A data-driven and dynamic threshold-based power consumption anomaly diagnosis system and method

Abstract

The application provides a data-driven and dynamic threshold power consumption anomaly diagnosis system and method, and belongs to the technical field of data analysis, and specifically comprises: determining the change of the power consumption load characteristics of different power users based on a recognition processing method, determining the dynamic updating of the power consumption threshold of the feature recognition analysis user based on the recognition processing data and the recognition processing method of the power consumption load characteristics of different power users, determining the updating method of the feature recognition analysis user based on the recognition processing data of the feature recognition analysis user and the change of the power consumption load characteristics of the power user, determining the early warning processing target of the power user based on the updating data of the feature recognition analysis user and the change of the power consumption load characteristics of the power user, and improving the accuracy of the power consumption anomaly diagnosis result.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a data-driven and dynamic threshold-based system and method for diagnosing abnormal electricity consumption. Background Technology

[0002] Traditional electricity consumption anomaly diagnosis systems often employ fixed threshold methods (e.g., setting a threshold of 30% above the monthly average for anomaly) or simple rule bases (e.g., triggering an alarm if nighttime electricity consumption is higher than daytime consumption). These methods ignore the differences in electricity consumption behavior among different users—residential users and industrial users have drastically different electricity consumption patterns, and the same user's electricity consumption patterns also fluctuate significantly across different seasons, weekdays, and holidays. Existing technologies struggle to distinguish between "genuine electricity consumption anomalies (e.g., electricity theft, leakage, equipment malfunction)" and "normal changes in user electricity consumption behavior (e.g., adding new appliances, seasonal air conditioning use)," resulting in high false alarm rates and requiring extensive manual verification. This makes them unsuitable for the efficient processing needs of large-scale smart meter data.

[0003] To address the aforementioned technical problems, the invention patent application CN202511313146.6, "Method and System for Anomaly Detection of User Electricity Bill Data," establishes a system based on historical electricity consumption data and meteorological data. This system utilizes a seasonally sensitive baseline model to automatically learn the electricity consumption fluctuation patterns of a specific season in a region. Furthermore, a seasonal adjustment factor is introduced into the seasonally sensitive baseline model, and deviation correction is performed using temperature data. Through year-on-year and month-on-month anomaly detection calculations, abnormal changes in electricity consumption can be detected in a timely manner, determining whether there is abnormal electricity consumption behavior or a power system fault. However, it suffers from the following technical shortcomings: When users diagnose and process electricity anomalies based on their historical electricity consumption data, it is inevitable that early warnings will be delayed or inaccurate. Therefore, identifying and processing the electricity load characteristics of different users, and determining different early warning processing methods based on the identification and processing results, in order to improve the reliability and timeliness of early warning processing, has become an urgent technical problem to be solved.

[0004] Therefore, there is an urgent need for a data-driven and dynamic threshold-based system and method for diagnosing abnormal electricity consumption. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides a data-driven and dynamic threshold-based method for diagnosing abnormal electricity consumption, which includes: S1 uses the monitoring data of the power user to determine the benchmark electricity consumption of the power user, and updates the electricity consumption threshold of the power user based on the deviation between the power user's electricity consumption and the benchmark electricity consumption. S2 determines the electricity consumption variation data of the electricity user, determines the electricity consumption stability type of the electricity user based on the electricity consumption variation data of different electricity users, and determines the identification and processing method of the electricity load characteristics of the electricity user according to the electricity user data and the different electricity consumption stability types of the electricity user. S3 determines the changes in the electricity load characteristics of different power users based on the identification and processing method, and determines the update processing method for the feature identification and analysis users among the power users based on the identification and processing data of the electricity load characteristics of different power users and the identification and processing method. S4 performs dynamic updating of the electricity consumption threshold of the user identified by the feature analysis, and determines the early warning processing target of the user based on the updated data of the user identified by the feature analysis and the changes in the electricity load characteristics of the user.

[0006] Furthermore, the monitoring data includes the electricity consumption of the power user in different monitoring periods.

[0007] Furthermore, the baseline electricity consumption of the electricity user is determined based on the average electricity consumption on different dates.

[0008] Furthermore, the process of updating the electricity consumption threshold for the aforementioned electricity users specifically includes: Based on the electricity consumption data of electricity users on different dates, the mean and standard deviation are calculated and used as the benchmark electricity consumption and benchmark standard deviation for the electricity users. The daily electricity consumption is compared with the baseline average. If the absolute value of the deviation between the electricity consumption and the baseline electricity consumption is a preset multiple of the standard deviation, then the date is placed in the abnormal electricity consumption array. The percentage of abnormal electricity consumption in the total number of reported dates is taken as the percentage of abnormal dates. If the percentage of abnormal dates is within the preset range, the electricity consumption in the abnormal electricity consumption array within the most recent preset time period is sorted to obtain the fluctuation range of the electricity consumption threshold for the power user. If the percentage of abnormal dates is not within the preset range, the abnormal electricity consumption situation of the power user is determined according to the electricity consumption abnormality early warning method. When no early warning is issued, the fluctuation range of the electricity consumption threshold for the power user is still updated.

[0009] Understandably, based on the deviation from the benchmark electricity consumption, the abnormal electricity consumption arrays for a specified number of days above the benchmark electricity consumption are sorted from smallest to largest, and the abnormal electricity consumption arrays for a specified number of days below the benchmark electricity consumption are selected as the fluctuation range of the electricity consumption threshold for the electricity user.

[0010] Furthermore, the electricity consumption variation data of the electricity user is determined based on the rate of change of the electricity consumption of the electricity user between different monitoring periods.

[0011] Furthermore, the method for determining the electricity consumption stability type of the electricity user is as follows: Based on the electricity consumption data of the electricity users, determine the rate of change of electricity consumption of the electricity users between different monitoring periods; The electricity consumption change rate of the power user is determined based on the average of the change rates of electricity consumption between different monitoring periods. The electricity consumption stability type of the electricity user is determined based on the electricity consumption change rate of the electricity user.

[0012] Furthermore, the method for determining the update processing method for the feature identification and analysis of electricity users is as follows: Based on the identification and processing data of different power users' electricity load characteristics, power users with identification and processing data of electricity load characteristics are identified and treated as identification and processing users. Based on the power user identification and processing method, power users of the relaxed identification method are identified among the power users, and the power users of the relaxed identification method are designated as relaxed identification power users. The update processing method for feature identification analysis users among the power users is determined by considering the number of identification processes for different identification processing users and the loosely identified power users among the power users.

[0013] Secondly, this application provides a data-driven and dynamic threshold-based power consumption anomaly diagnosis system, employing the aforementioned data-driven and dynamic threshold-based power consumption anomaly diagnosis method, specifically including: Threshold update module, identification processing module, early warning processing module; The threshold update module updates the electricity consumption threshold of the power user. The identification processing module is responsible for determining the identification processing method for the electricity load characteristics of the power users; The early warning processing module is responsible for determining the early warning processing targets for the power users.

[0014] The beneficial effects of this invention are as follows: Based on the deviation between electricity users' electricity consumption and the benchmark electricity consumption, the electricity consumption threshold for electricity users is updated. According to the changes in electricity users' electricity consumption, the electricity consumption threshold is dynamically updated. This fully takes into account the fact that the electricity consumption of different electricity users is prone to fluctuations, avoids the situation where the early warning results are not accurate due to the use of fixed thresholds, and also lays the foundation for targeted early warning processing for different electricity users.

[0015] Based on the power user data and the different power consumption stability types of different power users, a method for identifying and processing the power load characteristics of the power users is determined. Specifically, based on the number of power users and the different power consumption stability types of different power users, differentiated power load characteristic identification and processing methods are generated for different power users. This not only avoids the technical problem of excessive data processing difficulty caused by frequent power load characteristic identification and processing, but also takes into account the changes in power load and fully considers the identification and processing difficulty under different power load characteristic identification and processing methods. Based on the identification and processing difficulty and the power consumption stability type of the power users, different power load identification and processing methods are determined for different power users, ensuring that power users can have their power load characteristics identified and processed in a timely and effective manner when there are abnormalities in power consumption.

[0016] Based on feature identification analysis of user update data and changes in the electricity load characteristics of the power users, the early warning processing targets for the power users are determined. Specifically, based on the feature identification analysis of user update data, an overall assessment of the reliability of the current early warning processing is conducted. Furthermore, combined with the changes in the electricity load of the power users, an overall assessment of the drastic change in the overall electricity load of the power users is conducted. This achieves the determination of the early warning method for abnormal electricity consumption of power users from the perspectives of the reliability of the current early warning processing and the drastic change, which ensures the reliability and timeliness of the early warning, while also avoiding the technical problems of excessively high user processing difficulty caused by frequent early warnings.

