Abnormal power consumption behavior intelligent identification method fusing power consumption trajectory clustering and default feature rules

Abstract

The application discloses a method for intelligently identifying abnormal power utilization behavior by fusing power utilization track clustering and default feature rules, and relates to the technical field of intelligent power utilization monitoring, and comprises the following steps: establishing a unified time reference line in the process of intelligently identifying abnormal power utilization behavior, arranging power utilization data from different power equipment in time sequence, and extracting the change rhythm of power fluctuation, so that all power utilization data have a unified time starting point and time step length. By establishing a unified time reference line, the application eliminates the time difference between the output of clustering analysis and the default rule engine, ensures accurate comparison within the same time range, avoids false alarms, and improves the accuracy of abnormal identification. In addition, through short-time delay processing and smooth calculation, the system effectively balances instantaneous fluctuation and long-term trend, stabilizes the determination result, reduces fluctuation, and enhances the stability of the system through dynamic adjustment, thereby reducing manual intervention.

Description

Technical Field

[0001] This invention relates to the field of intelligent electricity consumption monitoring technology, specifically to an intelligent identification method for abnormal electricity consumption behavior that integrates electricity consumption trajectory clustering and default feature rules. Background Technology

[0002] The integration of electricity consumption trajectory clustering and default feature rules refers to combining time-series pattern analysis of long-term user electricity consumption data with expert-defined default feature discrimination rules during the identification of abnormal electricity consumption behavior. Through the fusion of algorithmic clustering and knowledge rules, multi-dimensional anomaly detection is achieved. Specifically, the electricity consumption trajectory clustering model first performs cluster analysis on the electricity consumption curves of different types of users, such as commercial, industrial, and residential users, extracting their normal electricity consumption trajectory features to establish a behavioral profile for each user type. Then, these clustering results are combined with default feature rules (such as sudden drops in electricity consumption exceeding thresholds, imbalance between live and non-live current, and abnormal meter opening) to re-screen and cross-validate users deviating from normal trajectories, determining whether there is a tendency for default or electricity theft. The resulting intelligent identification of abnormal electricity consumption behavior, through joint modeling of big data, machine learning, and rule engines, performs real-time monitoring and self-learning analysis of user electricity consumption behavior. It quickly identifies abnormal behaviors such as abnormal operation, hidden default, and malicious meter tampering from complex electricity consumption fluctuations, achieving intelligent judgment and early warning of electricity theft and default.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, abnormal power consumption identification methods that integrate clustering models and default rule engines typically employ a phased triggering mechanism. The clustering model outputs trajectory deviation results over a longer time window, while the default rule engine outputs immediate alarm results via short-cycle rule scanning. However, because the execution cycles of the two models are not unified on the time axis, the judgment windows overlap, leading to contradictory judgments from both models within the same time period. When the clustering model is still outputting a normal result based on the stable trajectory of the previous cycle, while the rule engine has already triggered an abnormal alarm due to a sudden change in load, the system cannot correctly determine the final state, causing the judgment result to switch repeatedly between adjacent cycles, resulting in periodic false alarms. This problem has not been effectively solved in existing technologies, easily causing continuous oscillations in the abnormal identification signal, leading to frequent alarm triggering. In severe cases, it can cause instability in the default identification link, sample contamination during the model's self-learning phase, and incorrect assignment of on-site inspection tasks, thus affecting the reliability and practicality of the intelligent identification system.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent identification method for abnormal electricity consumption behavior that integrates electricity consumption trajectory clustering and default feature rules, so as to solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent identification method for abnormal electricity consumption behavior that integrates electricity consumption trajectory clustering and default feature rules, comprising the following steps:

[0008] In the process of intelligent identification of abnormal electricity consumption behavior, a unified time reference line is established. Electricity consumption data from different power devices are sorted in chronological order and the rhythm of power fluctuations is extracted so that all electricity consumption data have a unified time starting point and time step, providing a unified time benchmark for subsequent cluster analysis and rule judgment.

[0009] Based on a unified time reference line, the trajectory results output by cluster analysis and the alarm information generated by the default feature rule judgment are reordered in the same time order, and comparative judgment is performed within the same time range to form a synchronous judgment interval, which is used to eliminate the time difference between the output of the clustering model and the output of the rule engine.

[0010] Based on the synchronous judgment interval setting result conflict handling area, a short time delay processing is implemented for the judgment information where there is a contradiction between the cluster analysis result and the rule judgment result, and the output of the immediate abnormal result is temporarily suspended. During the short time delay stage, the sudden fluctuation signal is balanced and a time delay table containing time correction parameters is generated.

[0011] Based on the time delay table, the changes in electricity consumption characteristics that occur in a short period of time are smoothed and the instantaneous change trend is integrated with the long-term stable trend to obtain a continuous and stable judgment trajectory, which is used to avoid repeated fluctuations in the abnormal identification results.

[0012] Based on the continuous and stable judgment trajectory, dynamic adjustment is performed. Reverse time adjustment and periodic smoothing are carried out according to the time correction parameters in the time delay table. The overlapping judgment time periods are gradually removed through the information feedback mechanism, so that the intelligent identification system for abnormal electricity consumption behavior can continuously output stable and reliable identification results during dynamic operation.

