic weight values ​​corresponding to each feature term For the first The big data weights corresponding to each feature item For the first The relevance weights corresponding to each feature term For the first Consistency weights corresponding to each feature term.

5. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 4, characterized in that: The big data configuration unit is used to call a historical anomaly information database. This database stores a number of historical anomaly information entries, including anomaly feature values ​​corresponding to several feature items and anomaly evaluation results. The big data configuration unit is pre-configured with an evaluation result table, which stores several historical evaluation values ​​indexed by the anomaly evaluation results. The big data configuration unit matches the corresponding historical evaluation value based on the anomaly feature value of the feature item. ,in, For the first The big data sub-weights corresponding to each feature term For the first The abnormal feature values ​​corresponding to each feature term For the first matching relationship There are 10 abnormal feature values. For the first matching relationship The historical evaluation values ​​corresponding to each abnormal feature value The number of abnormal feature values ​​with matching relationships, if satisfying , If the preset matching difference benchmark is used, then the abnormal feature value is considered to be related to the first... The feature items have an abnormal matching relationship; the big data configuration unit calculates the weights of the big data through a weight balancing algorithm, the weight balancing algorithm being... ,in, As a preset big data trade-off value, This refers to the total number of big data weights generated by the user through the big data configuration unit.

6. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 4, characterized in that: The relevance configuration unit pre-configures several feature associations and corresponding association weight sub-values ​​for each feature association. When any two feature items' corresponding abnormal feature values ​​satisfy a feature association relationship, the corresponding weight association sub-value is obtained. After filtering all feature associations, all weight association sub-values ​​for each feature item are obtained. The relevance configuration unit calculates the relevance weights using a weight balancing algorithm. The weight balancing algorithm is... ,in, For the first The weight associated sub-values ​​of each feature term The pre-defined relevance trade-off value. The total number of correlation weights generated for this user through the correlation configuration unit.

7. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 4, characterized in that: The consistency configuration unit is configured with consistency conditions. Each consistency condition corresponds to a consistency benchmark value for different feature items. The consistency conditions are matched against all feature items in the user's electricity consumption information. The consistency deviation for each feature item is calculated based on the consistency benchmark value corresponding to the consistency condition. A weighted consistency sub-value for each feature item is obtained based on the consistency deviation. This weighted consistency sub-value is positively correlated with the consistency deviation. The consistency configuration unit calculates the consistency weight using a weight balancing algorithm. ,in, For the first The weights of each feature term are consistent sub-values. The preset consistency trade-off value, This represents the total number of consistency weights generated by the user through the consistency configuration unit.

8. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 4, characterized in that: The consistency configuration unit is configured with a discrete mean algorithm for calculating the weighted consistency sub-value. The discrete mean algorithm is as follows: ,in, For consistency deviation, The mean deviation is the average of the consistency deviations of this feature item in the electricity consumption information of different users. The standard deviation is the standard deviation of the consistency deviation of this feature item of electricity consumption information of different users.

9. The power metering device anomaly handling and evaluation system based on multi-source information fusion as described in claim 4, characterized in that: The anomaly evaluation module is configured with an anomaly evaluation algorithm to calculate the anomaly evaluation value. The anomaly evaluation algorithm is as follows: ,in, This is an abnormal evaluation value. For the first The abnormal feature values ​​corresponding to each feature term For the preset first The solid weight values ​​corresponding to each feature term This represents the number of feature items in the user's electricity consumption information.

Images