[0017] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0019] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart of a data-driven and dynamic threshold-based method for diagnosing abnormal electricity consumption. Figure 2 This is a flowchart for updating the electricity consumption thresholds for power users; Figure 3 This is a flowchart illustrating the method for determining the electricity consumption stability type of electricity users; Figure 4 This is a flowchart illustrating the method for determining the identification and processing of electricity load characteristics of power users. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0022] Example 1 like Figure 1 As shown, this application provides a data-driven and dynamic threshold-based method for diagnosing abnormal electricity consumption, specifically including: S1 uses the monitoring data of the power user to determine the benchmark electricity consumption of the power user, and updates the electricity consumption threshold of the power user based on the deviation between the power user's electricity consumption and the benchmark electricity consumption. S2 determines the electricity consumption variation data of the electricity user, determines the electricity consumption stability type of the electricity user based on the electricity consumption variation data of different electricity users, and determines the identification and processing method of the electricity load characteristics of the electricity user according to the electricity user data and the different electricity consumption stability types of the electricity user. S3 determines the changes in the electricity load characteristics of different power users based on the identification and processing method, and determines the update processing method for the feature identification and analysis users among the power users based on the identification and processing data of the electricity load characteristics of different power users and the identification and processing method. S4 performs dynamic updating of the electricity consumption threshold of the user identified by the feature analysis, and determines the early warning processing target of the user based on the updated data of the user identified by the feature analysis and the changes in the electricity load characteristics of the user.

[0023] Furthermore, the monitoring data includes the electricity consumption of the power user in different monitoring periods.

[0024] Furthermore, the baseline electricity consumption of the electricity user is determined based on the average electricity consumption on different dates.

[0025] Suppose that the daily electricity consumption of a certain electricity user is summed up over all days and then divided by the total number of days to obtain the user's baseline electricity consumption.

[0026] This step establishes a personalized electricity consumption reference standard for users. Its significance lies in avoiding misjudgments caused by using industry-standard values ​​by calculating the average value based on the user's own historical data.

[0027] Specifically, such as Figure 2 As shown, the process of updating the electricity consumption threshold for the electricity user specifically includes: Based on the electricity consumption data of electricity users on different dates, the mean and standard deviation are calculated and used as the benchmark electricity consumption and benchmark standard deviation for the electricity users. The mean refers to the statistical average obtained by summing up the electricity consumption of all days and dividing by the number of days; the standard deviation refers to the square root of the arithmetic mean of the squared deviations of the electricity consumption of each day from the mean.

[0028] Suppose that the daily electricity consumption of a certain electricity user for all dates is substituted into the mean formula and the standard deviation formula respectively to calculate the user's baseline electricity consumption and baseline standard deviation.

[0029] This step determines two parameters at the same time: the mean and the standard deviation. Its significance lies in establishing a central value of electricity consumption through the benchmark electricity consumption, while quantifying the fluctuation range of electricity consumption through the benchmark standard deviation.

[0030] The daily electricity consumption is compared with the baseline average. If the absolute value of the deviation between the electricity consumption and the baseline electricity consumption is a preset multiple of the standard deviation (e.g., 4 times the standard deviation), then the date is placed in the abnormal electricity consumption array. The absolute value of the deviation is the standard deviation of the preset multiple, which means that the absolute value of the difference between the daily electricity consumption and the benchmark electricity consumption is equal to the product of the preset multiple and the benchmark standard deviation; the abnormal electricity consumption array refers to the container that stores the electricity consumption data of all dates that are judged to be abnormal.

[0031] If the absolute value of the difference between a user's daily electricity consumption and the benchmark electricity consumption is greater than or equal to the product of a preset multiple and the benchmark standard deviation, then the daily electricity consumption data will be added to the abnormal electricity consumption array.

[0032] This step establishes a preliminary screening mechanism for abnormal electricity consumption data. Its significance lies in constructing a dynamic boundary for anomaly determination by multiplying a preset multiple by the standard deviation.

[0033] The percentage of abnormal electricity consumption in the total number of reported dates is taken as the percentage of abnormal dates. If the percentage of abnormal dates is within the preset range, the electricity consumption in the abnormal electricity consumption array within the most recent preset time period is sorted to obtain the fluctuation range of the electricity consumption threshold for the power user. If the percentage of abnormal dates is not within the preset range, the abnormal electricity consumption situation of the power user is determined according to the electricity consumption abnormality early warning method. When no early warning is issued, the fluctuation range of the electricity consumption threshold for the power user is still updated.

[0034] The percentage of abnormal dates refers to the ratio of the number of dates in the abnormal electricity consumption array to the total number of monitored dates; the fluctuation range refers to the reasonable range of electricity consumption centered on the benchmark electricity consumption.

[0035] It should be noted that when the percentage of abnormal dates is within the preset range, it indicates that the user's abnormal electricity consumption is an occasional fluctuation. Therefore, the recent abnormal electricity consumption is sorted to update the fluctuation range. When the percentage of abnormal dates exceeds the preset range, it indicates that the user may have a continuous abnormal electricity consumption, and the system triggers an early warning process.

[0036] This step involves classifying and processing power outages based on the percentage of abnormal dates. Its significance lies in adopting differentiated response strategies for occasional and persistent power outages.

[0037] Understandably, based on the deviation from the benchmark electricity consumption, the abnormal electricity consumption arrays for a specified number of days above the benchmark electricity consumption are sorted from smallest to largest, and the abnormal electricity consumption arrays for a specified number of days below the benchmark electricity consumption are selected as the fluctuation range of the electricity consumption threshold for the electricity user.

[0038] The sorting of deviations from smallest to largest refers to arranging the absolute values ​​of the differences between the electricity consumption on each date and the baseline electricity consumption in the abnormal electricity consumption array in ascending order; the specified number of dates above the baseline electricity consumption refers to selecting the minimum value among several electricity consumptions above the baseline electricity consumption from the sorted abnormal electricity consumption as the upper limit; the specified number of dates below the baseline electricity consumption refers to selecting the maximum value among several electricity consumptions below the baseline electricity consumption from the sorted abnormal electricity consumption as the lower limit.

[0039] Assuming that the electricity consumption in the abnormal electricity consumption array is sorted by the size of the deviation, the minimum value of the electricity consumption corresponding to the number of specified dates above the benchmark electricity consumption is selected as the upper limit, and the maximum value of the electricity consumption corresponding to the number of specified dates below the benchmark electricity consumption is selected as the lower limit, so as to ensure that the difference between the upper and lower limits is controlled within the preset range.

[0040] This step refines the method for determining the fluctuation range. Its significance lies in constructing a compact and reasonable dynamic electricity consumption range by taking the boundary values ​​of abnormal electricity consumption on both the upper and lower sides of the benchmark.

[0041] In this complete and specific embodiment, the number of power users is set to 100, the monitoring period is 90 days, the preset time period is the most recent 30 days, the preset multiple is 4 times, the specified number of dates is 3 days, and the fluctuation range control threshold is 10 degrees.

[0042] For step S1: Use the daily electricity consumption data of 100 electricity users over a 90-day monitoring period as monitoring data. Each user's electricity consumption records include 90 days within the monitoring period.

[0043] Baseline electricity consumption determination: Taking user U1 as an example, the average daily electricity consumption data over 90 days (baseline electricity consumption) is calculated to be 45.6 kWh, with a baseline standard deviation of 8.2 kWh. The baseline electricity consumption for the remaining 99 users is calculated using the same method.

[0044] Anomaly detection and abnormal electricity consumption array: Taking user U1 as an example, the anomaly detection boundary value = 4 × 8.2 = 32.8 kWh. Traversing U1's 90 days of data, comparing it with the baseline electricity consumption of 45.6 kWh, dates with an absolute deviation exceeding 32.8 kWh are included in the abnormal electricity consumption array. A total of 8 days for U1 were identified as abnormal days.

[0045] Fluctuation range determination (independent for each user): Taking the abnormal electricity consumption of U1 in the last 30 days as an example, sort them. Assume that the specified number of days is 2 days. Select the maximum value of 51.8 kWh among the electricity consumption above the baseline of the 2 days with the largest deviation (51.8 kWh, 51.5 kWh) as the upper limit. Select the minimum value of 39.2 kWh among the electricity consumption below the baseline of the 2 days with the largest deviation (39.5 kWh, 39.2 kWh) as the lower limit. Therefore, the fluctuation range of U1 is [39.2 kWh, 51.8 kWh].

[0046] Abnormal Date Percentage Judgment: The preset abnormal date percentage range is set at [0%, 15%]. U1 exceeded its fluctuation range [39.2 degrees, 51.8 degrees] for 5 days within 90 days. The abnormal date percentage = 5 ÷ 90 ≈ 5.6%, which is within the preset range.

[0047] Furthermore, the electricity consumption variation data of the electricity user is determined based on the rate of change of the electricity consumption of the electricity user between different monitoring periods.

[0048] Furthermore, such as Figure 3 As shown, the method for determining the electricity consumption stability type of the electricity user is as follows: In this embodiment, the fluctuation of electricity consumption of the power user is assessed based on the rate of change of electricity consumption between different monitoring periods, i.e., between different dates. The electricity consumption stability type of the power user is determined based on the assessment results, which also lays the foundation for determining the identification and processing method of the power user's electricity load characteristics based on the electricity consumption stability type.

