Intelligent electric meter system with abnormal electricity consumption behavior identification function

Abstract

The invention belongs to the technical field of intelligent electric meters, and discloses an intelligent electric meter system with an abnormal power consumption behavior recognition function. The system is composed of a data acquisition module, an electric energy quality monitoring module, a data preprocessing module, a power utilization behavior feature extraction module, a short-time behavior monitoring module, an anomaly detection and diagnosis module, a behavior trend analysis module, an alarm and visualization module and a remote collaborative management module. Through deep mining and learning of historical power utilization data of a user, the system can accurately grasp power utilization habits, time period change rules and load characteristics of the user, along with dynamic evolution of power utilization conditions, the system can automatically optimize an anomaly detection threshold value, limitation brought by a traditional fixed threshold value is abandoned, and through the personalized and adaptive design, the power utilization efficiency of the user is improved. The system can exert the optimal efficiency in different regions and different types of users, and the application range and the practicability of the system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent electric meters, and particularly relates to an intelligent electric meter system with abnormal electric behavior identification function. BACKGROUND

[0002] Under the background of the rapid development of the intelligentization of the power industry, the intelligent electric meter system has become an important basic equipment for the informatization and digitization of the power grid. However, at present, most of the intelligent electric meter systems still mainly stay at the level of basic electric energy metering and simple load monitoring, and cannot meet the abnormal detection needs under the current complex electric environment. With the diversification of user electric modes and the continuous increase of emerging electric scenarios, the electric behavior presents high dynamics and individualization. In particular, new electric stealing methods such as electromagnetic interference, electronic tampering and fake load are continuously emerging, which brings great challenges to the safety and economy of electric power. At the same time, the aging, failure and hidden safety hazards caused by the long-term operation of electric power equipment have increasingly become important factors threatening the safety of the power grid. However, the existing intelligent electric meter system generally lacks real-time monitoring and accurate identification capability for these complex abnormal electric behaviors, which makes it difficult to discover potential risks in time and seriously affects the safety and operation efficiency of the electric power system.

[0003] In terms of abnormal detection technology, the existing system mainly relies on single threshold judgment or simple machine learning method. The fixed threshold cannot be individually adapted to different users, different time periods or special electric power application scenarios, which leads to a significantly high false positive rate and false negative rate, increases the work burden of electric power operation and maintenance personnel, reduces the system reliability, and the basic machine learning algorithm has some improvements, but still has problems such as complex model, insufficient real-time performance, high occupation of computing resources and the like, which cannot adapt to the low-power and low-delay operation requirements of edge devices, and the detection capability for new and small sample abnormal electric behaviors is obviously insufficient. In addition, the current system has a series of technical shortcomings such as low data acquisition frequency, poor data quality, easy interruption of communication, algorithm not supporting edge deployment, inadequate protection of user privacy, high network delay in remote areas and the like, which further limits the actual application effect of the abnormal electric behavior detection technology. SUMMARY

[0004] The purpose of the present application is to provide an intelligent electric meter system with abnormal electric behavior identification function to solve the problems raised in the background.

[0005] In order to achieve the above purpose, the present application provides the following technical scheme: an intelligent electric meter system with abnormal electric behavior identification function, comprising: A data acquisition module: with a second-level sampling period, high-precision sensors are used to collect basic electric parameters such as voltage and current, and real-time data streams are output after signal conditioning, which provides original data support for subsequent modules and is the beginning of system data processing; Power quality monitoring module: receives raw electrical parameters, calculates key indicators using digital signal processing algorithms, provides early warning of abnormalities based on threshold values, and outputs data with quality labels, converting raw data into a sequence with abnormality labels for the data preprocessing module to process; Data preprocessing module: applies filtering algorithms to denoise the labeled raw data sequence, eliminates dimensional effects through standardization processing, and outputs a high-quality data sequence to improve data quality and lay the foundation for electricity behavior feature extraction; Electricity behavior feature extraction module: based on high-quality data, calculates various electricity features, integrates multi-dimensional features using dimension reduction algorithms, forms a feature vector output, extracts key information reflecting electricity behavior from data, and transmits it to the short-term behavior monitoring module; Short-term behavior monitoring module: uses a sliding window to monitor the feature vector, captures sudden abnormal fluctuations, and outputs monitoring results after focusing on abnormal samples to quickly identify short-term electricity abnormalities and provide clues for anomaly detection and diagnosis; Anomaly detection and diagnosis module: receives short-term monitoring results, uses multiple algorithms for joint diagnosis, dynamically adjusts thresholds, and outputs anomaly detection results and type labels to accurately determine abnormal conditions and provide a basis for behavior trend analysis; Behavior trend analysis module: based on anomaly detection results, uses time series models to analyze long-term electricity trends, identifies potential risks, and outputs trend graphs and warning information to mine electricity rules and risks from a long-term perspective and transmit them to the alarm and visualization module; Alarm and visualization module: generates hierarchical alarms based on the severity of anomalies, visualizes data and anomalies in multiple formats, outputs interfaces after alarm aggregation and deduplication, and presents analysis results intuitively to facilitate user understanding while providing information for remote collaborative management; Remote collaborative management module: synchronizes visualization information to the backend, uses data compression and encryption techniques for transmission, and implements multi-meter collaborative analysis and management in the backend to output control instructions and analysis results, completing the full-process closed loop from data collection to remote management.