[0013] Preferably, the steps for establishing a unified time reference line during the intelligent identification of abnormal electricity consumption behavior are as follows:

[0014] The raw electricity consumption data collected from different power devices is time-stamped and organized. Timestamp information is extracted from the raw data of each device, and all timestamps are converted into a unified reference time format so that the time points of all electricity consumption data records are aligned to a unified time axis.

[0015] After the time stamps are sorted, the sorted data is rearranged in chronological order. Time sequence deviations are identified by scanning and the data order is adjusted. Missing time periods are filled with adjacent data periods to keep the time series continuous.

[0016] After obtaining continuous time series data, the power fluctuation rhythm on the time reference line is extracted, the power change amplitude and trend in each time period are analyzed, and a power fluctuation rhythm curve reflecting the load change pattern is generated.

[0017] After extracting the power fluctuation rhythm, the time step of the unified time reference line is dynamically adjusted according to the periodic characteristics of the power fluctuation rhythm, so that the time reference line has flexible time resolution in different time periods, providing a unified time benchmark for subsequent cluster analysis and rule judgment.

[0018] Preferably, the steps for reordering the trajectory results output by cluster analysis and the alarm information generated by the default feature rule judgment according to the same time sequence based on a unified time reference line are as follows:

[0019] The trajectory results output by cluster analysis are arranged in chronological order according to a unified time reference line. By connecting the timestamp information in the trajectory results with the unified time reference line, the trajectory output at each time point can be mapped to a unified time base.

[0020] The alarm information generated by the judgment of the default feature rules is sorted by time, and each alarm information is mapped to a unified time reference line through timestamp mapping, so that the cluster analysis results and alarm information are consistent in time.

[0021] After sorting the trajectory results and alarm information by time, the data of the two are compared synchronously. At the same time point, the cluster analysis results and the judgment results of the alarm information are compared to determine abnormal behavior.

[0022] Based on the results of synchronous comparison, a synchronous judgment interval is formed. Within this judgment interval, abnormal behavior is confirmed according to the correspondence between cluster analysis results and alarm information, ensuring the time synchronization and stability of anomaly identification and judgment.

[0023] Preferably, when forming a synchronous judgment interval, the time correspondence between the cluster analysis results and the judgment results of the default feature rules is continuously compared. Consistent results are directly confirmed within the synchronous judgment interval, and results with differences are delayed. The final anomaly identification result is output after the cluster analysis results and the judgment results of the default feature rules reach consistency within the time window, so as to ensure the continuity and stability of anomaly judgment.

[0024] Preferably, the steps for setting the result conflict handling area based on the synchronization judgment interval are as follows:

[0025] Determine the time range for conflict handling, define a time period near the point where the cluster analysis results and rule judgment results contradict each other, and dynamically set the conflict handling area according to the fluctuation characteristics of power data, so that the cluster analysis results and rule judgment results temporarily suspend the output of abnormal judgments during the conflict handling stage.

[0026] Implement short-time delay processing, delay the time point of the conflict within the conflict processing area, temporarily suspend the output of abnormal results through the delay mechanism, and adjust the delay time length according to the changes in power load;

[0027] Balance calculations are performed on sudden fluctuation signals during short time delays. By analyzing the power fluctuation trend within a time period, short-term fluctuations are compared and balanced with historical fluctuation patterns.

[0028] A time delay table is generated based on the balanced data, and time correction parameters and delay adjustment coefficients are recorded to provide a basis for subsequent time synchronization and anomaly detection.

[0029] Preferably, during the generation of the time delay table, the power fluctuation data after balancing within the conflict handling area is analyzed hourly, and corresponding time correction parameters are generated based on the delay time length and correction coefficient at each time point. The time correction parameters are then stored in the time delay table in chronological order, so that subsequent time synchronization and anomaly identification result adjustments have continuity and traceability.

[0030] Preferably, the steps for smoothing changes in electricity consumption characteristics occurring within a short period of time based on the time delay table are as follows:

[0031] Based on the time correction parameters in the time delay table, the changes in electricity consumption characteristics that occur in a short period of time are smoothed out. The time correction parameters buffer short-term fluctuations and balance instantaneous load changes within the time range.

[0032] After smoothing, short-term fluctuations are integrated with long-term stable trends. By analyzing the correspondence between short-term fluctuations and long-term trends, abnormal behavior and temporary fluctuations are distinguished.

[0033] After integrating short-term fluctuations with long-term stable trends, a continuous and stable judgment trajectory is generated. Short-term fluctuations are smoothed through time series analysis to preserve long-term trends.

[0034] By utilizing a continuous and stable judgment trajectory, repeated fluctuations in abnormal identification results are avoided. The judgment result is kept stable by dynamically adjusting the judgment window, and the real electricity consumption behavior identification result is continuously output.