[0049] The electricity consumption variation data refers to data reflecting the degree of change in electricity consumption of power users between different monitoring periods; the electricity consumption variation rate between different monitoring periods refers to the ratio of the difference in electricity consumption of users in two adjacent monitoring periods to the electricity consumption in the previous period.

[0050] If a user's electricity consumption in two adjacent monitoring periods is A and B respectively, then the formula for calculating the rate of change is (|BA|÷A)×100%, which reflects the relative change in electricity consumption between the two periods.

[0051] This step introduces a cross-cycle electricity consumption change analysis dimension. Its significance lies in capturing the dynamic changes in users' electricity consumption behavior through the rate of change index, thereby providing a quantitative basis for stability classification.

[0052] In this embodiment, the fluctuation of electricity consumption of the power user is assessed based on the rate of change of electricity consumption between different monitoring periods, i.e., between different dates. The electricity consumption stability type of the power user is determined based on the assessment results, which also lays the foundation for determining the identification and processing method of the power user's electricity load characteristics based on the electricity consumption stability type.

[0053] The assessment of changes in electricity consumption refers to a comprehensive evaluation of the regularity of users' electricity consumption behavior based on the magnitude and distribution characteristics of the rate of change during each monitoring period.

[0054] If we calculate the rate of change in a user's electricity consumption over multiple consecutive periods, and the rate of change is small and the fluctuation range is narrow in each period, then the user's electricity consumption is judged to be highly stable and classified as a stable type.

[0055] This step establishes a logical link from the calculation of the rate of change to the classification of stability. Its significance lies in transforming the quantitative rate of change index into a qualitative stability classification, thereby providing a classification basis for determining the subsequent differentiated identification and processing methods.

[0056] S11 uses the electricity consumption data of the electricity user to determine the rate of change of the electricity consumption of the electricity user between different monitoring periods; The change data of electricity consumption refers to the calculated difference in electricity consumption during each week; the change rate of electricity consumption between different monitoring periods refers to the percentage value obtained by dividing the change data by the baseline electricity consumption of the previous period.

[0057] Assuming a user's electricity consumption in two adjacent monitoring periods is 3500 kWh and 3700 kWh respectively, the change is 100 kWh, and the change rate is (100 ÷ 3500) × 100% ≈ 2.86%.

[0058] This step completes the calculation and output of the rate of change. Its significance lies in transforming the original change data into a normalized percentage indicator, making the degree of change of users of different sizes comparable.

[0059] S12 determines the electricity consumption change rate of the power user based on the average of the change rates of electricity consumption between different monitoring periods; The average value of the rate of change refers to the arithmetic mean of the rates of change calculated over multiple monitoring periods, reflecting the average change in user electricity consumption over the overall monitoring period.

[0060] For example, suppose a user calculates that the change rates over several consecutive periods are 2.1%, 3.5%, 1.8%, and 2.9%, then the average change rate is (2.1% + 3.5% + 1.8% + 2.9%) ÷ 4 ≈ 2.58%, which can be used as the user's electricity consumption change rate.

[0061] This step smooths out random fluctuations between periods through averaging, and its significance lies in obtaining a single indicator that can comprehensively reflect the stability of users' electricity consumption behavior.

[0062] S13 determines the electricity consumption stability type of the electricity user based on the electricity consumption change rate of the electricity user.

[0063] The electricity consumption stability type refers to the stability level of electricity consumption behavior classified according to the magnitude of the electricity consumption change rate; Class I stability type refers to the electricity consumption behavior type with a low electricity consumption change rate and a narrow fluctuation range; Class II stability type refers to the electricity consumption behavior type with a moderate electricity consumption change rate and certain regular fluctuations; Class III stability type refers to the electricity consumption behavior type with a high electricity consumption change rate and no obvious fluctuation pattern.

[0064] Assuming the upper limit of the change rate for the first type of stable electricity consumption is set at 5%, and the upper limit of the change rate for the second type of stable electricity consumption is set at 15%, then users with a change rate of less than 5% are classified as first-type stable electricity consumption, those between 5% and 15% are classified as second-type stable electricity consumption, and those above 15% are classified as third-type stable electricity consumption.

[0065] This step completes the stability classification of users' electricity consumption behavior. Its significance lies in dividing all users into three stable types based on the rate of change threshold, thereby providing a classification basis for determining the subsequent differentiated identification and processing methods.

[0066] It is understood that the electricity consumption stability type of the electricity user is determined based on the preset stability type corresponding to the electricity consumption change rate of the electricity user.

[0067] The preset stability type refers to three stability levels and their corresponding fluctuation rate interval boundary values ​​that are pre-set based on the statistical results of the fluctuation rate distribution of a large number of users.

[0068] This step clarifies the correspondence between stable types and variability ranges. Its significance lies in standardizing the classification rules so that different users can obtain consistent classification results under the same rules.

[0069] The electricity consumption stability types include three types: Type I, Type II, and Type III. Type I has a higher degree of electricity consumption stability than Type II, and Type II has a higher degree of electricity consumption stability than Type III.

[0070] The first type of stable type refers to users whose electricity consumption behavior is highly regular and whose electricity consumption remains within a narrow range; the second type of stable type refers to users whose electricity consumption behavior fluctuates with periodic or seasonal patterns; and the third type of stable type refers to users whose electricity consumption behavior is highly random and fluctuates drastically without obvious patterns.

[0071] Assuming that among all users, those whose electricity consumption remains within the average range of 5% for several consecutive months are classified as Category I stable users, those whose electricity consumption shows a clear alternation between weekends and weekdays are classified as Category II stable users, and those whose electricity consumption fluctuates wildly and without any discernible pattern are classified as Category III stable users.

[0072] This step clarifies the essential differences among the three stable types, which is significant in helping to understand the classification basis for the differentiated design of subsequent identification and processing methods.

[0073] In this embodiment, based on the number of electricity users and the different electricity consumption stability types of different electricity users, differentiated electricity load characteristic identification and processing methods are generated for different electricity users. This not only avoids the technical problem of excessive data processing difficulty caused by frequent electricity load characteristic identification and processing, but also takes into account the changes in electricity load and fully considers the identification and processing difficulty under different electricity load characteristic identification and processing methods. Based on the identification and processing difficulty and the electricity consumption stability type of the electricity user, different electricity load identification and processing methods are determined for different electricity users, ensuring that electricity users can have their electricity load characteristics identified and processed in a timely and effective manner when there are abnormalities in electricity consumption.

[0074] The method for identifying and processing the characteristics of electricity load refers to the differentiated load analysis strategy adopted for users with different stability types, including lenient identification methods and strict identification methods.

[0075] This step clarifies the design principles of the differentiated identification processing method, which is significant in achieving a balance between diagnostic accuracy and computational efficiency by comprehensively considering user scale and stable type distribution.

[0076] Specifically, such as Figure 4 As shown, the method for determining the identification and processing method of the electricity load characteristics of the electricity user is as follows: In this embodiment, based on the number of electricity users and the different electricity consumption stability types of different electricity users, differentiated electricity load characteristic identification and processing methods are generated for different electricity users. This not only avoids the technical problem of excessive data processing difficulty caused by frequent electricity load characteristic identification and processing, but also takes into account the changes in electricity load and fully considers the identification and processing difficulty under different electricity load characteristic identification and processing methods. Based on the identification and processing difficulty and the electricity consumption stability type of the electricity user, different electricity load identification and processing methods are determined for different electricity users, ensuring that electricity users can have their electricity load characteristics identified and processed in a timely and effective manner when there are abnormalities in electricity consumption.

[0077] S21 determines the stability weight value of the power user based on the different power consumption stability types of the power user, and determines the load stability value based on the average of the stability weight values ​​of the different power users; The stability weight value refers to the quantitative weight coefficient assigned according to the stability type of the user, which is used to reflect the relative importance of different stability types in the overall stability evaluation; the load stability value refers to the comprehensive index obtained by arithmetically averaging the stability weight values ​​of all users, which is used to characterize the stability of the overall electricity consumption behavior of all users.

[0078] Assuming a higher weight coefficient is assigned to the first type of stable type, a medium weight to the second type, and a lower weight to the third type, then each user receives a corresponding stable weight value based on their type, and the average of the stable weight values ​​of all users is the load stability value.

[0079] This step establishes a transformation mechanism from qualitative classification to quantitative weighting and then to comprehensive aggregation. Its significance lies in aggregating the individual stability information of all users into a single numerical indicator that can be compared horizontally through load stability value, providing a quantitative decision-making basis for subsequent control strategies.

[0080] S22 determines the number of power users based on the power user data; The number of electricity users refers to the cumulative number of all electricity users participating in this electricity consumption anomaly diagnosis and analysis, extracted from the system database.

[0081] Assuming the system records the electricity consumption data of a batch of electricity users during the monitoring period, the total number of these users is the number of electricity users involved in this analysis.

[0082] This step completes the statistical confirmation of the total number of users. Its significance lies in providing a numerical basis for subsequent comparison with the preset threshold, thereby triggering different branch processing logic and determining the corresponding identification and processing methods.