[0006] Preferably, the data acquisition module comprises: (1) Efficient perception and parameter calculation of the data acquisition module: The data acquisition module, as the data cornerstone of the smart meter system, is responsible for the core task of real-time acquisition of user electricity basic electrical parameters. It collects voltage, current, active power, reactive power, apparent power, and power factor at a sampling period of seconds. The module integrates high-precision voltage transformers, current transformers, and other sensors, which can accurately capture subtle changes in various electrical parameters due to their high sensitivity. Based on the active power calculation formula and the reactive power calculation formula, the module accurately quantifies electrical energy consumption by calculating voltage and current parameters, providing accurate basic data for the system; Active power calculation formula ; In the formula: Active power (unit: W); This is the effective value of the voltage (unit: V). This is the effective value of the current (unit: A). The phase angle between voltage and current (unit: °); Reactive power calculation formula ; Reactive power (unit: var); (2) Data processing and output transmission of the data acquisition module: The acquired raw electrical parameter data will undergo preliminary processing within the module. By using signal conditioning techniques such as amplification and filtering, some interference signals are effectively removed, improving data quality. Subsequently, the processed data is output in the form of a real-time data stream through a high-speed data transmission interface. This high-quality raw data provides reliable data support for the subsequent power quality monitoring module to analyze indicators such as voltage fluctuations and harmonic distortion rates, and for the electricity consumption behavior feature extraction module to calculate various electricity consumption characteristics. It is an important prerequisite for the entire system to achieve abnormal electricity consumption behavior identification.

[0007] Preferably, the power quality monitoring module includes: (1) In-depth analysis of electrical parameters and calculation of key indicators: As the core component of the smart meter system for measuring power quality, the power quality monitoring module receives basic electrical parameters such as voltage and current from the data acquisition module and embarks on a journey of in-depth analysis. During the analysis process, based on the reactive power calculation formula, the reactive power is accurately calculated through specific operations on parameters such as voltage and current, thereby assessing the exchange of reactive power during power transmission. At the same time, advanced digital signal processing algorithms such as Fast Fourier Transform (FFT) are used to analyze the frequency domain characteristics of electrical parameters, and then based on the current harmonic distortion rate formula, the current harmonic distortion rate (THD) is obtained through complex calculations, and indicators such as voltage fluctuation and current fluctuation are calculated to obtain key data on power quality in all aspects; Voltage Harmonic Distortion Rate (THDV): ; Where: THDV is the voltage harmonic distortion rate (unit: %); This is the fundamental effective value of the voltage (unit: V). For the first Effective value of subharmonic voltage (unit: V); The highest harmonic order (usually taken as 40). Harmonic distortion rate (THDI) of current.

[0008] ; THD = 100 * (I2h - I1h) / I1h I1h = I1h / N I2h = I2h / N (2) Abnormal early warning, data labeling and output: after completing the calculation of various indicators, the power quality monitoring module monitors the voltage fluctuation, current fluctuation and harmonic distortion rate according to the scientifically set threshold. When THD suddenly increases, voltage abnormally fluctuates, etc. are detected, the potential abnormal power consumption early warning mechanism is triggered immediately to warn of possible power consumption problems. The module will also assign appropriate power quality labels to each data sample, such as "normal", "voltage fluctuation abnormal", "harmonic distortion abnormal", etc. Finally, the original data sequence labeled by power quality is output to the data preprocessing module, providing a clear and accurate data basis for subsequent data processing and abnormal power consumption behavior identification.

[0009] Preferably, the data preprocessing module comprises: (1) Noise elimination processing of the data preprocessing module: the data preprocessing module is the key link between power quality monitoring and subsequent data analysis, and receives the original data sequence output by the power quality monitoring module. The original data is easily disturbed during collection and transmission, so the primary task of this module is to remove noise and pulse interference. The moving average filtering and median filtering algorithms are used. The moving average filtering algorithm smoothes the data curve by performing moving average filtering operation on multiple consecutive data points, effectively suppressing random noise; the median filtering algorithm selects the middle value in the data sequence to replace the current data point, accurately removing pulse interference and purifying the data basis for subsequent processing; Moving average filtering formula: ; In the formula: Data after moving average filtering; Collected original data; Sliding window size; (2) Standardization and quality improvement of the data preprocessing module: after denoising, the data preprocessing module processes the different dimension differences of different electrical parameter data using Z-score standardization and minimum-maximum normalization technology. Z-score standardization adjusts data according to the overall distribution characteristics of the data, and minimum-maximum normalization maps the data to a unified interval. After this series of operations, the dimension influence between data is eliminated, and the data has high comparability. The final output of the denoised and standardized high-quality data sequence provides high-quality and reliable data support for the power consumption behavior feature extraction module to accurately calculate various power consumption features.

[0010] Preferably, the electricity consumption behavior feature extraction module includes: (1) Accurate Calculation of Key Electricity Consumption Characteristics: The electricity consumption behavior characteristic extraction module, based on the high-quality data output by the data preprocessing module, undertakes the important task of mining key electricity consumption characteristics. This module performs in-depth analysis of the data through the average power calculation formula, calculating key indicators such as average power, load factor, peak-to-valley difference, volatility, and harmonic distortion rate. When calculating the average power, based on relevant principles, the active power over a period of time is integrated and calculated (similar to accumulating the active power over a specific period of time and then distributing it evenly over that period of time), thereby obtaining an average value that reflects the user's electricity consumption intensity. At the same time, the load factor is obtained by comparing the actual power with the rated power. These accurately calculated characteristics provide an important basis for a comprehensive understanding of the user's electricity consumption behavior. Average power calculation formula: ; In the formula: Average power Number of sampling points; For the first Power values ​​at each sampling point; (2) Multidimensional Feature Fusion and Feature Vector Output: After completing the calculation of key electricity consumption characteristics, the electricity consumption behavior feature extraction module faces the challenge of multidimensional data processing. In order to analyze and utilize these data more efficiently, this module adopts dimensionality reduction algorithms such as principal component analysis (PCA) to fuse features of multiple dimensions such as average power and load rate. Through the algorithm's screening and integration of data features, the complex multidimensional data is transformed into a few representative comprehensive features, thereby forming the electricity consumption behavior feature vector for the current period. This feature vector, as the output of the module, provides core data support in a concise and effective data form for the subsequent short-term behavior monitoring module to perform abnormal fluctuation detection and the anomaly detection and diagnosis module to carry out accurate diagnosis.