[0035] Preferably, the steps for performing dynamic adjustment based on a continuous and stable judgment trajectory are as follows:

[0036] Based on the time correction parameters in the time delay table, reverse time adjustment is performed, and time deviation is corrected by backtracking the time series to make each data point consistent with the actual time point of power load fluctuation;

[0037] After completing the reverse time adjustment, periodic smoothing is performed. By analyzing the power fluctuation amplitude and frequency, the time series is smoothed to reduce the impact of sudden fluctuations and maintain data continuity.

[0038] An information feedback mechanism is initiated on the smoothed judgment trajectory to feed back the recognition results of the previous stage to the subsequent judgment stage in order to optimize the recognition criteria and achieve self-adjustment;

[0039] By calculating the overlapping parts of the judgment time periods and gradually removing the overlapping judgment time periods, only the valid judgment intervals are retained to optimize the final output results and ensure the stability and accuracy of the anomaly identification process.

[0040] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0041] This invention effectively eliminates conflicts caused by time differences between models by establishing a unified time reference line and synchronizing the clustering analysis results with the output of default feature rules. Through this method, the system can ensure accurate comparison and judgment within the same time range, avoiding false alarms caused by time misalignment. This synchronization process greatly improves the accuracy of anomaly identification, enabling the system to stably identify abnormal electricity consumption behavior and avoiding misjudgments and repeated fluctuations.

[0042] This invention further stabilizes the judgment results by combining short-time delay processing and smoothing calculation. During the balancing process of sudden fluctuation signals, the system can automatically adjust the relationship between instantaneous fluctuations and long-term stable trends, thereby avoiding abnormal identification result fluctuations caused by short-term fluctuations. This dynamic adjustment mechanism makes the entire intelligent identification process smoother, effectively improving the system's stability and reliability, and reducing unnecessary human intervention. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0044] Figure 1 This is a flowchart of the intelligent identification method for abnormal electricity consumption behavior that integrates electricity consumption trajectory clustering and default feature rules according to the present invention. Detailed Implementation

[0045] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0046] This invention provides, for example Figure 1 The intelligent identification method for abnormal electricity consumption behavior, which integrates electricity consumption trajectory clustering and default feature rules, includes the following steps:

[0047] In the process of intelligent identification of abnormal electricity consumption behavior, a unified time reference line is established. Electricity consumption data from different power devices are sorted in chronological order and the rhythm of power fluctuations is extracted so that all electricity consumption data have a unified time starting point and time step, providing a unified time benchmark for subsequent cluster analysis and rule judgment.

[0048] When conducting intelligent identification of abnormal electricity consumption behavior, it is first necessary to ensure that electricity consumption data from different power devices are processed uniformly over time, so that subsequent clustering analysis and rule judgment can be performed on the same time reference. To achieve this goal, a series of operational steps are required to ensure that all electricity data is consistent across the timeline, thus providing a stable foundation for subsequent analysis and judgment. The specific implementation process is as follows:

[0049] The raw electricity consumption data collected from various power devices is time-stamped and organized. This process first requires extracting the timestamp information for each record from the raw data of each device. Due to differences in time synchronization accuracy, sampling frequency, and data recording starting points among different devices, the timestamps of the electricity data from different devices are inconsistent. To solve this problem, a method is adopted to convert all timestamps into a unified reference time format and uniformly adjust them to a standardized time starting point. Specifically, the timestamps at the device end are offset according to a unified benchmark, so that the time points of all electricity data records are aligned to a unified time axis. This unified time benchmark point is the starting point of the unified time reference line. The timestamp of each electricity consumption record is mapped to a specific time position on this time axis, thereby ensuring that records from different devices have a consistent starting point in the time dimension, laying the foundation for subsequent analysis.

[0050] After completing the time stamping, the data is rearranged chronologically. Since timestamps from different devices may have intervals during data collection, or slight misalignments in time sequence may occur due to data transmission delays in some cases, data processing is necessary to ensure all records are arranged in strict chronological order. First, the entire dataset is scanned to identify chronological discrepancies, and data with discrepancies is adjusted. For records with intermittent missing data or brief data interruptions due to communication problems, the gaps are filled using adjacent data periods. During the filling process, a smoothing algorithm is used to transition the values ​​of adjacent data, allowing the time interval of the missing portion to naturally transition to the next time period, avoiding discontinuities in the time series caused by missing data. Through this processing, the chronological order of the data is perfected, and all timestamps accurately correspond to a specific point in time on a unified time reference line, providing a consistent and complete temporal basis for subsequent time series analysis.

[0051] After obtaining the integrated time-series data, the extraction of the power fluctuation rhythm begins. The power fluctuation rhythm reflects the dynamic changes in electricity load over time and is crucial for subsequent electricity consumption trajectory clustering analysis. In this step, the power changes within each time period are first calculated, determining the power increase or decrease at each moment. The amplitude and trend of power fluctuations are analyzed by comparing power values ​​at adjacent time points. In particular, abrupt power fluctuations (such as drastic instantaneous load fluctuations) require special labeling and analysis to identify abnormal electricity consumption behavior. The power fluctuation rhythm corresponding to each segment of electricity consumption data represents the dynamic changes in electricity load within a specific time period. To extract valuable electricity consumption rhythm information from a global perspective, the power fluctuations of all recorded data need to be comprehensively considered. Through a comprehensive analysis of these fluctuation amplitudes and rates of change, the power fluctuation rhythm curve for each time period is obtained. This rhythm curve not only clearly shows the fluctuations in electricity load but also reflects the changing patterns of electricity load across different time periods, providing a more accurate temporal structure for subsequent anomaly identification.