[0083] S23 uses the number of power users, load stability value, and power consumption stability type of the power users to determine the identification and processing method of the power user's power load characteristics.

[0084] The method for identifying and processing electrical load characteristics refers to the differentiated load analysis strategy adopted for different combinations of conditions, including different types such as lenient identification methods and strict identification methods.

[0085] Assuming that the specific identification and processing method for each user is determined based on a combination of conditions such as whether the number of users has reached a preset threshold, whether the load stability value exceeds a preset stability threshold, and the stability type of the user.

[0086] This step completes the final allocation decision of the identification and processing method. Its significance lies in transforming the quantitative analysis results of S21 and S22 into an executable differentiated strategy, so that the subsequent load characteristic identification and processing has clear methodological guidance.

[0087] Furthermore, if the number of power users is less than a preset power user number threshold, then the method for identifying the electricity load characteristics of all power users is a lenient identification method. That is, as long as the number of days in the monitoring period after the last electricity load characteristic identification time is less than the electricity consumption threshold in the system is above a preset date number threshold, the electricity load characteristics of the power users are identified, and the stability level of their electricity load characteristics is determined, that is, the stability of the use of electrical equipment in different time periods.

[0088] As explained in the above steps, if the number of electricity users is less than the preset threshold for the number of electricity users, then the method for identifying the electricity load characteristics of all electricity users is a lenient identification method. The lenient identification method refers to a load identification strategy with relatively lenient triggering conditions. When the number of days with electricity consumption below the threshold reaches a certain requirement, the identification analysis is initiated. The preset threshold for the number of days refers to the lower limit of the number of days required to trigger the identification analysis under the lenient identification method.

[0089] Assuming the number of users participating in the analysis is small, even with a lenient identification strategy, the amount of data processing is within a controllable range. Therefore, the system uniformly assigns a lenient identification method to all users, that is, identification is initiated after the number of days with electricity consumption below the threshold during the monitoring period reaches a certain number.

[0090] This step establishes a unified handling strategy for small-scale user scenarios. Its significance lies in reducing the system's computational burden in small-scale user scenarios by simplifying the judgment logic, while ensuring that effective load identification can still be achieved.

[0091] Furthermore, if the number of electricity users is not less than a preset electricity user number threshold, the following applies: Scenario 1: If the load stability value is greater than the preset stability threshold, then the load of different power users is relatively stable. Therefore, even if an overly aggressive identification method is adopted, the data processing difficulty is relatively small. The identification method for determining the power load characteristics of all power users is as follows: as long as the number of days in the monitoring period after the last identification time of the power load characteristics is less than the power consumption threshold in the system is greater than the preset number of days threshold, the power load characteristics of the power users are identified, and the stability level of their power load characteristics is determined, that is, the stability level of the use of electrical equipment in different time periods.

[0092] Case 1: If the load stability value is greater than the preset stability threshold, then the identification and processing method for the electricity load characteristics of all power users is determined to be a lenient identification method.

[0093] The load stability value being greater than the preset stability threshold means that the load stability value calculated by S21 exceeds the system's preset stability lower bound, indicating that the overall electricity consumption behavior of all users is relatively stable.

[0094] Assuming that the load stability value is at a high level, it indicates that the overall electricity consumption of the users involved in the analysis has small fluctuations and strong regularity. In this case, even if the identification and analysis scope is expanded, the complexity of data processing will not increase significantly, and the electricity consumption will not frequently fall below the threshold. Therefore, a relaxed identification method is uniformly adopted to improve the efficiency of handling.

[0095] This step, scenario 1, establishes a unified handling path for highly stable scenarios. Its significance lies in the fact that when the overall electricity consumption behavior is highly stable, there is no need to differentiate users in detail. A uniform, lenient strategy can be adopted to improve efficiency while ensuring effectiveness.

[0096] Case 2: If the load stability value is not greater than the preset stability threshold, and if the power user's power consumption stability type is a type of stability, then the method for identifying the power user's power load characteristics is as follows: if, within the monitoring period after the last power load characteristic identification time, the number of days in which the power user's power consumption is less than the power consumption threshold in the system is greater than the preset number of days threshold, then the power user's power load characteristics are identified. The load stability value not exceeding the preset stability threshold indicates that the overall user electricity consumption behavior is not stable enough, but an individual user may still belong to a stable type; the stable type refers to an electricity consumption behavior type with a low electricity consumption fluctuation rate.

[0097] Assuming that although the overall load stability value is low, there are still some stable users in the system. These users have relatively regular electricity consumption behavior, so a lenient identification method can still be assigned to them, thereby reducing the data processing burden on these users while ensuring the diagnostic effect.

[0098] In step 2, a relaxed identification path is established separately for a stable type of user. The significance of this is that even when the overall stability is poor, a stable type of user can still be identified and a relaxed strategy can be adopted, thereby achieving a balance between system resources and diagnostic effectiveness.

[0099] Scenario 3: If the electricity consumption stability type of the power user does not belong to a stable type, obtain the proportion of the number of stable types among the power users, and determine whether the proportion of the number of stable types among the power users is greater than a preset proportion threshold. If not, the method for identifying the electricity load characteristics of the power user is to identify the electricity load characteristics of the power user as long as the number of days in the monitoring period after the last electricity load characteristic identification time is greater than the target date number threshold. If yes, the method for identifying the electricity load characteristics of the power user is to identify the electricity load characteristics of the power user as long as the number of days in the monitoring period after the last electricity load characteristic identification time is greater than the target date number threshold and no user load characteristic identification has been performed in the most recent preset period (e.g., 30 days).

[0100] Specifically, if the electricity consumption stability type of the power user does not belong to a stable type, the proportion of the number of stable types among the power users is obtained, and it is determined whether the proportion of the number of stable types among the power users is greater than a preset proportion threshold.

[0101] The percentage of a stable user type refers to the proportion of users of a stable user type in the total number of users; the preset percentage threshold refers to the lower bound of the percentage used to determine whether a user type occupies a dominant position in the group.

[0102] If a certain percentage of users in the system belong to a stable type, this percentage is compared with a preset percentage threshold to determine whether a large number of users are currently using a lenient approach to identify their electricity load characteristics, thereby deciding what refined identification strategy to adopt subsequently.

[0103] If the proportion of a stable type is greater than a preset proportion threshold, then the identification and processing method for the electricity load characteristics of the power user is determined to be a dual-condition constraint identification method.

[0104] The dual-condition constraint refers to an identification strategy that adds a time interval condition to the date quantity condition. That is, a new round of identification analysis is triggered only when the date quantity condition is met and the time interval since the last identification analysis has exceeded a preset time interval.

[0105] If a stable user type accounts for a relatively high proportion in the group, it indicates that the user type of that type in the entire group has a certain scale. In this case, a strict identification method with dual constraints is adopted for non-user types. This ensures effective monitoring of non-user types while avoiding overly frequent identification analysis through time interval conditions. This branch establishes a strict identification strategy for scenarios where non-user types but a high proportion of the stable user type exist. Its significance lies in avoiding the waste of computing resources caused by overly frequent identification without reducing the ability to identify anomalies by adding time interval conditions.

[0106] It is understandable that the target date number threshold is greater than the preset date number threshold.

[0107] This complete and specific embodiment follows S1, takes the electricity users entering S2 as the analysis object, sets the preset electricity user number threshold to 5, the preset stability threshold to 0.60, the preset date number threshold to 5 days, the target date number threshold to 8 days, and the preset percentage threshold to 30%.

[0108] For determining the stable user types S11 to S13: Taking all 100 users who transitioned from S1 to S2 as the analysis subjects, the electricity consumption change rate for each user during the monitoring period was calculated. The upper limit of the change rate for Type I stable user was set at 5%, and for Type II stable user at 15%. Users with change rates exceeding 15% were classified as Type III stable users. The classification results are as follows: 35 users (35%) were Type I stable users, 42 users (42%) were Type II stable users, and 23 users (23%) were Type III stable users.

[0109] For calculating the S21 stability weight value and load stability value: Assume the weight coefficient for Category I stability is 1.0, for Category II stability is 0.7, and for Category III stability is 0.4. Taking 35 Category I users as an example, each user has a weight value of 1.0; 42 Category II users each have a weight value of 0.7; and 23 Category III users each have a weight value of 0.4. The load stability value = sum of all user weight values ​​÷ total number of users = (35 × 1.0 + 42 × 0.7 + 23 × 0.4) ÷ 100 = (35 + 29.4 + 9.2) ÷ 100 = 73.6 ÷ 100 = 0.736.

[0110] For S22, the number of users is determined as follows: the total number of users participating in this analysis is 100.