[0011] Preferably, the short-term behavior monitoring module includes: (1) Real-time monitoring under the sliding window mechanism: The short-term behavior monitoring module, as the front-line outpost for capturing abnormal fluctuations in electricity consumption, relies on the sliding window mechanism to continuously monitor the electricity consumption feature vector output by the electricity consumption behavior feature extraction module. The size and step size of the sliding window can be flexibly adjusted according to the actual scenario, such as the common 10-minute window and 1-minute step size setting. During operation, based on the sliding window abnormal monitoring formula, the module analyzes the feature vector within the window in real time and continuously scans the dynamic changes in electricity consumption data. With this efficient monitoring method, it can keenly capture sudden abnormal fluctuations in electricity consumption and build a solid first line of defense for abnormal identification; Sliding window anomaly detection formula: ; In the formula: Power difference in sliding window; Window length; Starting point of current window; (2) Abnormal sample processing and result output: once the short-term behavior monitoring module detects an abnormal sample, it starts the short-period focus processing program. Referring to the judgment logic of the sliding window abnormal monitoring formula, the module increases the analysis frequency of abnormal samples, collects more dimensional data, and deeply excavates abnormal characteristics to effectively avoid the risk of abnormal data being covered by conventional processing methods. After careful analysis and processing, the module will finally generate short-term power consumption abnormal monitoring results, providing accurate abnormal clues for subsequent abnormal detection and diagnosis modules, and assisting the system to achieve rapid response and accurate determination of abnormal power consumption behavior.

[0012] Preferably, the abnormal detection and diagnosis module comprises: (1) Preliminary screening and filtering of abnormal data: as the core hub of the system for accurately determining abnormal power consumption, the abnormal detection and diagnosis module starts a rigorous detection process as soon as it receives the abnormal monitoring results output by the short-term behavior monitoring module. First, use the 3σ rule to preliminarily screen the data, and quickly identify data points that deviate from the data mean by 3 times the standard deviation as abnormal points. Through this efficient preliminary filtering mechanism, abnormal data that deviates significantly from the normal range can be quickly screened out, reducing the burden of subsequent more in-depth analysis, and also laying the foundation for accurate positioning of abnormalities; 3σ rule formula: ; Current power value; Historical mean; Historical standard deviation; (2) Multi-algorithm collaborative diagnosis and threshold dynamic adjustment: after completing the preliminary screening, the abnormal detection and diagnosis module uses multiple advanced algorithms for deep joint diagnosis. Use the Isolation Forest algorithm to build a unique Isolation Tree structure, and through isolation analysis of the data, accurately excavate a small amount of hidden anomalies hidden in normal data. Then, use the LSTM autoencoder to deeply learn the time series characteristics of power consumption data. When inputting abnormal data, it will produce a significant reconstruction error by virtue of its memory of normal time series characteristics, thereby effectively identifying abnormalities. In addition, the module dynamically adjusts the abnormal judgment threshold value according to the rules of historical data and the change trend of real-time data, realizes the organic cooperation of multiple models, and finally outputs accurate abnormal detection results and corresponding abnormal type labels, ensuring accurate diagnosis of various abnormal power consumption behaviors.

[0013] Preferably, the behavior trend analysis module comprises: (1) In-depth analysis of long-term electricity consumption trend: As a key part of the system to understand the long-term electricity consumption law, the behavior trend analysis module receives the abnormal results output by the abnormal detection and diagnosis module, and starts the in-depth mining of the long-term electricity consumption trend. With the help of advanced time series analysis tools such as ARIMA model and LSTM model, and combined with the core principle of moving average trend formula, the module systematically analyzes the massive historical electricity consumption data. Through the moving average trend formula, the data is dynamically analyzed, the short-term fluctuation interference is effectively filtered, and the internal law and trend characteristics of the electricity consumption data changing with time are accurately captured, laying a solid analysis foundation for subsequent abnormal behavior prediction; Moving average trend formula: ; In the formula: Moving average trend value; Sliding window length; (2) Electricity risk prediction and early warning output: Based on the in-depth analysis of electricity consumption data, the behavior trend analysis module can sensitively identify the abnormal patterns of periodic electricity stealing behavior, predict the potential risks of equipment aging failure, and explore the hidden tracks of potential electricity abnormal development. For example, by analyzing the trend evolution of equipment electricity parameters in historical data, combined with the trend judgment logic similar to the moving average trend formula, the time node and probability of possible equipment aging failure are predicted. Finally, the module presents the analysis results in the form of intuitive electricity trend chart, and outputs the predictive risk warning information with foresight, providing scientific decision-making basis for power management personnel to take preventive measures and optimize power resource allocation, effectively improving the safety and stability of the power system operation.