[0052] After extracting the power fluctuation rhythm, the time step is dynamically adjusted. Because power load fluctuations are time-sequential, changes may be drastic in certain time periods, while remaining relatively stable in others. Therefore, the time step of the time reference line is not fixed but dynamically adjusted based on the actual power change rhythm. In this stage, based on the characteristics of the extracted power fluctuation rhythm, the periodicity of power fluctuations is analyzed to determine the most suitable time step for each time period. Specifically, shorter time intervals are used for periods with larger fluctuations to capture more details, while the time step can be appropriately increased for periods with relatively stable changes to reduce computational load. This step not only avoids excessive processing burden due to frequent calculations but also adaptively adjusts the time scale according to the data's changing characteristics, making data analysis more accurate. By dynamically adjusting the time step, the time reference line has flexible time resolution in different time periods, ensuring that subsequent abnormal power consumption behavior identification achieves higher accuracy during periods of high fluctuation and maintains high computational efficiency during periods of low fluctuation.

[0053] Based on a unified time reference line, the trajectory results output by cluster analysis and the alarm information generated by the default feature rule judgment are reordered in the same time order, and comparative judgment is performed within the same time range to form a synchronous judgment interval, which is used to eliminate the time difference between the output of the clustering model and the output of the rule engine.

[0054] After establishing a unified time reference line, the next crucial step is to reorder the trajectory results output by cluster analysis and the alarm information generated by the default feature rule engine in the same chronological order. This allows for comparison and judgment within the same time frame, forming a synchronized judgment interval and eliminating the impact of time differences. The core purpose of this step is to ensure time alignment across multiple data sources with different time granularities, thereby making subsequent anomaly detection more accurate and stable. The specific steps are as follows:

[0055] The trajectory results output from cluster analysis need to be organized chronologically according to a unified time reference line. The trajectory results from cluster analysis typically contain electricity consumption behavior patterns of different types of users, which are derived from historical electricity consumption data. Since the clustering process usually uses a relatively long time window for data aggregation, there may be time delays or inconsistencies in the time intervals of different users' trajectories. Therefore, it is first necessary to align the timestamp information in these trajectory results with the previously established unified time reference line to ensure that the trajectory output at each time point corresponds to a unified time base. The key technical approach in this step is to use a unified time standard to map the time information of each trajectory onto a unified time axis and to arrange the trajectory data from different time periods sequentially, thereby ensuring that each result from the clustering output has a clear time indication and achieves temporal consistency with data from other data sources.

[0056] The alarm information generated by the default feature rule engine is sorted by time. The default feature rule engine monitors for potential abnormal behavior based on real-time power data, and the generated alarm information is typically output immediately within a short time window. Since the trajectory results output by cluster analysis and the alarm information output by the default feature rule engine are usually based on different time intervals, the outputs of these two types of data may not be consistent within the same time period, and there may even be time misalignments. To eliminate this difference, the alarm information output by the rule engine first needs to be processed to align it with the cluster analysis results according to a unified time reference line. Specifically, all generated alarm information should be sorted by timestamp, mapping each alarm information to its corresponding time point under a unified time reference line, thereby ensuring that the alarm information corresponds one-to-one with the cluster analysis results in time, forming an ordered time series.

[0057] After sorting the trajectory results and alarm information by time, the next step is to synchronize and compare the two sets of data. At this point, all trajectory data and alarm information are arranged in chronological order, and the time points are perfectly aligned. To make comparisons within the same time frame, the trajectory results output by the clustering analysis and the alarm information output by the rule engine must be synchronized. This means that at each identical time point, the system needs to compare whether there are any contradictions between the clustering analysis results and the alarm information, and determine whether the abnormal behavior during that period should be marked as a confirmed anomaly or abnormal behavior. For example, at a certain moment, the clustering model may consider the user's electricity consumption trajectory to be normal, but the rule engine may trigger an abnormal alarm due to a sudden change in load. In this case, the system will make a judgment based on specific rules, combined with the synchronized comparison results in chronological order, to decide whether to classify the electricity consumption behavior at that moment as abnormal. Through this synchronized comparison step, data conflicts and judgment biases caused by time differences can be effectively eliminated, thereby improving the accuracy of the final judgment result.

[0058] Based on the results of the synchronous judgment, a stable synchronous judgment interval is formed. Establishing this synchronous judgment interval is the core of the entire process, effectively avoiding repeated judgments caused by time errors. Within the synchronous judgment interval, the clustering analysis results are compared with the rule engine alarm information, and confirmation is made according to the set rules. If the clustering model output results are consistent with the rule engine output results within a certain time period, the abnormal behavior in that period can be confirmed and output. If there is a difference between the two results, the system will perform cross-validation or delay processing according to predetermined rules, and output the final judgment only after all results reach consensus within a defined time window. This process ensures that in data analysis with a large time span, the system can output the final abnormal electricity consumption behavior judgment result in a short time and stably, avoiding periodic false alarms or repeated state switching problems caused by time differences.