[0111] Regarding the identification and processing method determined in S23: the preset threshold for the number of electricity users is 50, and 100 > 50, therefore, the branch judgment of "the number of electricity users is not less than the preset threshold for the number of electricity users" is entered. The load stability value is 0.736 > the preset stability threshold is 0.60. According to Case 1, the identification and processing method for the electricity load characteristics of all 100 users is determined to be a lenient identification method. That is, in the monitoring period after the last identification time, if the number of days when the electricity consumption is less than the electricity consumption threshold (i.e., the minimum value of the fluctuation range determined in step S1) is more than 5 days above the preset date number threshold, then the electricity load characteristics are identified. Among them, the target date number threshold is greater than the preset date number threshold, that is, more than 7 days above.

[0112] Furthermore, the variation in the electricity load characteristics of the electricity users is determined based on the deviation of the time period data under different electricity load characteristics on different dates.

[0113] Furthermore, the method for determining the update processing method for the feature identification and analysis of electricity users is as follows: In this embodiment, based on existing electricity user identification and processing data, the users undergoing electricity load characteristic identification and the number of identification processes for different users are determined. This enables the determination of the reliability and timeliness of identification and processing for different electricity users under the current identification and processing method, based on the identification and processing data of electricity load characteristics and the number of identification processes for different users. Based on the reliability and timeliness, differentiated feature identification analysis users are generated, and a method for updating users' features for real-time identification and processing and reliable updating of electricity consumption thresholds using the identification and processing results is determined. This not only reduces the difficulty of identification and processing but also improves the reliability of early warning processing by reliably updating feature identification analysis users.

[0114] The core objective of this embodiment is to determine the update processing method for users with characteristic identification and analysis. By analyzing the electricity load characteristic identification and processing data of different users, the reliability and timeliness of the identification and processing are evaluated, and differentiated update strategies are generated. The core logic is: electricity load characteristics are determined by power equipment. If the power equipment changes, the electricity load characteristics will change accordingly, manifested as changes in the current curve characteristics; when there are a large number of electricity load characteristic groups on different dates, it indicates that the load of the power equipment changes frequently. The overall logic follows the process of "user screening for identification and processing → reliability assessment → update method determination → electricity consumption threshold construction".

[0115] S31 identifies power users with identified load characteristics based on the power load characteristics of different power users, and uses these power users as identified users. The electricity load characteristics refer to the current curve characteristics generated by power equipment during operation. Different electricity load characteristics correspond to different combinations of power equipment. When power equipment changes, the corresponding electricity load characteristics change accordingly. The identified users refer to power users who have undergone electricity load characteristic identification processing at least once within the historical monitoring period. These users have historical electricity load characteristic data that can be used for subsequent analysis.

[0116] This step completes the basic screening of electricity users. Its significance lies in screening out the user group with electricity load characteristic identification and processing data from all electricity users, providing analysis objects for subsequent reliability assessment and update method determination, avoiding the inclusion of users who do not meet the analysis conditions in the processing flow, thereby improving the overall diagnostic efficiency.

[0117] The above steps include the following: S311 Obtain the proportion of identified users among the power users, and determine whether the proportion of identified users among the power users is greater than the preset identification user proportion threshold. If so, the identification processing reliability is high, that is, it can be identified in time when changes occur in the electricity load characteristics. Then, the update processing method of the feature identification and analysis users among the power users is determined to be the basic update method. If not, proceed to step S312.

[0118] The proportion of users undergoing identification processing refers to the ratio of the number of users undergoing identification processing to the total number of electricity users, reflecting the proportion of all electricity users capable of electricity load characteristic identification processing. The basic update method refers to the update processing strategy adopted for users with relatively more electricity load characteristics. This method calculates the electricity consumption of each time period group and sums them to construct an electricity consumption threshold.

[0119] This step determines the overall reliability by identifying the proportion of users who have processed the data. The significance of this step is that when most electricity users have the ability to identify and process data, it indicates that the overall monitoring system is sound and can promptly identify changes in electricity load characteristics. Therefore, a more efficient basic update method can be adopted. When the proportion is too low, it is necessary to further evaluate the reliability of individual users.

[0120] It should be noted that the method for determining the basic update method is as follows: The time periods in which the similarity coefficient of the user's electricity load characteristics meets the requirements are divided into the same time period group. If the number of time period groups for the user on different dates is greater than the preset number of time period groups, then the user is determined to be a user of feature identification analysis.

[0121] The time period group refers to a set of time periods on different dates where the same group of electrical equipment exhibits similar electricity load characteristics. In other words, the same group of electrical equipment will display similar current curve characteristics during operation, and these similar time periods are grouped into the same time period group. The similarity coefficient of the electricity load characteristics refers to the degree of similarity between the electricity load characteristic curves of different time periods; a higher coefficient indicates that the combinations of electrical equipment corresponding to the two time periods are more similar. The number of time period groups refers to the number of time period groups a user is divided into within a day. Each time period group corresponds to a group of electrical equipment; the more equipment, the more time period groups.

[0122] This step clarifies the criteria for determining the basic update method. Its significance lies in: by dividing time period groups, a correspondence is established between the user's electricity load characteristics and electrical equipment; if the number of time period groups for a user in a day reaches the preset requirement, it indicates that the user has multiple independently identifiable electrical equipment, and the electricity consumption threshold can be calculated based on each group, thereby improving the accuracy of threshold construction.

[0123] It should be noted that if the user belongs to the feature identification analysis user, the electricity load feature identification processing is performed on different dates to construct the electricity consumption under the same time period group, that is, the same group of power equipment, and the electricity consumption under different time period groups is used to construct the electricity consumption threshold for different dates.

[0124] The electricity consumption within the same time period group refers to the electrical energy consumed by the same group of electrical devices within a specific time period. This electricity consumption is obtained by integrating the current curves of the same group of electrical devices over time. The electricity consumption thresholds for different dates refer to the lower limit thresholds for electricity consumption constructed for each date, calculated based on the electricity consumption of each time period group within that date.

[0125] This step illustrates the method for constructing user electricity consumption thresholds through feature identification and analysis. Its significance lies in: calculating electricity consumption separately for time period groups (electrical equipment groups), and then summing the electricity consumption of each group to obtain the total electricity consumption threshold. This hierarchical calculation method based on the actual composition of electrical equipment can more accurately reflect the actual electricity consumption patterns of users and avoid threshold deviations caused by equipment changes.

[0126] S312, determine whether there are any power users among the power users whose identification processing count exceeds the preset identification processing count threshold. If yes, proceed to step S32. If no, the reliability of the identification processing at this time is not high. Therefore, determine the update processing method of the feature identification analysis user among the power users as the first update method.

[0127] The number of identification processing operations refers to the total number of times a single electricity user undergoes electricity load characteristic identification processing within a monitoring period. A higher number of operations indicates a more comprehensive understanding of the user's electricity load characteristics. The first update method refers to an update processing strategy adopted when the current identification processing method is insufficient. This method determines whether a user is a feature-identified user by evaluating the stability of the time period group coverage.

[0128] This step determines the reliability of a single user by the number of identification processes. The significance of this step is that when there are users with a sufficient number of identification processes, it means that the users have sufficient monitoring data to support their identification and analysis using the existing identification and analysis methods, indicating that the identification and analysis methods at this time have a high degree of reliability. When there are no users with a sufficient number of identification processes, the overall reliability is not high, so the first update method is adopted.

[0129] It should be noted that the first update method is to determine that the user belongs to the feature identification analysis user if the percentage of the number of days in which the number of user time period groups is greater than or equal to the number of preset time period groups is greater than or equal to the first percentage.

[0130] The first percentage refers to the proportion of days in which the number of user time-segment groups reaches the preset requirement to the total number of days within the monitoring period. This proportion reflects the stability of the user's device configuration within the monitoring period. The requirement for the number of time-segment groups in the first update method is the same as that in the basic update method, that is, the number of time-segment groups must reach or exceed the preset number. The difference is that the first update method is more lenient, requiring only that the user reaches or exceeds the first percentage to be identified as a user for feature identification and analysis.

[0131] This step explains the specific criteria for the first update method. Its significance lies in further examining the stability of the user's device configuration within the monitoring period, based on the requirement for the number of time period groups. Only when the device configuration meets the requirements for most days is the user identified as a feature identification and analysis user, thus achieving a balance between reliability and update efficiency.

[0132] S32, Based on the power user identification processing method, determine the power users of the loose identification method among the power users, and designate the power users of the loose identification method among the power users as loosely identified power users.

[0133] The aforementioned lenient identification method refers to an identification processing approach that has relatively low requirements for the similarity of electricity load characteristics. This method uses a more lenient similarity coefficient threshold when determining whether time periods belong to the same time period group, allowing more time periods to be grouped together. The aforementioned leniently identified electricity users refer to electricity users whose electricity load characteristics are identified using the lenient identification method. The more such users there are, the higher the reliability of load change identification.

[0134] This step completes the classification of electricity users. Its significance lies in: dividing users into two categories based on different identification and processing methods, namely, using a lenient identification method and using a strict identification method, determining the reliability of the identification and processing under the current identification and processing method, and laying the foundation for determining the update method for different types of users in the future.

[0135] S33 determines the update processing method for feature identification analysis users among the power users by considering the number of identification processing times for different identification processing users and the loosely identified power users among the power users.

