Electric energy meter remote monitoring management system based on edge calculation

Abstract

The invention belongs to the technical field of electric energy meter monitoring and management, and discloses an electric energy meter remote monitoring and management system based on edge computing. The system is composed of a multi-protocol electric energy data acquisition module, an edge data preprocessing module, an edge intelligent anomaly detection module, an edge priority alarm module, an edge dynamic routing and load balancing module, a cloud depth prediction analysis module, a cloud-edge collaborative strategy issuing module and a remote user interaction and strategy configuration module. And an edge adaptive learning optimization module. A sliding window adaptive threshold and edge lightweight SVM linkage algorithm is innovatively adopted through an edge intelligent anomaly detection module, the sliding window adaptive threshold dynamically calculates data features and adaptively adjusts threshold preliminary screening anomaly, and the edge lightweight SVM linkage algorithm accurately recognizes electric leakage, overload and other abnormal states through historical data training. And the edge priority alarm module automatically classifies anomalies according to a preset level, and performs top-speed transmission on high-priority alarms through a low-delay link in combination with a link state.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric energy meter monitoring and management, and specifically relates to an electric energy meter remote monitoring and management system based on edge computing. Background Art

[0002] With the rapid development of smart grids, the scale of power grids continues to expand, and distributed energy, smart electricity terminals and IoT devices are connected in large numbers. As the basic collection unit of electricity consumption data, the number of electricity meters is showing an exponential growth trend. Faced with such a large access scale, remote monitoring and refined management of electricity meters have become an important part of the digital transformation of the power industry. At this stage, users have put forward higher requirements for the real-time, accuracy and transparency of electricity consumption data, hoping to be able to grasp the electricity consumption dynamics at any time and reasonably optimize the electricity consumption structure, so as to achieve safe, energy-saving and intelligent electricity management.

[0003] However, the existing remote monitoring and management systems for electricity meters mostly adopt a "cloud-based centralized processing" architecture, which has significant bottlenecks. First, in terms of data transmission, massive electricity meters continuously upload raw data to the cloud, resulting in a huge load on the transmission link. The data link is prone to delays, packet loss, congestion and other problems, which seriously restricts the real-time monitoring, especially in the event of sudden anomalies and cannot respond quickly.

[0004] Secondly, the existing system lacks compatibility with electricity meter protocols. It cannot efficiently adapt to heterogeneous electricity meters with multiple communication protocols such as DL / T645, Modbus, and CJ / T188, resulting in protocol barriers and access restrictions in data collection. In addition, the current system's anomaly detection, alarm processing, and policy adjustment mostly use static configuration and fixed thresholds, lacking intelligent dynamic adjustment capabilities, and are difficult to adapt to complex changes in different electricity usage scenarios, resulting in abnormal response delays and frequent misjudgments and missed judgments, seriously affecting power operation and maintenance efficiency and electricity safety. Summary of the Invention

[0005] The purpose of the present invention is to provide an electric energy meter remote monitoring and management system based on edge computing to solve the problems raised in the above background technology.

[0006] In order to achieve the above objectives, the present invention provides the following technical solutions: an electric energy meter remote monitoring and management system based on edge computing, comprising: Multi-protocol power data acquisition module: Equipped with multiple communication interfaces, it can connect to different power meters to collect various data such as electricity consumption, automatically identify protocols and perform standardized processing, support protocol self-adaptation and format conversion, and provide basic data for subsequent processing; Edge data preprocessing module: This module receives collected data, filters it, removes noise, restructures it, and compensates for packet loss. It uses an improved Kalman filter to dynamically optimize the sampling frequency, forming a standard time series to provide high-quality data for anomaly detection. Edge intelligent anomaly detection module: Based on pre-processed time series data, it uses a sliding window and lightweight SVM algorithm for joint detection. It dynamically adjusts the threshold based on historical data to quickly identify abnormal power usage and pass the results to the alarm module. Edge priority alarm module: Receives anomaly detection results, classifies them by level, monitors communication link status, uses local algorithms to sort alarm data, and selects low-latency links for transmission to ensure that important alarms are delivered in a timely manner. The data then enters the routing module. Edge dynamic routing and load balancing module: Based on link state parameters, it uses genetic algorithms to optimize communication paths, dynamically allocate traffic, and ensure reliable data transmission to the cloud, providing data support for in-depth analysis; Cloud-based deep prediction and analysis module: Receives uploaded data, uses the LSTM model to predict electricity consumption trends, classifies abnormal patterns, and forms a map of electricity consumption behavior, providing a decision-making basis for grid management. The results are used for strategy generation. Cloud-edge collaborative policy delivery module: Generates various policy instructions based on cloud analysis results and uses a differentiated distribution mechanism to deliver them to edge nodes based on priority, enabling cloud-edge collaborative control and guiding edge node operation. Remote user interaction and policy configuration module: This module provides users with a multi-terminal access interface, supports data monitoring, alarm subscription, and policy configuration, and enables personalized remote management. Its configuration information can assist in edge optimization. Edge adaptive learning optimization module: Combines user configuration and operation data, uses reinforcement learning algorithms to optimize edge node parameters and strategies, supports dynamic optimization throughout the entire life cycle, and improves system intelligence and stability.