[0059] Based on the synchronous judgment interval setting result conflict handling area, a short time delay processing is implemented for the judgment information where there is a contradiction between the cluster analysis result and the rule judgment result, and the output of the immediate abnormal result is temporarily suspended. During the short time delay stage, the sudden fluctuation signal is balanced and a time delay table containing time correction parameters is generated.

[0060] After confirming the synchronous judgment interval, the next crucial step is to handle potential contradictions between the clustering analysis results and the rule-based judgment results. Since clustering analysis typically uses a relatively long time window, while rule-based judgment relies on short-term real-time data, their judgment results may contradict each other within certain time periods. To ensure the accuracy and stability of the final abnormal electricity consumption behavior identification results, these contradictory judgment information must be effectively processed. The specific steps are as follows:

[0061] The time frame for conflict handling is determined. The conflict handling area is set based on the synchronous judgment interval; specifically, it's a suitable time period defined around the point where two results contradict each other. Because the time intervals between clustering analysis results and rule judgment outputs differ, conflicts may occur simultaneously. To address this, the system sets a reasonable time interval to handle potential conflicts based on the fluctuation patterns in historical data. The length of the conflict handling area is dynamic, determined by the fluctuation characteristics of the power data. Within this time frame, the judgment results of the clustering model and rule engine are temporarily "frozen," meaning abnormal judgment results are not immediately output; instead, the system enters the conflict handling phase. This phase aims to avoid misjudgments caused by sudden load fluctuations or short-term data fluctuations. Setting a conflict handling area helps ensure that data within the time frame undergoes stable calculations, thereby improving the reliability of the judgment results.

[0062] Implement short-time delay processing. After identifying the conflict handling area, the next step is to implement short-time delay processing for the moments when clustering analysis and rule judgment conflict. When the results of clustering analysis and the output of rule judgment contradict each other, the system does not immediately output the final anomaly judgment, but instead employs a short-time delay mechanism. The purpose of the delay is to allow the system sufficient time to assess and mitigate potential anomalies. During this process, the system temporarily ignores the immediate anomaly output and enters a delay state. The length of the delay phase is not fixed but is automatically adjusted based on the dynamic changes in power load. For example, in some cases, when there are short-term, sharp fluctuations in power load, the delay duration may be longer to allow for judgment after the fluctuations stabilize; while in other cases, if the fluctuations are small, the delay time can be appropriately shortened. This short-time delay processing ensures that the system is not misled by instantaneous fluctuations, but rather balances and suppresses short-term anomaly judgments through delay.

[0063] A balance calculation is performed for sudden fluctuation signals. Sudden fluctuation signals are usually short-term changes caused by instantaneous load fluctuations or equipment startup. These fluctuations may not represent true abnormal behavior, so a balance calculation is necessary. During the short-delay phase, the system analyzes power fluctuation data within a time period to determine whether these fluctuations belong to normal load fluctuations. At this time, the system evaluates the dynamic trend of the data, identifies short-term power changes, and performs smoothing. The key to the balance calculation is to compare these fluctuations with normal fluctuation patterns in historical data to determine whether the fluctuation is temporary. If the fluctuation exceeds the normal fluctuation range, it may be considered abnormal, and an abnormal judgment will be output; if the fluctuation is within the normal range, it will be excluded to avoid misjudgment. In this balance phase, the system uses a weighted average of the short-term fluctuations and long-term trends of the data to ensure that the final judgment accurately reflects the actual electricity consumption behavior.

[0064] A time delay table containing time correction parameters is generated. Through the preceding steps, fluctuations within the short time delay phase have been balanced and adjusted. The system generates a time delay table based on the adjusted data. This table records the time correction parameters at each point in time within the conflict resolution area, including the delay length and the corresponding adjustment coefficient. The time delay table not only records the fluctuation adjustments during conflict resolution but also provides a basis for subsequent correction processes. Using this table, the system can make corresponding adjustments to data in other time periods, ensuring the continuous consistency and accuracy of anomaly detection results. The generated time delay table not only provides the necessary parameters for subsequent time synchronization but also provides a scientific basis for the final judgment of abnormal electricity consumption behavior, ensuring the stability of the entire identification process in a dynamic environment.

[0065] Based on the time delay table, the changes in electricity consumption characteristics that occur in a short period of time are smoothed and the instantaneous change trend is integrated with the long-term stable trend to obtain a continuous and stable judgment trajectory, which is used to avoid repeated fluctuations in the abnormal identification results.