[0136] This step comprehensively considers two factors—the number of identification processes and the number of users with lenient identification—to determine the update method. Its significance lies in the fact that, based on the basic update method and the first update method, the proportion of users with lenient identification is further introduced to form a more comprehensive reliability assessment system, thereby determining whether to adopt the basic update method or the second update method.

[0137] Furthermore, the above steps include the following: S331 identifies power users whose identification processing count exceeds a preset threshold as reliable identification users, and determines whether the proportion of reliable identification users among the power users is greater than a preset threshold for the proportion of reliable identification users. If so, the update processing method for feature identification analysis users among the power users is determined to be the basic update method; otherwise, proceed to step S332.

[0138] The term "reliable identified users" refers to electricity users whose identification processing frequency exceeds a preset threshold. These users have sufficient electricity load characteristic data, and their identification reliability is high under the current identification processing method. The term "reliable identified user ratio" refers to the ratio of the number of reliable identified users to the total number of electricity users, reflecting the proportion of users with reliable identification processing capabilities among all users.

[0139] This step makes a preliminary judgment based on the proportion of reliably identified users. Its significance is that when the proportion of reliably identified users meets the requirements, it indicates that the identification reliability under the current identification processing method is relatively high, and the overall reliability is relatively high. Therefore, the basic update method is directly adopted. When the proportion is insufficient, it is necessary to further introduce the factor of relaxed identification users for comprehensive evaluation.

[0140] S332, obtain the proportion of loosely identified power users among the power users, and determine the identification reliability value based on the average of the proportion of loosely identified power users and the proportion of reliable identified power users among the power users. Determine whether the identification reliability value is greater than a preset reliability threshold. If yes, determine the feature identification analysis user update processing method among the power users as the basic update method. If no, determine the feature identification analysis user update processing method among the power users as the second update method.

[0141] The identification reliability value refers to the comprehensive reliability index obtained by averaging the proportion of users with lenient identification and the proportion of users with reliable identification. This index considers both the proportion of users using the lenient identification method and the proportion of users with reliable identification capabilities, determining the identification reliability under the current identification processing method. The second update method refers to the update processing strategy adopted by power users when the reliability assessment is at a medium level. This method's requirements for time-group load variability are between those of the first update method and the basic update method.

[0142] This step makes a final judgment by identifying a reliability value. Its significance lies in: comprehensively considering the proportion of the two types of users to obtain a more comprehensive reliability assessment; when the identified reliability value meets the requirements, the basic update method is used, and when the identified reliability value is insufficient, the second update method is used, so as to achieve an effective match between the reliability level and the update method.

[0143] Furthermore, the second update method is to determine that the user belongs to the feature identification analysis user if the percentage of the number of days in which the number of user time period groups is greater than or equal to the number of preset time period groups is greater than or equal to the second percentage.

[0144] The second percentage refers to the proportion of days in which the number of user time-segment groups reaches a preset requirement within the monitoring period, out of the total number of days. This proportion requirement is higher than the first percentage. The first percentage is less than the second percentage, indicating that the second update method has lower requirements for the variability of time-segment groups than the first update method, but higher requirements than the basic update method.

[0145] This step explains the specific criteria for the second update method. Its significance lies in setting an intermediate level between the basic update method and the first update method. Users who meet the second proportion requirement are identified as feature identification and analysis users, thus forming a three-level differentiated update strategy system.

[0146] Furthermore, the dynamic update process for the user's electricity consumption threshold based on the aforementioned feature identification analysis specifically includes: Construct the electricity consumption under the same time period group, that is, the same group of power devices, and use the electricity consumption under different time period groups to construct the electricity consumption threshold for different dates.

[0147] This step clarifies the method for dynamically updating the electricity consumption threshold of users based on feature identification and analysis. Its significance lies in: calculating the electricity consumption of each group (power equipment group) within a specific time period, and then summing the electricity consumption of different groups to obtain the total electricity consumption threshold for that date; this hierarchical calculation method based on the actual composition of power equipment can accurately reflect the actual electricity consumption level of users on different dates.

[0148] Assuming there are 100 electricity users in the industrial park, the monitoring period is 90 days, the preset threshold for the number of groups in a given time period is 3, the preset threshold for the proportion of identified users is 0.70, the preset threshold for the number of identification processing times is 50, the preset threshold for the proportion of reliable identified users is 0.40, the preset threshold for reliability is 0.55, the first proportion is 0.60, and the second proportion is 0.75.

[0149] S31 Identify and process the user: Of the 100 electricity users, 72 underwent electricity load characteristic identification processing during the monitoring period; these users were marked as identified users. The proportion of identified users is 72 divided by 100, which equals 0.72.

[0150] S311 determines and identifies the percentage of users processed: Since the proportion of users identified and processed (0.72) is greater than the preset threshold of 0.70, it is necessary to further determine whether the update processing method for each user is the basic update method.

[0151] Determining the basic update method: Taking user U1 as an example, the number of time groups in each day of the 30-day period reached more than 3. Therefore, U1 was identified as a user for feature identification and analysis, and the basic update method was adopted.

[0152] Construction of U1's power consumption threshold under the basic update method: U1 was divided into four time groups (G1, G2, G3, G4) during the historical monitoring period, each corresponding to a different group of power equipment.

[0153] Suppose that U1 has 180 time periods assigned to group G1 over a historical 90-day period, and the total historical electricity consumption of group G1 over these 180 time periods is 1170 kWh. Therefore, the historical average electricity consumption of group G1 per unit time period is 1170 divided by 180, which equals 6.5 kWh per time period. U1 has 270 time periods assigned to group G2, and the total historical electricity consumption of group G2 over these 270 time periods is 2430 kWh. Therefore, the historical average electricity consumption of group G2 per unit time period is 2430 divided by 270, which equals 9 kWh per time period. U1 has 180 time periods assigned to group G3, and the total historical electricity consumption of group G3 over these 180 time periods is 990 kWh. Therefore, the historical average electricity consumption of group G3 per unit time period is 990 divided by 180, which equals 5.5 kWh per time period. There are 90 time periods that are assigned to Group G4. The total historical electricity consumption of Group G4 in these 90 time periods is 360 kWh. Therefore, the historical average electricity consumption of Group G4 in a unit time period is 360 divided by 90, which equals 4 kWh per time period.

[0154] Then, the electricity consumption threshold is calculated based on the actual runtime of each group on the current date. Let's assume that on the current target date, the runtime of each group for U1 is consistent with the historical average: Group G1 runs for 2 periods, Group G2 runs for 3 periods, Group G3 runs for 2 periods, and Group G4 runs for 1 period. Then, on the current date, the electricity consumption of Group G1 is 6.5 x 2 = 13 kWh; the electricity consumption of Group G2 is 9 x 3 = 27 kWh; the electricity consumption of Group G3 is 5.5 x 2 = 11 kWh; and the electricity consumption of Group G4 is 4 x 1 = 4 kWh. Summing the electricity consumption of the four groups on the current date, we get the electricity consumption threshold for U1 on the current date as 13 + 27 + 11 + 4 = 55 kWh.

[0155] In another possible embodiment, if the proportion of identified users (0.42) is not greater than the preset user proportion threshold (0.70), the process proceeds to S32 to determine the leniently identified electricity users. Of the 100 electricity users, 58 were identified using a lenient identification method. The percentage of leniently identified electricity users is 58 divided by 100, which equals 0.58.

[0156] S33 Comprehensive Determination of Update Processing Method: First, electricity users who have undergone identification and processing more than 5 times are considered reliable identification users. If 42 out of 100 electricity users have undergone identification and processing more than 5 times, then the percentage of reliable identification users is 0.42.

[0157] Since the proportion of reliably identified users (0.42) is greater than the preset threshold of 0.40, the update processing method for the feature identification analysis of all power users is determined as the basic update method.

[0158] As a comparison: If the proportion of reliably identified users is 0.35, which is less than the preset reliable user identification threshold of 0.40, then a reliable identification value needs to be calculated. Assuming the proportion of loosely identified electricity users is 0.58 and the proportion of reliably identified users is 0.35, then the reliable identification value equals 0.58 plus 0.35 divided by 2, which equals 0.465. Since 0.465 is less than the preset reliable threshold of 0.55, the second update method should be used in this scenario.

[0159] Determination of the second update method: If user U3 has 3 or more time groups in 70 out of 90 days, accounting for 0.78%, which is greater than the second percentage of 0.75, then U3 is identified as a user for feature identification analysis and the second update method is adopted.

[0160] This embodiment, through the complete process from S31 to S33, has three core values: First, it accurately reflects the actual electricity consumption patterns of users with multiple sets of power equipment by statistically analyzing the average electricity consumption per unit time period of each group within its corresponding time period and dynamically calculating the electricity consumption based on the actual running time of each group on the current date. Second, it achieves differentiated processing for users with different reliability levels by setting three levels: basic update method, first update method, and second update method. Third, it calculates the electricity consumption threshold based on power equipment groups, ensuring the accuracy and reliability of the threshold.