[0007] Preferably, the multi-protocol power data acquisition module includes: (1) Multi-source data acquisition: The multi-protocol power data acquisition module achieves physical connection with various types of power meters through the locally deployed multi-protocol acquisition unit and a rich communication interface. During the acquisition process, it not only obtains power consumption data such as voltage and current in real time, but also comprehensively collects information such as the power meter's operating status, communication quality, and ambient temperature and humidity, providing a multi-dimensional and comprehensive data foundation for system operation; (2) Intelligent protocol processing: This module has powerful protocol processing capabilities and a built-in protocol feature library that can automatically identify various electricity meter communication protocols such as DL / T645 and Modbus. By sending a detection signal, the protocol type is accurately determined based on the matching results of the returned data and the feature library. Using a unified standardization formula for electric energy sampling, data in different protocol formats is formatted to achieve unified input of heterogeneous protocol electric energy data. At the same time, it supports protocol self-adaptation and automatic format conversion, which can quickly be compatible with new protocol electric energy meters, significantly improving the system's adaptability to various types of electric energy meters.

[0008] Unified standardized formula for electric energy sampling: ; Where: is the standardized electric energy data (dimensionless); The original sampling data of the electric energy meter (unit: V, A, kWh); The maximum value of the protocol sampling data corresponding to the electric energy meter; It is the minimum value of the protocol sampling data corresponding to the electric energy meter.

[0009] Preferably, the edge data preprocessing module includes: (1) Data purification and regularization: The edge data preprocessing module uses the output data of the acquisition module as the basis, uses the digital filtering algorithm to filter out noise interference, and then uses the denoising algorithm to accurately remove abnormal data points to improve the purity and accuracy of the data. At the same time, the data is reorganized in time sequence, arranged in chronological order, and the packet loss compensation algorithm is used to estimate and fill in the lost data based on the data change trend to ensure data integrity and consistency; (2) Dynamic sampling frequency optimization: The module uses an improved Kalman filter algorithm to optimize the data sampling frequency. Based on the Kalman filter dynamic sampling optimization formula, it achieves dynamic optimization of sampling noise. This formula adjusts the sampling frequency in real time based on the data variation characteristics and system resource usage: when the data is stable, the sampling frequency is reduced to alleviate transmission and processing pressure; when the data fluctuates violently, the sampling frequency is increased to ensure that key data is not missed. Through dynamic optimization, it is effectively different from the static sampling strategy, forming a standard time series, and providing high-quality data support for subsequent modules.

[0010] Kalman filter dynamic sampling optimization formula: ; Where: The best estimate at the current moment; The estimated value at the previous moment; Kalman gain at the current moment; Actual observation value at the current moment (sampled data); Observation matrix.

[0011] Preferably, the edge intelligent anomaly detection module includes: (1) Dual-algorithm collaborative anomaly detection: The edge intelligent anomaly detection module uses a sliding window adaptive threshold algorithm and an edge lightweight SVM algorithm based on pre-processed time series data. The former dynamically calculates data mean, variance and other features according to the set window to preliminarily screen anomalies; the latter trains a classification model based on historical data to accurately identify abnormal power consumption conditions such as leakage and overload. The two work together to achieve efficient anomaly detection; (2) Dynamic Threshold Optimization Mechanism: This module uses a sliding window anomaly detection threshold adjustment formula to dynamically optimize the anomaly determination threshold based on historical and current data trends. This formula can adapt to on-site power fluctuations in real time, avoiding misjudgments or missed detections due to environmental changes. Through dynamic threshold adjustment and dual algorithm linkage, an edge self-learning anomaly detection system is formed, which efficiently completes detection tasks locally, reduces cloud computing load, and improves the accuracy and timeliness of anomaly detection.