[0066] After generating the time delay table, the next crucial step is to smooth out short-term changes in electricity consumption characteristics, thereby fusing instantaneous trends with long-term stable trends. The purpose of this step is to avoid overreacting to sudden fluctuations when identifying abnormal electricity consumption behavior, thus ensuring a stable and continuous final judgment trajectory. By fusing instantaneous fluctuations and long-term trends, the system can accurately identify genuine anomalies without being misled by short-term fluctuations and making repeated judgments. The specific implementation steps are as follows:

[0067] Based on the time correction parameters provided by the time delay table, the system smooths short-term changes in electricity consumption characteristics. In this step, the time delay table records the delay time and adjustment parameters for each conflict handling area, providing a basis for subsequent smoothing calculations. Short-term electricity consumption characteristics typically include sudden load fluctuations, instantaneous power increases or decreases, etc., which may have an excessive impact on the final anomaly judgment result. Therefore, the system first uses the time correction parameters in the delay table to buffer short-term fluctuations. Specifically, the system applies the time correction parameters recorded in the time delay table to short-term data to mitigate the impact of these short-term fluctuations, preventing them from directly affecting the final judgment result. Through this processing, the system ensures that the electricity consumption characteristics of each time period are appropriately smoothed when fluctuations are severe, avoiding over-response to instantaneous load changes.

[0068] After smoothing, the system integrates short-term fluctuations with long-term stable trends. Short-term fluctuations may be caused by factors such as equipment startup or instantaneous load increases, while long-term stable trends reflect users' normal electricity consumption behavior and habits. Therefore, after smoothing short-term fluctuations, the system further analyzes the relationship between these fluctuations and long-term stable trends. By comparing and integrating short-term fluctuations with long-term stable trends, the system can effectively distinguish between genuine abnormal behavior and temporary fluctuations. For example, if electricity consumption fluctuations change drastically in the short term but are consistent with historical long-term trends, these changes may only be temporary load fluctuations, rather than genuine abnormal electricity consumption behavior. Through this integration process, the system can combine short-term fluctuations and long-term trends to arrive at more accurate judgments and avoid misjudgments caused by accidental fluctuations.

[0069] After integrating short-term fluctuations with long-term stable trends, a continuous and stable judgment trajectory is generated. The goal of this stage is to eliminate the interference of short-term fluctuations on the system's judgment results, forming a continuous and stable judgment trajectory. In this step, the system performs further time-series analysis on the smoothed electricity consumption data, smoothing out short-term fluctuations to retain the long-term trend reflecting normal electricity consumption behavior. In this way, the judgment trajectory generated by the system accurately reflects the user's actual electricity consumption without being affected by short-term fluctuations. The generated judgment trajectory not only stably reflects the changing trends of user electricity consumption behavior but also provides a more stable and reliable basis for subsequent anomaly judgments.

[0070] A continuous and stable judgment trajectory is used to avoid repeated fluctuations in anomaly identification results. In actual power consumption, sudden fluctuations often lead to repeated judgments by the identification system, especially without smoothing calculations, where short-term fluctuations can easily be misjudged as abnormal behavior. Therefore, through the aforementioned smoothing processing and trend fusion, the system can find a balance between short-term fluctuations and long-term trends, avoiding overreaction to short-term load changes. In this process, the system dynamically adjusts the judgment window to ensure that even with instantaneous fluctuations, the judgment results remain stable, and the final output of abnormal power consumption behavior identification results can consistently and stably reflect the real situation. This stability is guaranteed through the application of a time delay table, the fusion of short-term fluctuations and long-term trends, and multiple calculations and processing to generate a continuous judgment trajectory.

[0071] Based on the continuous and stable judgment trajectory, dynamic adjustment is performed. Reverse time adjustment and periodic smoothing are carried out according to the time correction parameters in the time delay table. The overlapping judgment time periods are gradually removed through the information feedback mechanism, so that the abnormal electricity consumption behavior intelligent identification system can continuously output stable and reliable identification results during dynamic operation.

[0072] After generating continuous and stable judgment trajectories, the next step is to dynamically adjust these trajectories to ensure that the identification results of abnormal electricity consumption behavior can be output stably and continuously in a variable power load environment. The key to this process is to achieve the stability of the results through reverse time adjustment, periodic smoothing, and information feedback mechanisms, thereby eliminating potentially overlapping judgment time periods and avoiding misjudgments and fluctuations. The specific steps are as follows:

[0073] Based on the time correction parameters in the time delay table, reverse time adjustment is performed. The time delay table records detailed time delay information for each judgment point, including the error at that time point and the correction parameters. To ensure the accuracy of time judgment, the system first backtracks to each time point that needs adjustment. Specifically, after the correction factor for conflicting time periods is applied, the system performs reverse time correction in the "past" time series. This process first identifies moments that do not match the time markers in the time delay table, and then adjusts the original time sequence in reverse according to these markers. By backtracking to the accurate time point, the system ensures that each data point is completely consistent with the actual power load fluctuation time point. The reverse adjustment process not only corrects time deviations caused by system delays or external factors, but also ensures that each data segment is correctly positioned in the real time series. This step ensures that the system can rearrange the time axis so that all subsequent judgments can reflect a more accurate correlation with the actual power load.