[0161] Furthermore, the second update method is to determine that the user belongs to the feature identification analysis user if the percentage of the number of days in which the number of user time period groups is greater than or equal to the number of preset time period groups is greater than or equal to the second percentage.

[0162] It should be noted that the first proportion is less than the second proportion.

[0163] Furthermore, the dynamic update process for the user's electricity consumption threshold based on the aforementioned feature identification analysis specifically includes: Construct the electricity consumption under the same time period group, that is, the same group of power devices, and use the electricity consumption under different time period groups to construct the electricity consumption threshold for different dates.

[0164] Furthermore, the method for determining the early warning processing target of the power user is as follows: In this application, based on the updated user data identified by feature recognition, an overall assessment of the reliability of the current early warning processing is conducted. Furthermore, by combining the changes in the electricity load of power users, an overall assessment of the drastic changes in the electricity load of power users is conducted. This enables the determination of an early warning method for abnormal electricity consumption of power users from the perspectives of the reliability of the current early warning processing and the drastic changes. This ensures the reliability and timeliness of the early warning, while also avoiding the technical problem of excessively high user processing difficulty caused by frequent early warnings.

[0165] S41 determines the proportion of the feature-identified users among the power users based on the updated data of the feature-identified users, and uses the proportion of the feature-identified users among the power users as the identification and analysis proportion.

[0166] The feature identification and analysis users refer to those identified in S3 using the basic update method, the first update method, or the second update method, and these users are capable of constructing electricity consumption thresholds based on time period groups. The identification and analysis ratio refers to the ratio of the number of feature identification and analysis users to the total number of electricity users.

[0167] This step completes the assessment of user coverage for feature identification and analysis. Its significance lies in the fact that the identification and analysis ratio determines the reliability of the overall electricity consumption threshold construction.

[0168] Specifically, if the identification and analysis ratio is less than the preset identification and analysis ratio threshold, then the construction method of the electricity consumption threshold for a large number of electricity users is not accurate enough, and it is impossible to achieve accurate electricity consumption anomaly warning. Therefore, in order to prevent omissions, the electricity consumption anomaly warning method for electricity users excluding feature identification and analysis users is determined to be: if the proportion of abnormal dates is not within the preset abnormal date proportion range, then the warning processing for the electricity user is performed.

[0169] This step explains the early warning handling method when the identification and analysis ratio is insufficient. Its significance lies in the fact that when the electricity consumption thresholds for a large number of users are not constructed accurately, a more sensitive early warning strategy needs to be adopted for users who are not identified and analyzed by features.

[0170] It should also be noted that if the identification and analysis ratio is not less than the preset identification and analysis ratio threshold, then step S42 will be performed for further analysis of the number of fluctuation groups.

[0171] This step, as a judgment branch, is significant in that it guides subsequent analysis into S42, based on identifying and processing changes in the user's electricity load characteristics, to further determine the severity of the early warning processing.

[0172] S42 determines the average value of time period groups on different dates in the most recent identification analysis of the user based on the changes in the user's electricity load characteristics, and uses the average value of time period groups on different dates in the most recent identification analysis of the user as the number of fluctuation groups of the user.

[0173] The variation in electricity load characteristics refers to the degree of change in the electricity load characteristics of the identified users across different time periods within the most recent identification and analysis cycle. The number of fluctuation groups refers to the average number of groups for each date and time period during the most recent identification and analysis process, reflecting the overall drastic degree of change in electricity load characteristics.

[0174] This step completes a quantitative assessment of the degree of drastic change in electricity load characteristics. Its significance lies in the fact that the more time period groups there are, the more complex the power equipment configuration is and the more unstable the load characteristics are.

[0175] The above steps include the following: Determine whether the power user belongs to the user identification and processing category. If yes, proceed to step S43. If no, the power user cannot determine the trend of its power load characteristics because it has not undergone power load characteristic identification and processing. Therefore, a strict power consumption anomaly early warning method is required. Thus, the early warning processing target for the power user is determined to be that if the percentage of abnormal dates is not within the preset abnormal date percentage range, then the power user will be subject to early warning processing.

[0176] This step branches the processing based on whether the user belongs to the identification processing user category. The significance of this is that since non-identification processing users have not undergone electricity load characteristic identification, it is impossible to determine the changing trend of their electricity load characteristics. Therefore, it is necessary to directly adopt a strict early warning method as a conservative strategy.

[0177] S43 uses the identification analysis ratio and the number of fluctuation groups of different identification processing users to determine the early warning processing target of the power user.

[0178] This step comprehensively considers two factors—the identification and analysis ratio and the number of fluctuation groups—to determine the early warning and handling targets. Its significance lies in: assessing the overall reliability through the identification and analysis ratio, and assessing the severity of load fluctuations through the number of fluctuation groups.

[0179] Furthermore, the above steps include the following: Electricity users subject to strict electricity consumption anomaly early warning methods are designated as strictly controlled electricity users. The proportion of strictly controlled electricity users among all electricity users is determined to be greater than a preset control user proportion threshold. If so, the electricity consumption anomaly early warning method for electricity users excluding strictly controlled electricity users and feature identification analysis users is determined as follows: If the proportion of abnormal dates is not within the preset abnormal date proportion range, and there are more than the target number of abnormal dates within the most recent preset time period, then early warning processing is performed for the electricity user; otherwise, the electricity consumption anomaly early warning method for the identified electricity users is determined based on the number of different fluctuation groups of identified electricity users.

[0180] This step branches out the processing based on the proportion of strictly controlled electricity users. The significance of this is that when the proportion of strictly controlled electricity users is too high, it indicates that the sensitivity of the early warning processing is high. In order to avoid frequent early warnings, it is necessary to adopt a stricter early warning strategy for other users as well.

[0181] Specifically, if the average number of fluctuation groups of different identified power users is greater than the preset threshold for the number of fluctuation groups, it indicates that the overall fluctuation of the power load of the identified power users is relatively high. Therefore, in order to avoid false alarms and excessive difficulty in investigation, the power consumption anomaly warning method for identified power users excluding those identified by feature identification analysis, that is, power users excluding those under strict control and those identified by feature identification analysis, is as follows: if the percentage of abnormal dates is not within the preset percentage of abnormal dates, and there are abnormal dates exceeding the target number within the most recent preset time period, then the power user is given an early warning.

[0182] This step explains the early warning handling method when the power load fluctuates drastically. Its significance is that when the load fluctuates drastically, it is necessary to add the condition of "abnormal dates exceeding the target number within the most recent preset time period" to raise the early warning trigger threshold and avoid false alarms.

[0183] Additionally, it should be noted that if the average number of fluctuation groups of different identified power users is not greater than the preset threshold for the number of fluctuation groups, it indicates that the overall fluctuation of the power load of the identified power users is not high. For identified power users excluding those identified by feature identification analysis, that is, power users excluding those under strict control and those identified by feature identification analysis, the power consumption anomaly warning method is as follows: if the percentage of abnormal dates is not within the preset percentage of abnormal dates, and there are more than the target number of abnormal dates in the most recent preset time period, or there are consecutive abnormal dates in the most recent preset time period, then the power user is given an early warning.

[0184] This step explains the early warning handling method when the power load fluctuation is not severe. Its significance is that when the power load fluctuation is not high, the early warning conditions can be set relatively loosely, and "the existence of consecutive abnormal dates" can be added as an alternative condition to trigger the early warning.

[0185] It should be noted that if the electricity user belongs to the user identified by feature analysis, the date on which the difference between the electricity consumption threshold and the electricity consumption is greater than the electricity consumption threshold will be regarded as the electricity consumption deviation date. If the number of electricity consumption deviation dates of the electricity user does not meet the requirements, an early warning will be issued immediately.

[0186] This step explains the warning triggering conditions for user feature identification analysis. Its significance is that user feature identification analysis has a reliable electricity consumption threshold built based on time period groups. As long as the number of days with electricity consumption deviation reaches the requirement, the warning will be triggered immediately to ensure the timeliness of the warning.

[0187] Assuming there are 100 electricity users, a monitoring period of 90 days, a preset identification and analysis ratio threshold of 0.70, a preset control user ratio threshold of 0.20, a preset fluctuation group number threshold of 3.5, a preset abnormal date percentage range of 0.10 to 0.30, a preset time period of 15 days, and a target date number of 5 days.

[0188] S41 Determine the identification and analysis ratio: Based on the feature identification and analysis user update processing method defined in S3, it is assumed that out of 100 electricity users, 75 users are identified as feature identification and analysis users. The identification and analysis ratio is 75 divided by 100, which equals 0.75.

[0189] Since the identification and analysis ratio of 0.75 is greater than the preset identification and analysis ratio threshold of 0.70, it indicates that most users are able to construct electricity consumption thresholds based on reliable time period group divisions, and the overall reliability of the early warning processing is high. Therefore, further analysis of the number of fluctuation groups will be performed in S42.

[0190] S42 Determine the number of fluctuation groups: Of the 100 electricity users, 88 are identified and processed users, and 12 are unidentified and processed users (strictly controlled electricity users).