[0012] Sliding window anomaly detection threshold adjustment formula: ; Where: Adaptively adjusted anomaly detection threshold; Current anomaly detection threshold; Threshold adjustment coefficient (0< <1); The number of abnormal points in the current window; The total number of sampling points in the current window.

[0013] Preferably, the edge priority alarm module includes: (1) Automatic classification of abnormal alarms: The edge priority alarm module receives the results of the edge intelligent abnormal detection module and automatically classifies abnormal power usage according to the preset three-level standards of emergency, minor, and prompt. Among them, situations such as leakage that may cause serious consequences are classified as emergency, minor overload is classified as minor, and communication quality degradation is classified as prompt, laying the foundation for subsequent alarm processing; (2) Priority transmission strategy optimization: During the data transmission phase, the module monitors the status parameters of 4G, NB-IoT, LoRa, and other communication links in real time, such as delay and bandwidth. Using the alarm priority assignment formula, it quantitatively associates the abnormality level with the link parameters, and assigns accurate priority values ​​to the alarm data. Based on this, priority sorting is performed, and for high-priority alarm data, the lowest latency link is preferred for transmission. This formula provides quantitative support for priority sorting, enabling local alarm priority sorting and dynamic link selection at the edge node, improving the timeliness of alarm transmission and reducing cloud processing pressure.

[0014] Alarm priority assignment formula: ; Where: No. The priority score of the alarm; Alarm weight coefficient; Alarm severity level (grading value) Frequency of alarms (times / hour); Alarm duration (unit: seconds).

[0015] Preferably, the edge dynamic routing and load balancing module includes: (1) Multi-link data monitoring and processing: The edge dynamic routing and load balancing module receives the alarms and normal data from the edge priority alarm module, and monitors the core parameters such as delay, packet loss rate, signal-to-noise ratio, etc. of multiple communication links such as 4G, NB-IoT, and LoRa in real time. By continuously collecting and analyzing link status information, it provides accurate data support for subsequent communication path optimization and data traffic allocation, ensuring dynamic control of link status; (2) Genetic algorithm-driven path optimization: Based on the link parameters obtained from monitoring, the module uses genetic algorithms to simulate the biological evolution mechanism and perform selection, crossover, and mutation operations on potential communication paths. Combined with the link optimization path cost function, factors such as link delay and bandwidth utilization are quantified as path costs, and the path selection scheme is continuously optimized iteratively based on this. During data transmission, traffic is dynamically allocated according to the load conditions of each link to achieve load balancing, driving edge nodes to dynamically select the optimal communication link, breaking through the limitations of traditional fixed links, ensuring low-latency and high-reliability data transmission, and improving the system network adaptability.

[0016] Link optimization path cost function: ; Where: No. The path cost of the links; Link real-time delay (unit: ms); Link packet loss rate (%); Link signal-to-noise ratio (unit: dB); Path weight parameter.

[0017] Preferably, the cloud-based deep prediction and analysis module includes: (1) Data in-depth analysis and trend prediction: The cloud-based in-depth prediction and analysis module receives data transmitted from the edge and performs in-depth processing using the LSTM model. With the help of the LSTM power consumption trend prediction formula, time series analysis is performed on historical power consumption data to explore the time dependency of the data and accurately predict the future load change trend of the electricity meter. At the same time, abnormal data is classified and sorted, and different abnormal patterns are identified to provide basic data support for subsequent power system operation and maintenance; LSTM electricity consumption trend prediction formula: ; Where: The current hidden layer state; Activation function (such as tanh or ReLU) Input weight matrix; Current moment input (sampled power data); State transition weight matrix; Bias term; (2) Graph Construction and Collaborative Learning Optimization: This module combines massive historical data with user electricity usage behavior to construct an intuitive electricity usage behavior graph, clearly presenting electricity usage habits and peak and valley periods, facilitating large-scale grid load forecasting and regional fault warning. Its innovative cloud-based multi-node collaborative learning mechanism integrates the computing and storage resources of each node, deepens predictive analysis based on the electricity usage behavior graph, significantly improves forecast accuracy and reliability, and lays a solid data foundation for power system planning and scheduling.