[0014] Preferably, the alarm and visualization module comprises: (1) Hierarchical alarm and abnormal information processing: As the key window for system and user interaction, the alarm and visualization module immediately starts the hierarchical alarm mechanism after receiving the electricity trend chart and abnormal results output by the behavior trend analysis module. According to the influence degree of abnormality on the operation of the power system, it is divided into different levels such as first-level alarm (serious abnormality), second-level alarm (relatively serious abnormality), and third-level alarm (general abnormality). When processing abnormal information, the abnormal event aggregation formula is used to intelligently integrate similar abnormal events. By analyzing key elements such as abnormal type and occurrence location, the alarm information of the same type and same location is merged, effectively reducing the redundant information caused by false alarms, and making the alarm information more concise and accurate; Abnormal event aggregation formula: ; In the formula: Alarm state (1 for alarm, 0 for normal); Alarm trigger threshold; (2) Multi-form visual display and interface output: After completing the classification and aggregation processing of alarm information, the alarm and visualization module uses diversified presentation methods to perform real-time visual display of power consumption data and abnormal conditions. With intuitive charts, dynamic power consumption curves, and clear abnormal event annotations, users can quickly understand power consumption conditions and abnormal details. Charts can display power consumption trends in different time periods, power consumption curves can reflect power changes in real time, and abnormal event annotations can highlight the time and nature of abnormal occurrences. Ultimately, the module outputs include charted alarm information and real-time power consumption abnormality visualization interfaces, providing power management personnel and users with a convenient and efficient power consumption information viewing and abnormality warning platform to help take timely measures.

[0015] Preferably, the remote collaborative management module comprises: (1) Information remote synchronization and optimized transmission: The remote collaborative management module serves as a bridge between the smart meter system and the power back-end management platform, and undertakes the important task of multi-meter information remote synchronization. After receiving the information output by the alarm and visualization module, the module uploads the data to the power back-end management platform via wireless networks (NB-IoT, 4G / 5G). To improve transmission efficiency and security, the module uses efficient data compression algorithms to reduce data transmission volume, and uses encryption algorithms to ensure data integrity and security during transmission, reducing network bandwidth requirements and providing a stable data transmission foundation for collaborative analysis of multi-meter abnormal data; (2) Back-end collaborative analysis: In the power back-end management platform, the remote collaborative management module performs collaborative analysis on the abnormal data of multiple smart meters based on the principle of multi-meter collaborative abnormality statistical formula. By integrating abnormal information from different meters and following specific statistical logic, the module can mine the correlations and rules among abnormal data and achieve remote centralized management. With this collaborative analysis mechanism, the system can support low-latency rapid response and timely detection and handling of abnormal conditions. Ultimately, the module outputs remote collaborative control instructions and centralized abnormality analysis results, providing strong support for power management personnel to make accurate decisions and efficiently handle abnormal power consumption issues, optimizing overall power consumption monitoring and management efficiency; Multi-meter collaborative abnormality statistical formula: ; In the formula: Collaborative abnormality ratio; is the alarm state of the i-th meter.

[0016] The beneficial effects of the present application are as follows: ​1. This invention uses a data acquisition module to acquire electrical parameters at high frequency and a power quality monitoring module to analyze power quality indicators in depth. Both provide rich data support for anomaly detection. The short-term behavior monitoring module captures instantaneous abnormal fluctuations in real time, while the behavior trend analysis module provides in-depth analysis of long-term power consumption trends. This achieves full-time coverage of abnormal behavior from a time dimension. Whether it is a sudden anomaly caused by equipment short circuit, a continuous anomaly caused by electricity theft, or a power quality problem, it can be accurately identified by the system, greatly improving the detection accuracy and effectively reducing the false alarm and missed alarm rates, thus building a reliable defense for the safe and stable operation of the power system.

[0017] 2. Through in-depth mining and learning of users' historical electricity consumption data, the system can accurately grasp users' electricity consumption habits, time period change patterns, and load characteristics. As the electricity consumption situation dynamically evolves, the system can automatically optimize the anomaly detection threshold, abandoning the limitations brought by the traditional fixed threshold. This personalized and adaptive design allows the system to perform at its best in different regions and for different types of users, significantly improving the system's application scope and practicality.

[0018] 3. This invention, by employing a lightweight model at the algorithm level, effectively reduces computational resource consumption, enabling edge devices to maintain high real-time performance under low power consumption. It is particularly suitable for the intelligent transformation of power facilities in remote areas and aging power grids. In terms of remote collaborative management, the remote collaborative management module achieves remote synchronization of information from multiple meters via a wireless network, enabling centralized control and collaborative analysis on the power back-end management platform. This model breaks regional limitations, improves the cross-regional anomaly collaborative analysis capability of multiple meters, and allows power management personnel to respond quickly to abnormal situations. Simultaneously, it reduces system maintenance costs, optimizes overall electricity monitoring efficiency, and provides strong support for the intelligent upgrading of power systems. Attached Figure Description

[0019] Figure 1 This is a flowchart of the smart meter system with abnormal electricity consumption behavior identification function of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment of the invention provides a smart meter system with abnormal electricity consumption behavior identification function, including: Data acquisition module: Collects basic electrical parameters such as voltage and current with high-precision sensors at a sampling period of seconds, outputs real-time data stream after signal conditioning, and provides raw data support for subsequent modules, marking the beginning of system data processing. Power quality monitoring module: Receives raw electrical parameters, calculates key indicators using digital signal processing algorithms, and outputs data with quality labels after adding threshold-based early warning exceptions. It converts raw data into a sequence with anomaly labels for the data preprocessing module. Data preprocessing module: Denoises the labeled raw data sequence using filtering algorithms and eliminates dimensional effects through standardization processing. It outputs high-quality data sequences to improve data quality and lay the foundation for electric behavior feature extraction. Electric behavior feature extraction module: Based on high-quality data, it calculates various electric features and uses dimension reduction algorithms to fuse multi-dimensional features to form a feature vector output. It extracts key information reflecting electric behavior from data and passes it to the short-term behavior monitoring module. Short-term behavior monitoring module: Uses a sliding window to monitor feature vectors, captures sudden abnormal fluctuations, and outputs monitoring results after focusing on abnormal samples. It realizes rapid identification of short-term electric anomalies and provides clues for anomaly detection and diagnosis. Anomaly detection and diagnosis module: Receives short-term monitoring results, uses multiple algorithms for joint diagnosis, dynamically adjusts thresholds, and outputs anomaly detection results and type labels. It accurately judges abnormal situations and provides a basis for behavior trend analysis. Behavior trend analysis module: Based on anomaly detection results, it analyzes long-term electric trends using time series models, identifies potential risks, and outputs trend charts and warning information. It excavates electric rules and risks from a long-term perspective and passes them to the alarm and visualization module. Alarm and visualization module: Generates hierarchical alarms based on the severity of anomalies, visualizes data and anomalies in multiple forms, and outputs interfaces after alarm aggregation and deduplication. It presents analysis results intuitively, making it easy for users to understand, and provides information for remote collaborative management. Remote collaborative management module: Synchronizes visualization information to the background remotely, uses data compression and encryption technology for transmission, and realizes multi-meter collaborative analysis and management in the background. It outputs control instructions and analysis results, completing the full-process closed loop from data acquisition to remote management.