[0191] For the 88 identified users, the time-period grouping was analyzed during the most recent identification analysis (90-day period). Taking user U1 as an example, the number of time-period groups for each day of the 90 days was counted, and the average was taken to obtain a fluctuation group number of 3.2 for U1. The fluctuation group number for all 88 identified power users was calculated using the same method, and then the average of these 88 values ​​was taken to obtain an average fluctuation group number of 2.8 for the identified power users.

[0192] S43 Determine the target for early warning processing: First, calculate the proportion of strictly controlled electricity users. The number of strictly controlled electricity users is 12, and the proportion is 12 divided by 100, which equals 0.12.

[0193] Since the proportion of strictly controlled electricity users (0.12) is less than the preset control threshold of 0.20, it is necessary to further determine the early warning method based on the number of fluctuating electricity user groups.

[0194] Since the average number of fluctuation groups is 2.8, which is greater than the preset threshold of 3, it indicates that the power load fluctuation of the identified power users is relatively high. Therefore, for the identified power users excluding those identified by feature identification analysis, the power consumption anomaly warning method is as follows: if the proportion of abnormal dates is not within the preset range of abnormal date proportion, i.e., greater than 0.15, and there are more than 3 abnormal dates in the last 15 days, then a warning will be issued.

[0195] For the 12 strictly controlled electricity users, since their electricity load characteristics were not identified, a strict early warning method was adopted: if the proportion of abnormal days is greater than 0.15, an early warning will be issued immediately.

[0196] For 75 users with feature identification and analysis, the date on which the difference between the electricity consumption threshold and the electricity consumption is greater than the electricity consumption threshold is defined as the electricity consumption deviation date. The preset requirement for the number of electricity consumption deviation dates is 2 days. If the number of electricity consumption deviation dates is greater than 2 days, an early warning will be issued immediately.

[0197] This embodiment, through the complete process from S41 to S43, determines differentiated early warning methods for different types of users' abnormal electricity consumption. Its core value lies in three aspects: First, by identifying and analyzing the proportion of abnormal electricity consumption, it assesses the overall reliability, ensuring that the early warning strategy matches the ability to construct electricity consumption thresholds; second, by assessing the severity of load fluctuations through the number of fluctuation groups, it achieves a balance between early warning sensitivity and investigation difficulty; and third, by constructing differentiated early warning methods, it ensures both the reliability and timeliness of early warnings while avoiding the problem of excessively high user processing difficulty caused by frequent early warnings.

[0198] Example 2 Secondly, this application provides a data-driven and dynamic threshold-based power consumption anomaly diagnosis system, employing the aforementioned data-driven and dynamic threshold-based power consumption anomaly diagnosis method, specifically including: Threshold update module, identification processing module, early warning processing module; The threshold update module updates the electricity consumption threshold of the power user. The identification processing module is responsible for determining the identification processing method for the electricity load characteristics of the power users; The early warning processing module is responsible for determining the early warning processing targets for the power users.

[0199] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0200] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0201] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A data-driven and dynamic threshold-based method for diagnosing abnormal electricity consumption, characterized in that, Specifically, it includes: Acquire monitoring data of electricity users, including the electricity consumption of the electricity users in different monitoring periods. Based on the electricity consumption data of the electricity users on different days, calculate the mean and standard deviation, which serve as the benchmark electricity consumption and benchmark standard deviation of the electricity users. The electricity consumption for each date is compared with the baseline electricity consumption. If the absolute value of the deviation between the electricity consumption and the baseline electricity consumption is greater than the standard deviation of a preset multiple, then the date is placed in the abnormal electricity consumption array. The percentage of dates in the abnormal electricity consumption array to the total number of reported dates is taken as the percentage of abnormal dates, and the fluctuation range of the electricity consumption threshold of the power user is determined based on the percentage of abnormal dates. The data on the change in electricity consumption of each power user is determined. Based on the data on the change in electricity consumption of each power user, the stable electricity consumption type of each power user is determined. Based on the power user data and the stable electricity consumption type of each power user, the method for identifying and processing the electricity load characteristics of each power user is determined. Based on the aforementioned identification and processing method, the variation of electricity load characteristics of each electricity user is determined; Based on the changes in the electricity load characteristics of each power user and the identification and processing method, an update processing method for each power user is determined, and based on the update processing method, the characteristic identification and analysis users among the power users are identified. The power consumption threshold of the user is dynamically updated based on the user update processing method described above. Based on the dynamic update results of the electricity consumption threshold and the changes in the electricity load characteristics of each electricity user, differentiated early warning processing is carried out for the electricity user. The electricity users include: feature identification and analysis users, which are electricity users determined based on the update processing method and have the condition of constructing electricity consumption thresholds based on time period groups; Users are identified and processed, and the identified and processed users are power users with historical identification and processing records that have electricity load characteristics; users are not identified and processed, and the users are power users without historical identification and processing records that do not have electricity load characteristics. For the feature-identified users, early warning is issued based on the dynamic update of their electricity consumption threshold. For other users among the identified users, early warning is issued based on the proportion of abnormal dates and changes in electricity load characteristics. For the non-identified users, early warning is issued based on the proportion of abnormal dates.

2. The data-driven and dynamic threshold-based power consumption anomaly diagnosis method as described in claim 1, characterized in that, The baseline electricity consumption of the electricity user is determined based on the average electricity consumption over different dates.

3. The data-driven and dynamic threshold-based abnormal power consumption diagnosis method as described in claim 1, characterized in that, The process of updating the electricity consumption threshold for the aforementioned electricity users specifically includes: Based on the electricity consumption data of electricity users on different dates, the mean and standard deviation are calculated and used as the benchmark electricity consumption and benchmark standard deviation for the electricity users. The electricity consumption for each date is compared with the baseline electricity consumption. If the absolute value of the deviation between the electricity consumption and the baseline electricity consumption is greater than the standard deviation of a preset multiple, then the date is placed in the abnormal electricity consumption array. The percentage of dates in the abnormal electricity consumption array to the total number of reported dates is taken as the percentage of abnormal dates, and the fluctuation range of the electricity consumption threshold of the power user is determined based on the percentage of abnormal dates. If the percentage of abnormal dates is not within the preset percentage of abnormal dates, the abnormal electricity consumption situation of the power user is determined according to the electricity consumption abnormality early warning method. When no early warning is issued, the fluctuation range of the power consumption threshold of the power user is still updated.

4. The data-driven and dynamic threshold-based abnormal power consumption diagnosis method as described in claim 3, characterized in that, Based on the deviation from the benchmark electricity consumption, the abnormal electricity consumption is sorted from smallest to largest. The maximum value of the abnormal electricity consumption array for a specified number of days above the benchmark electricity consumption and the minimum value of the abnormal electricity consumption array for a specified number of days below the benchmark electricity consumption are selected as the fluctuation range of the electricity consumption threshold for the electricity user.

5. The data-driven and dynamic threshold-based power consumption anomaly diagnosis method as described in claim 1, characterized in that, The method for determining the identification and processing method of the electricity load characteristics of the electricity user is as follows: Based on the different electricity consumption stability types of electricity users, determine the stability weight value of the electricity users, and based on the average of the stability weight values ​​of different electricity users, determine the load stability value; Based on the electricity user data, the number of electricity users is determined; A method for identifying and processing the electricity load characteristics of the electricity users by utilizing the number of electricity users, load stability values, and electricity consumption stability types of the electricity users.

6. The data-driven and dynamic threshold-based power consumption anomaly diagnosis method as described in claim 5, characterized in that, If the number of electricity users is less than a preset threshold for the number of electricity users, then the method for identifying the electricity load characteristics of all electricity users is a lenient identification method.

7. The data-driven and dynamic threshold-based abnormal power consumption diagnosis method as described in claim 1, characterized in that, The variation in the electricity load characteristics of the power users is determined based on the deviation of the data composed of different time periods under different electricity load characteristics on different dates.

8. The data-driven and dynamic threshold-based abnormal power consumption diagnosis method as described in claim 1, characterized in that, The method for determining the early warning processing targets for the aforementioned power users is as follows: Based on the updated data of the feature-identified users, the proportion of the feature-identified users among the power users is determined, and the proportion of the feature-identified users among the power users is used as the identification and analysis proportion. Based on the changes in the electricity load characteristics of the power users, the average value of time period groups on different dates in the most recent identification and analysis process of the identified users is determined, and the average value of time period groups on different dates in the most recent identification and analysis process of the identified users is taken as the number of fluctuation groups of the identified users. By using the identification and analysis ratio and the number of fluctuation groups of different identification and processing users, the early warning processing targets of the power users are determined.

9. A data-driven and dynamic threshold-based power consumption anomaly diagnosis system, employing the data-driven and dynamic threshold-based power consumption anomaly diagnosis method according to any one of claims 1-8, specifically comprising: Threshold update module, identification processing module, early warning processing module; The threshold update module updates the electricity consumption threshold of the power user. The identification processing module is responsible for determining the identification processing method for the electricity load characteristics of the power users; The early warning processing module is responsible for determining the early warning processing targets for the power users.

Images