[0018] Preferably, the cloud-edge collaborative strategy delivery module includes: (1) Intelligent strategy generation: The cloud-edge collaborative strategy delivery module uses the prediction results of the cloud-based deep prediction and analysis module as a basis to intelligently generate parameter adjustment instructions, abnormal response strategies, and communication link optimization strategies. Through in-depth analysis of data such as power consumption trends and abnormal patterns, various strategies that meet actual needs are formulated to provide strategic support for the optimized operation of the system; (2) Differentiated instruction distribution: During the instruction distribution phase, the module adopts a differentiated edge instruction distribution mechanism, combined with a cloud-edge policy priority ranking formula, to dynamically prioritize instructions. This formula comprehensively considers factors such as edge node functions, performance, working status, and instruction importance and urgency, and assigns precise priorities to cloud multi-node strategies. Important and urgent instructions are issued first, and general instructions are flexibly arranged according to node load, realizing intelligent optimization of multi-node concurrent scheduling, efficiently achieving cloud-edge real-time collaborative control, and significantly improving system operation efficiency and responsiveness.

[0019] Cloud-edge policy delivery priority formula: ; Where: No. Priority scoring for issuing policies; Current edge node utilization (%); The current number of alarms for the node; Average alarm priority level; Priority weight parameter.

[0020] Preferably, the remote user interaction and policy configuration module includes: (1) Diversified remote interactive management: The remote user interaction and policy configuration module creates multiple access channels for users, such as mobile terminal applications and web platforms. Through these interfaces, users can monitor electricity consumption data in real time and comprehensively view operating parameters such as voltage and current of electricity meters; subscribe to alarm messages to timely grasp abnormal power consumption dynamics; and conveniently perform management operations such as edge node policy adjustment, parameter configuration and firmware upgrade, thus building an efficient remote management and control system; (2) Personalized strategy dynamic optimization: This module uses the user strategy influencing factor adjustment formula to achieve personalized dynamic adjustment of remote strategies. This formula comprehensively considers multiple influencing factors such as node type, alarm level, and power usage scenario. Users can flexibly adjust management strategies for specific edge nodes or alarm types. By quantifying the weights of each factor, it accurately adapts to different user needs, greatly improving management flexibility, ensuring the efficient operation of the system in complex scenarios, and meeting the diverse remote management needs of electricity meters.

[0021] User strategy impact factor adjustment formula: ; Where: No. Individual user strategy adjustment weights; User-defined alarm category priority; User-set alarm response time requirement (ms); The number of functions enabled set by the user.

[0022] Preferably, the edge adaptive learning optimization module includes: (1) Reinforcement learning-driven optimization mechanism: The edge adaptive learning optimization module integrates user configuration, historical operation data, and cloud-edge collaborative information, relies on the edge reinforcement learning parameter update formula (based on Q-Learning), and uses the reinforcement learning algorithm as the core driving force to achieve automatic optimization of edge node parameters and strategies. This formula continuously adjusts the optimal solution of each parameter through quantitative environmental feedback and reward mechanism, provides precise mathematical guidance for the self-optimization of edge nodes, and builds a basic framework for dynamic optimization; Edge reinforcement learning parameter update formula (based on Q-Learning): ; In the formula state Take action Estimated current value of Learning rate Immediate rewards after the current execution; Future reward discount factor (0 <y≤1); Next state; Next optional action; (2) Dynamic optimization of all-dimensional parameters: Based on the above formula, the module performs in-depth optimization of edge nodes throughout their entire life cycle. In terms of anomaly detection, the detection threshold is dynamically adjusted to improve the accuracy of anomaly identification; based on data transmission requirements and network conditions, the communication strategy is optimized to enhance data transmission efficiency; based on the characteristics of data changes and resource usage, the sampling period is reasonably adjusted to balance data quality and resource consumption. Through the continuous self-optimization of multi-dimensional parameters, the system is able to continuously adapt to complex power consumption environments, achieving a dual improvement in intelligence level and long-term stability.

[0023] The beneficial effects of the present invention are as follows: 1. The present invention uses a multi-protocol power data acquisition module with automatic protocol identification and format conversion functions to achieve seamless compatibility with various protocol power meters such as DL / T645 and Modbus, breaking down data access barriers. The edge data preprocessing module uses an improved Kalman filter algorithm to dynamically optimize the sampling frequency, combined with digital filtering, packet loss compensation and other technologies to ensure data purity, integrity and temporal consistency. The edge dynamic routing and load balancing module introduces a genetic algorithm to monitor parameters such as link delay and packet loss rate in real time, and dynamically adjusts communication paths and balances traffic through simulated evolutionary operations, significantly improving data transmission efficiency.