[0022] Among them, the data acquisition module samples the voltage, current, active power, reactive power, apparent power and power factor and other basic electrical parameters at a sampling period of seconds. Its internal integration of high-precision voltage transformer, current transformer and other sensors, with excellent sensitivity, can accurately capture the subtle dynamic changes of various electrical parameters. According to the active power and reactive power calculation formula, through the precise operation of voltage, current and other parameters, the actual consumption of electric energy and the exchange of reactive energy can be accurately quantified, and accurate basic data support is provided for the system.

[0023] In terms of data processing, the original electrical parameter data collected will be preliminarily processed in the module, with the help of amplification, filtering and other signal conditioning techniques, effectively eliminating some interference signals and significantly improving data quality. The high-quality data processed is output in the form of real-time data stream through the high-speed data transmission interface, laying a solid and reliable data foundation for the subsequent power quality monitoring module to analyze voltage fluctuation, harmonic distortion rate and other indicators, and the electrical behavior characteristic extraction module to calculate average power, load rate and other electrical characteristics, becoming an important prerequisite and key guarantee for the whole system to realize accurate identification of abnormal electrical behavior.

[0024] Among them, the power quality monitoring module quantifies the exchange of reactive energy in the process of electric energy transmission according to the reactive power calculation formula through accurate operation of voltage, current and other parameters, and provides basis for power system reactive power compensation decision. At the same time, with the help of advanced digital signal processing algorithms such as fast Fourier transform (FFT), the frequency domain characteristics of electrical parameters are deeply analyzed, and the current harmonic distortion rate (THD) is calculated according to the current harmonic distortion rate formula, while the voltage fluctuation, current fluctuation and other key indicators are calculated, and the core data of power quality is obtained in all directions and multiple dimensions.

[0025] After completing all the index calculations, the power quality monitoring module based on the scientifically and rigorously set thresholds, implements real-time dynamic monitoring on voltage fluctuation, current fluctuation and harmonic distortion rate and other indicators. Once THD surge, voltage abnormal fluctuation and other situations are detected, the potential electrical abnormality early warning mechanism is triggered immediately, and the system is warned in advance to prompt possible electrical hazards. In addition, the module will accurately assign the corresponding power quality labels to each data sample according to the calculation and monitoring results, such as "normal", "voltage fluctuation abnormality", "harmonic distortion abnormality" and so on. Finally, the original data sequence marked by power quality is output to the data preprocessing module, laying a clear and accurate data foundation for the subsequent data deep processing and accurate identification of abnormal electrical behavior.

[0026] Among them, the data preprocessing module receives the original data sequence output by the power quality monitoring module. Due to various disturbances in the data collection and transmission process, the primary task of the data preprocessing module is to perform noise elimination processing. By using the moving average filtering and median filtering algorithms, the moving average filtering performs specific moving average operations on multiple consecutive data points, effectively smoothing the data curve and greatly suppressing random noise interference. The median filtering replaces the current data point with the median value in the data sequence, accurately removing impulse interference and laying a solid foundation for subsequent processing.

[0027] After completing noise elimination, the data preprocessing module addresses the dimensional difference problem of different electrical parameters and uses Z-score standardization and minimum-maximum normalization techniques for in-depth processing. Z-score standardization accurately adjusts the data based on the overall distribution characteristics of the data, and minimum-maximum normalization maps the data to a unified interval. Through this series of scientific and rigorous operations, the dimensional influence between data is completely eliminated, making the data highly comparable. The final output of the high-quality data sequence after denoising and standardization provides solid and reliable data support for the power consumption behavior feature extraction module to accurately calculate average power, load rate, and other power consumption characteristics, effectively ensuring the accuracy and effectiveness of subsequent abnormal power consumption behavior identification and analysis.

[0028] Among them, the power consumption behavior feature extraction module uses operation logic such as the average power calculation formula to systematically analyze the data and accurately calculate core indicators such as average power, load rate, peak-valley difference, fluctuation rate, and harmonic distortion rate. When calculating the average power, the active power in a specific period is integrated and calculated, and the cumulative active power in that period is evenly distributed, thereby obtaining the key value reflecting the user's power consumption intensity. The load rate is obtained by comparing the actual power with the rated power. These accurately calculated features provide solid data support for comprehensively understanding user power consumption behavior patterns.

[0029] After completing the calculation of key power consumption characteristics, in the face of the complexity of processing multi-dimensional data, the power consumption behavior feature extraction module uses advanced dimension reduction algorithms such as principal component analysis (PCA) to deeply integrate multi-dimensional features such as average power and load rate. The algorithm intelligently selects and organically integrates data features, refines complex multi-dimensional data into a few highly representative comprehensive features, and then constructs the power consumption behavior feature vector of the current period. This feature vector provides core data support for subsequent short-term behavior monitoring modules to accurately capture abnormal fluctuations, abnormal detection and diagnosis modules to accurately diagnose, and plays a key role in the entire abnormal power consumption behavior identification process.