[0074] After completing the reverse time adjustment, periodic smoothing is performed. This step aims to integrate short-term fluctuations with long-term stable trends, eliminating unnecessary interference caused by sudden load fluctuations. During this process, the system smooths the data based on the amplitude and frequency of power fluctuations within each time period. Specifically, the system calculates the load change curves for each time period and smooths the data to reduce the impact of any sudden fluctuations. If there are sudden increases or decreases in power load during certain time periods, the system mitigates these fluctuations by calculating a weighted average of adjacent time periods, limiting the impact of these abnormal fluctuations to an acceptable range. At the same time, the system does not ignore these sudden fluctuations but adjusts them flexibly to ensure that the anomaly detection process can still identify genuine abnormal behavior without over-reliance on short-term fluctuations. Ultimately, through this periodic smoothing process, the data sequence exhibits stable and continuous characteristics, ensuring that subsequent anomaly identification is no longer interfered with by short-term load fluctuations.

[0075] An information feedback mechanism is initiated on the smoothed judgment trajectory. The purpose of this mechanism is to ensure that the identification results at each stage influence subsequent judgments, thereby enabling the system to self-adjust. During dynamic adjustment, the system fine-tunes subsequent judgment criteria based on the output results of the previous stage. Specifically, when the system identifies abnormal behavior within a certain time period, this information is immediately sent back to the detection module for the next time period to adjust the identification criteria for abnormal electricity consumption behavior. For example, if the system determines a certain fluctuation as abnormal electricity consumption behavior within a certain time period, this judgment will be fed back to the subsequent data processing stage to adjust the identification criteria for similar fluctuations in subsequent stages. This feedback mechanism allows the system to continuously optimize and adjust based on previous identification results, ensuring judgment accuracy in long-term operation and reducing the chain reaction caused by early judgment errors on subsequent identification results. By continuously updating and adjusting the identification results at each time point, the information feedback mechanism enables the system to maintain high identification accuracy in complex electricity consumption behaviors.

[0076] The final output is optimized by removing overlapping judgment time periods. Overlapping judgment time periods may be caused by time period conflicts in multiple judgments, especially when conflicts occur between the outputs of the clustering model and the rule engine, often resulting in overlapping judgment results. This overlap not only affects subsequent identification judgments but may also lead to unnecessary alarms and false alarms. At this stage, the system calculates the overlapping parts between time periods based on previous judgment trajectories and gradually removes these overlapping judgment time periods. Specifically, the system removes duplicate data of these time periods from the final output, retaining only the valid judgment intervals. This operation not only helps reduce the burden of redundant calculations but also avoids false alarms caused by overlapping judgments. Through this processing, the system can output clearer and more accurate anomaly identification results, avoiding repeated fluctuations and unnecessary duplicate judgments during the judgment process.

[0077] This invention effectively eliminates conflicts caused by time differences between models by establishing a unified time reference line and synchronizing the clustering analysis results with the output of default feature rules. Through this method, the system can ensure accurate comparison and judgment within the same time range, avoiding false alarms caused by time misalignment. This synchronization process greatly improves the accuracy of anomaly identification, enabling the system to stably identify abnormal electricity consumption behavior and avoiding misjudgments and repeated fluctuations.

[0078] This invention further stabilizes the judgment results by combining short-time delay processing and smoothing calculation. During the balancing process of sudden fluctuation signals, the system can automatically adjust the relationship between instantaneous fluctuations and long-term stable trends, thereby avoiding abnormal identification result fluctuations caused by short-term fluctuations. This dynamic adjustment mechanism makes the entire intelligent identification process smoother, effectively improving the system's stability and reliability, and reducing unnecessary human intervention.

[0079] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An intelligent method for identifying abnormal electricity consumption behavior that integrates electricity consumption trajectory clustering and default feature rules, characterized in that: Includes the following steps: In the process of intelligent identification of abnormal electricity consumption behavior, a unified time reference line is established, and the electricity consumption data from different power devices are sorted in chronological order and the rhythm of power fluctuation is extracted so that all electricity consumption data have a unified time starting point and time step. Based on a unified time reference line, the trajectory results output by cluster analysis and the alarm information generated by the default feature rule judgment are reordered in the same time order, and comparative judgment is performed within the same time range to form a synchronous judgment interval. Based on the synchronous judgment interval setting result conflict handling area, a short time delay processing is implemented for the judgment information where there is a contradiction between the cluster analysis result and the rule judgment result, temporarily suspending the output of immediate abnormal results, and performing balance calculation on the sudden fluctuation signal during the short time delay stage to generate a time delay table. Based on the time delay table, smooth calculations are performed on changes in electricity consumption characteristics that occur in a short period of time, and the instantaneous change trend is integrated with the long-term stable trend to obtain a continuous and stable judgment trajectory. Dynamic adjustment is performed based on the continuous and stable judgment trajectory. Reverse time adjustment and periodic smoothing are carried out according to the time correction parameters in the time delay table, and overlapping judgment time periods are gradually removed through the information feedback mechanism.

2. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 1, characterized in that, The steps for establishing a unified time reference line in the intelligent identification of abnormal electricity consumption behavior are as follows: The raw electricity consumption data collected from different power devices is time-stamped and organized. Timestamp information is extracted from the raw data of each device and all timestamps are converted into a unified reference time format. After the time tags are sorted, the sorted data is rearranged in chronological order. Time order deviations are identified by scanning and the data order is adjusted. Missing time periods are filled with adjacent data periods. After obtaining continuous time series data, the power fluctuation rhythm on the time reference line is extracted, the power change amplitude and trend in each time period are analyzed, and a power fluctuation rhythm curve reflecting the load change pattern is generated. After extracting the power fluctuation rhythm, the time step of the unified time reference line is dynamically adjusted according to the periodic characteristics of the power fluctuation rhythm, so that the time reference line has flexible time resolution in different time periods.

3. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 2, characterized in that, The steps for reordering the trajectory results output by cluster analysis and the alarm information generated by the default feature rule judgment according to the same time sequence based on a unified time reference line are as follows: The trajectory results output by cluster analysis are arranged in chronological order according to a unified time reference line. By connecting the timestamp information in the trajectory results with the unified time reference line, the trajectory output at each time point can be mapped to a unified time base. The alarm information generated by the judgment of the default feature rules is sorted by time, and each alarm information is mapped to a unified time reference line through timestamp mapping, so that the cluster analysis results and alarm information are consistent in time. After sorting the trajectory results and alarm information by time, the data of the two are compared synchronously. At the same time point, the cluster analysis results and the judgment results of the alarm information are compared to determine abnormal behavior. Based on the results of synchronous comparison, a synchronous judgment interval is formed. Within this judgment interval, abnormal behavior is confirmed according to the correspondence between cluster analysis results and alarm information.

4. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 3, characterized in that, When forming a synchronous judgment interval, the time correspondence between the cluster analysis results and the judgment results of the default feature rules is continuously compared. Consistent results are directly confirmed within the synchronous judgment interval, while results with discrepancies are delayed. The final anomaly identification result is output after the cluster analysis results and the judgment results of the default feature rules reach consistency within the time window.

5. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 3, characterized in that, The steps for setting up a conflict handling region based on the synchronous judgment interval are as follows: Determine the time range for conflict handling, define a time period near the point where the cluster analysis results and rule judgment results contradict each other, and dynamically set the conflict handling area according to the fluctuation characteristics of power data, so that the cluster analysis results and rule judgment results temporarily suspend the output of abnormal judgments during the conflict handling stage. Implement short-time delay processing, delay the time point of the conflict within the conflict processing area, temporarily suspend the output of abnormal results through the delay mechanism, and adjust the delay time length according to the changes in power load; Balance calculations are performed on sudden fluctuation signals during short time delays. By analyzing the power fluctuation trend within a time period, short-term fluctuations are compared and balanced with historical fluctuation patterns. A time delay table is generated based on the balanced data, and the time correction parameters and delay adjustment coefficients are recorded.

6. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 5, characterized in that, During the generation of the time delay table, the power fluctuation data after balance in the conflict handling area is analyzed hourly. Based on the delay time length and correction coefficient of each time point, the corresponding time correction parameters are generated and stored in the time delay table in chronological order.

7. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 5, characterized in that, The steps for smoothing electricity consumption characteristics changes that occur within a short period of time based on a time delay table are as follows: Based on the time correction parameters in the time delay table, the changes in electricity consumption characteristics that occur in a short period of time are smoothed out. The time correction parameters buffer short-term fluctuations and balance instantaneous load changes within the time range. After smoothing, short-term fluctuations are integrated with long-term stable trends. By analyzing the correspondence between short-term fluctuations and long-term trends, abnormal behavior and temporary fluctuations are distinguished. After integrating short-term fluctuations with long-term stable trends, a continuous and stable judgment trajectory is generated. Short-term fluctuations are smoothed through time series analysis to preserve long-term trends. By utilizing a continuous and stable judgment trajectory, repeated fluctuations in abnormal identification results are avoided. The judgment result is kept stable by dynamically adjusting the judgment window, and the real electricity consumption behavior identification result is continuously output.

8. The intelligent identification method for abnormal electricity consumption behavior based on the fusion of electricity consumption trajectory clustering and default feature rules as described in claim 7, characterized in that, The steps for dynamic adjustment based on a continuously stable trajectory are as follows: Based on the time correction parameters in the time delay table, reverse time adjustment is performed, and time deviation is corrected by backtracking the time series to make each data point consistent with the actual time point of power load fluctuation; After completing the reverse time adjustment, periodic smoothing is performed. By analyzing the power fluctuation amplitude and frequency, the time series is smoothed to reduce the impact of sudden fluctuations and maintain data continuity. An information feedback mechanism is initiated on the smoothed judgment trajectory to feed back the recognition results of the previous stage to the subsequent judgment stage in order to optimize the recognition criteria and achieve self-adjustment; By calculating the overlapping parts of the judgment time periods and gradually removing the overlapping judgment time periods, only the valid judgment intervals are retained to optimize the final output results and ensure the stability and accuracy of the anomaly identification process.

Images