[0030] The short-time behavior monitoring module relies on a flexible and adjustable sliding window mechanism to perform high-frequency and detailed short-time continuous monitoring on the power consumption feature vectors output by the power consumption behavior feature extraction module. The module can flexibly set the sliding window size and step according to different power consumption scenarios and monitoring needs, such as a common 10-minute window length with a 1-minute sliding step, to ensure comprehensive coverage and timely response to power consumption data changes. During operation, the module performs real-time deep analysis of the feature vectors within the window based on the core logic of the sliding window anomaly monitoring formula, like installing a sensitive "monitoring radar" for power consumption data changes, continuously scanning data dynamics, and quickly capturing sudden power consumption abnormal fluctuations with an efficient and accurate monitoring mechanism, building a solid first line of defense for abnormal power consumption behavior identification.

[0031] When an abnormal sample is detected, the short-time behavior monitoring module immediately starts the targeted short-period focus processing program. Strictly in accordance with the judgment criteria of the sliding window anomaly monitoring formula, the module actively increases the analysis frequency of abnormal samples and collects more dimensional data information to deeply mine abnormal features from multiple angles. This processing method effectively avoids the risk of abnormal data being smoothed and covered by conventional processing methods, ensuring that no minor abnormalities are missed. After rigorous and detailed analysis and processing, the module accurately outputs the short-time power consumption anomaly monitoring results, providing valuable abnormal clues for the subsequent abnormal detection and diagnosis module, and effectively promoting the system to achieve rapid response and accurate judgment of abnormal power consumption behavior, playing an indispensable role in the entire abnormal power consumption behavior identification process.

[0032] The abnormal detection and diagnosis module quickly starts a rigorous and efficient detection process after receiving the abnormal monitoring results output by the short-time behavior monitoring module. First, the module uses the 3σ criterion based on the normal distribution principle to perform preliminary screening on the data. This criterion takes the data mean as the benchmark and quickly identifies data points that deviate from the mean by 3 times the standard deviation as abnormal points, like setting a "rough screening checkpoint" for data, efficiently filtering out abnormal data that deviates significantly from the normal range, not only significantly reducing the subsequent analysis pressure, but also laying a solid foundation for accurate positioning of abnormalities.

[0033] After the preliminary screening, the anomaly detection and diagnosis module uses multiple advanced algorithms to start deep joint diagnosis. The Isolation Forest algorithm is used to build a unique isolated tree structure, which is like opening a path in the data forest, accurately isolating and mining a small number of hidden anomalies hidden in normal data. The LSTM autoencoder is used to perform deep learning on the time series characteristics of power consumption data. With its memory ability for normal power consumption behavior time series characteristics, when abnormal data is input, it will produce a significant reconstruction error, thereby accurately capturing anomalies. In addition, the module intelligently and flexibly adjusts the anomaly judgment threshold according to historical data rules and real-time data changes, realizing the organic cooperation of multiple algorithm models. Finally, it outputs accurate anomaly detection results and corresponding anomaly type labels, ensuring accurate diagnosis of various abnormal power consumption behaviors and providing reliable basis for subsequent power management and decision-making.

[0034] The behavior trend analysis module immediately conducts deep analysis of long-period power consumption trends after receiving the anomaly results output by the anomaly detection and diagnosis module. This module uses advanced time series analysis tools such as ARIMA models and LSTM models, and deeply integrates the core principles of moving average trend formulas to systematically analyze massive historical power consumption data. During the analysis process, the data processing logic contained in the moving average trend formula is used to dynamically process the power consumption data, effectively filtering out the interference caused by short-term fluctuations, like putting a "trend magnifying glass" on the data, accurately capturing the inherent laws and trend characteristics of power consumption data over time, and laying a solid data and analysis foundation for subsequent anomaly behavior prediction work.

[0035] Based on deep mining and analysis of power consumption data, the behavior trend analysis module can accurately detect abnormal patterns of periodic electricity theft behavior, scientifically predict potential risks of equipment aging failure, and accurately explore the hidden track of potential electricity anomaly development by carefully studying the trend evolution of equipment power consumption parameters in historical data and combining the trend judgment logic of the moving average trend formula. For example, it can identify the time node and probability of potential equipment aging failure. Finally, the module presents the analysis results in the form of an intuitive and easy-to-understand power consumption trend chart and outputs highly forward-looking predictive risk warning information, providing reliable scientific decision-making basis for power management personnel to plan intervention measures in advance and optimize power resource allocation, and effectively improving the safety and stability of the power system operation.

[0036] Among them, the alarm and visualization module receives the power consumption trend chart and abnormal result output by the behavior trend analysis module, and quickly starts the intelligent processing process. First of all, according to the potential harm of abnormality to the operation of the power system, the module strictly divides the multi-level alarm system of first-level alarm (serious abnormality), second-level alarm (relatively serious abnormality), third-level alarm (general abnormality), etc., to ensure that the alarm information can accurately reflect the emergency degree of the abnormality. When processing abnormal information, the logic principle of the abnormal event aggregation formula is used to deeply integrate similar abnormal events. Through the comprehensive analysis of key elements such as abnormal type, occurrence location, and occurrence time, the alarm information of the same type and same location is intelligently combined, like "slimming" the alarm information, greatly reducing the redundant information generated by false alarms, making the alarm content more concise and clear, and highlighting the key points.

[0037] After completing the classification and aggregation optimization of alarm information, the alarm and visualization module uses a variety of presentation forms to visually display the power consumption data and abnormal conditions in real time. Using intuitive column charts, line charts and other charts, it clearly presents the power consumption trend changes in different time periods; through the dynamic updating of the power curve, it reflects the fluctuation of power in real time; with the help of prominent abnormal event annotations, it accurately prompts the specific time, nature and impact range of the abnormality. Finally, the module outputs a comprehensive display platform integrating charted alarm information and real-time power abnormality visualization interface, providing an efficient and convenient information window for power management personnel to make decisions and for users to understand power consumption, and effectively promoting the timely response and proper handling of abnormal power consumption.

[0038] Among them, after receiving the information output by the alarm and visualization module, the remote collaborative management module immediately uploads the data to the power background management platform through NB-IoT, 4G / 5G and other wireless networks. In this process, to improve transmission efficiency and security, the module uses efficient data compression algorithms to greatly reduce data transmission volume, and at the same time, through encryption algorithms, it provides comprehensive protection for the integrity and security of data during transmission, effectively reduces the network bandwidth demand, and lays a stable data transmission foundation for subsequent collaborative analysis of multi-meter abnormal data.

[0039] In the power background management platform, the remote collaborative management module is based on the core principle of the multi-meter collaborative abnormality statistical formula to deeply integrate and collaboratively analyze abnormal data from multiple smart meters. Through systematic analysis of different meter abnormal information, according to specific statistical logic, the internal correlation and change rule between abnormal data are accurately mined, and remote centralized management of power consumption is realized. With this efficient collaborative analysis mechanism, the system can quickly respond to abnormal situations and effectively shorten problem processing time. Finally, the module outputs accurate remote collaborative control instructions and comprehensive centralized abnormality analysis results, providing strong support for power management personnel to make scientific decisions and efficiently handle abnormal power consumption problems, and significantly improving overall power consumption monitoring and management efficiency.

[0040] It should be noted that, in this text, relational terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between or among the entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover non-exclusive inclusions, so that a process, method, article, or apparatus that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0041] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A smart meter system with abnormal electricity consumption behavior identification function, characterized in that... The system consists of a data acquisition module, a power quality monitoring module, a data preprocessing module, a power consumption behavior feature extraction module, a short-term behavior monitoring module, an anomaly detection and diagnosis module, a behavior trend analysis module, an alarm and visualization module, and a remote collaborative management module. The data acquisition module uses high-precision sensors to collect basic electrical parameters such as voltage and current, and outputs a real-time data stream after signal conditioning. The power quality monitoring module receives the raw electrical parameters, uses digital signal processing algorithms to calculate key indicators, issues early warnings based on thresholds, adds quality labels to the data, and outputs the data, transforming the raw data into a sequence with anomaly labels. The data preprocessing module uses filtering algorithms to remove noise from the labeled raw data sequence, eliminates the influence of dimensions through standardization, and outputs a high-quality data sequence. After receiving the high-quality data sequence, the electricity consumption behavior feature extraction module calculates various electricity consumption features, uses dimensionality reduction algorithms to fuse multi-dimensional features, and outputs a feature vector. The short-term behavior monitoring module uses a sliding window to monitor feature vectors, captures sudden abnormal fluctuations, and outputs short-term monitoring results after focusing on abnormal samples. After receiving the short-term monitoring results, the anomaly detection and diagnosis module uses multiple algorithms to jointly diagnose and dynamically adjusts the threshold to output anomaly detection results and type labels. The behavioral trend analysis module analyzes long-term electricity consumption trends based on anomaly detection results and uses time series models to identify potential risks, output trend charts and early warning information, and explore electricity consumption patterns and risks from a long-term perspective. The alarm and visualization module generates graded alarms based on the severity of anomalies, visualizes data and anomalies in various forms, and outputs the analysis results intuitively after alarm aggregation and deduplication. The remote collaborative management module remotely synchronizes visualized information to the backend, uses data compression and encryption technology for transmission, and enables collaborative analysis and management of multiple meters in the backend, outputting control commands and analysis results.

2. The smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The data acquisition module includes: (1) Efficient sensing and parameter calculation of data acquisition module: With a sampling period of seconds, voltage, current, active power, reactive power, apparent power and power factor are collected at high frequency to accurately capture the subtle changes of various electrical parameters. Based on the active power calculation formula and reactive power calculation formula, the actual consumption of electrical energy is accurately quantified through the calculation of voltage and current parameters. Active power calculation formula: ； In the formula: Active power; RMS voltage value This is the effective value of the current; It is the phase angle between voltage and current; Reactive power calculation formula: ； Reactive power; (2) Data processing and output transmission of the data acquisition module: By using amplification and filtering signal conditioning technology, some interference signals are effectively removed. The processed data is output in the form of real-time data stream through a high-speed data transmission interface.

3. The smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The power quality monitoring module includes: (1) In-depth analysis of electrical parameters and calculation of key indicators: Based on the basic electrical parameters of voltage and current output by the data acquisition module, the reactive power is accurately calculated by specific calculation of voltage and current parameters according to the reactive power calculation formula. The frequency domain characteristics of electrical parameters are analyzed by using the advanced digital signal processing algorithm of fast Fourier transform. Then, the current harmonic distortion rate is obtained according to the current harmonic distortion rate formula, and the voltage fluctuation and current fluctuation indicators are calculated. Voltage harmonic distortion rate: ； Where: THDV is the voltage harmonic distortion rate; This represents the effective value of the fundamental voltage frequency. For the first RMS value of subharmonic voltage; The highest harmonic order; Current harmonic distortion rate: ； In the formula: THDI is the current harmonic distortion rate; This represents the effective value of the fundamental current wave. This represents the effective value of the h-th harmonic current. (2) Abnormal warning, data labeling and output: Based on the set threshold, voltage fluctuation, current fluctuation and harmonic distortion rate indicators are closely monitored. When a sudden increase in THD or abnormal voltage fluctuation is detected, the potential power consumption abnormality warning mechanism is immediately triggered. According to the calculation results, each data sample is assigned a corresponding power quality label, and the original data sequence after power quality labeling is output to the data preprocessing module.

4. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The data preprocessing module includes: (1) Noise removal processing of the data preprocessing module: After receiving the original data sequence output by the power quality monitoring module, the moving average filtering and median filtering algorithms are used. The moving average filtering smooths the data curve by performing moving average filtering operations on multiple consecutive data points. Moving average filter formula: ； In the formula: Data after moving average filtering; The raw data collected; Slide window size; (2) Standardization and quality improvement of data preprocessing module: Z-score standardization and min-max normalization techniques are used for processing. Z-score standardization adjusts the data according to the overall distribution characteristics of the data, while min-max normalization maps the data to a unified range, eliminating the influence of dimensions between data and making the data highly comparable.

5. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The electricity consumption behavior feature extraction module includes: (1) Accurate calculation of key power consumption characteristics: Through the average power calculation formula, the data is analyzed in depth to calculate key indicators such as average power, load rate, peak-to-valley difference, volatility, and harmonic distortion rate. Average power calculation formula: ； In the formula: Average power Number of sampling points; For the first Power values ​​at each sampling point; (2) Multidimensional feature fusion and feature vector output: Principal component analysis dimensionality reduction algorithm is adopted to fuse features of multiple dimensions such as average power and load rate. Through the algorithm to screen and integrate data features, complex multidimensional data is converted into a few representative comprehensive features, thereby forming the current electricity consumption behavior feature vector.

6. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The short-term behavior monitoring module includes: (1) Real-time monitoring under the sliding window mechanism: Based on the sliding window mechanism, the power consumption feature vector output by the power consumption behavior feature extraction module is continuously monitored in a short time. During operation, based on the sliding window anomaly monitoring formula, the module analyzes the feature vector in the window in real time, which can keenly capture sudden abnormal fluctuations in power consumption. Sliding window anomaly detection formula: ； In the formula: Power difference within the sliding window; Window length; Current window start point; (2) Abnormal sample processing and result output: When the short-term behavior monitoring module detects an abnormal sample, it immediately starts the short-cycle focusing processing program. Referring to the judgment logic of the sliding window abnormal monitoring formula, the module increases the analysis frequency of the abnormal sample, collects more data in more dimensions, deeply mines abnormal features, and finally outputs the short-term power consumption abnormal monitoring result.

7. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The anomaly detection and diagnosis module includes: (1) Preliminary screening and filtering of abnormal data: After receiving the abnormal monitoring results, the 3σ criterion is used to conduct preliminary screening of the data, and data points that deviate from the data mean by more than 3 times the standard deviation are quickly identified as abnormal points; 3σ criterion formula: ； Current power value; Historical average; Historical standard deviation; (2) Multi-algorithm collaborative diagnosis and dynamic threshold adjustment: After the initial screening is completed, a unique isolated tree structure is constructed using the isolated forest algorithm. The data is isolated and analyzed. Then, the LSTM autoencoder is used to learn the time-series characteristics of the electricity consumption data in depth, effectively identifying anomalies. Based on the patterns of historical data and the changing trends of real-time data, the anomaly judgment threshold is dynamically adjusted.

8. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The behavioral trend analysis module includes: (1) In-depth analysis of long-term electricity consumption trends: After receiving the abnormal results output by the anomaly detection and diagnosis module, long-term electricity consumption trend analysis is performed based on the ARIMA model and LSTM model. Combining the core principle of the moving average trend formula, historical electricity consumption data is systematically sorted out to effectively filter out short-term fluctuation interference. Moving average trend formula: ； In the formula: Moving average trend value; Sliding window length; (2) Electricity risk prediction and early warning output: Based on in-depth analysis of electricity data, it can keenly identify abnormal patterns of periodic electricity theft, predict potential risks of equipment aging and failure, and explore the hidden trajectory of potential abnormal electricity development.

9. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The alarm and visualization module includes: (1) Graded alarm and abnormal information processing: After receiving the power consumption trend map and abnormal results output by the behavior trend analysis module, different alarm levels are divided according to the degree of impact of the abnormality on the operation of the power system. When processing abnormal information, the abnormal event aggregation formula is used to intelligently integrate similar abnormal events. Abnormal event aggregation formula: ； In the formula: Alarm status; Alarm trigger threshold; (2) Multi-form visualization display and interface output: With the help of intuitive charts, dynamic power consumption curves and clear abnormal event labels, the final output includes charted alarm information and real-time power consumption abnormality visualization interface.

10. A smart meter system with abnormal electricity consumption behavior identification function according to claim 1, characterized in that: The remote collaborative management module includes: (1) Remote synchronization and optimized transmission of information: After receiving the information output by the alarm and visualization module, the data is uploaded to the power back-end management platform through the wireless network. The efficient data compression algorithm is used to reduce the amount of data transmission, and the encryption algorithm is used at the same time. (2) Back-end collaborative analysis: Based on the principle of multi-meter collaborative anomaly statistical formula, the abnormal data of smart meters are analyzed collaboratively. By integrating the abnormal information of different meters, the correlation and pattern between abnormal data are explored according to specific statistical logic. Multi-meter coordination anomaly statistical formula: ； In the formula: The proportion of abnormal collaborations; For the first Alarm status of each electricity meter.

Images