[0024] 2. The present invention innovatively adopts the sliding window adaptive threshold and edge lightweight SVM linkage algorithm through the edge intelligent anomaly detection module. The former dynamically calculates data features and adaptively adjusts the threshold to initially screen anomalies; the latter accurately identifies abnormal conditions such as leakage and overload through historical data training; the edge priority alarm module automatically classifies anomalies according to preset levels, and transmits high-priority alarms at high speed through low-latency links based on the link status; the cloud-edge collaborative strategy delivery module synchronously generates a response strategy to achieve a rapid closed loop of the entire process from detection to processing, greatly shortening the fault response time, and greatly reducing the safety hazards of electricity use and the risk of economic losses.

[0025] 3. The present invention provides multi-terminal access channels such as mobile terminals and web pages through remote user interaction and policy configuration modules. Users can not only monitor electricity consumption data and subscribe to alarms in real time, but also adjust the formula based on user policy influencing factors, flexibly customize management strategies according to nodes and alarm levels, and realize personalized remote management and control; the edge adaptive learning optimization module is based on reinforcement learning algorithm, integrating user configuration, historical data and cloud-edge collaborative information, and continuously optimizes anomaly detection, communication, sampling and other parameters through edge reinforcement learning parameter update formula; the system self-evolves throughout its life cycle, flexibly adapts to different scenarios, significantly improves the level of management intelligence and efficiency, and effectively reduces operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of the electric energy meter remote monitoring and management system based on edge computing in the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] like Figure 1 As shown, an embodiment of the present invention provides an electric energy meter remote monitoring and management system based on edge computing, including: Multi-protocol power data acquisition module: Equipped with multiple communication interfaces, it can connect to different power meters to collect various data such as electricity consumption, automatically identify protocols and perform standardized processing, support protocol self-adaptation and format conversion, and provide basic data for subsequent processing; Edge data preprocessing module: This module receives collected data, filters it, removes noise, restructures it, and compensates for packet loss. It uses an improved Kalman filter to dynamically optimize the sampling frequency, forming a standard time series to provide high-quality data for anomaly detection. Edge intelligent anomaly detection module: Based on pre-processed time series data, it uses a sliding window and lightweight SVM algorithm for joint detection. It dynamically adjusts the threshold based on historical data to quickly identify abnormal power usage and pass the results to the alarm module. Edge priority alarm module: Receives anomaly detection results, classifies them by level, monitors communication link status, uses local algorithms to sort alarm data, and selects low-latency links for transmission to ensure that important alarms are delivered in a timely manner. The data then enters the routing module. Edge dynamic routing and load balancing module: Based on link state parameters, it uses genetic algorithms to optimize communication paths, dynamically allocate traffic, and ensure reliable data transmission to the cloud, providing data support for in-depth analysis; Cloud-based deep prediction and analysis module: Receives uploaded data, uses the LSTM model to predict electricity consumption trends, classifies abnormal patterns, and forms a map of electricity consumption behavior, providing a decision-making basis for grid management. The results are used for strategy generation. Cloud-edge collaborative policy delivery module: Generates various policy instructions based on cloud analysis results and uses a differentiated distribution mechanism to deliver them to edge nodes based on priority, enabling cloud-edge collaborative control and guiding edge node operation. Remote user interaction and policy configuration module: This module provides users with a multi-terminal access interface, supports data monitoring, alarm subscription, and policy configuration, and enables personalized remote management. Its configuration information can assist in edge optimization. Edge adaptive learning optimization module: Combines user configuration and operation data, uses reinforcement learning algorithms to optimize edge node parameters and strategies, supports dynamic optimization throughout the entire life cycle, and improves system intelligence and stability.

[0029] Among them, the multi-protocol electricity data acquisition module uses a locally deployed multi-protocol acquisition unit and a variety of communication interfaces to physically connect with various types of electricity meters, and collects electricity consumption data such as voltage and current in real time, as well as multi-dimensional information such as the electricity meter's operating status, communication quality parameters, and ambient temperature and humidity, providing comprehensive data support for the system.

[0030] The module features a built-in protocol signature library that automatically identifies various protocols, including DL / T645 and Modbus. It accurately determines the protocol type by sending a detection signal and matching it with the signature library. It then formats data from different protocols using a standardized formula for power sampling, enabling unified input of heterogeneous power data. It also supports protocol adaptation and automatic conversion of acquisition formats, enabling rapid compatibility with new protocol power meters and significantly improving the system's applicability to various power meters.

[0031] The edge data preprocessing module uses the data output from the acquisition module as a basis. It first filters out noise using a digital filtering algorithm, and then uses a denoising algorithm to remove abnormal data points, improving data purity and accuracy. It also reorganizes the data into time series, arranging them in chronological order. Using a packet loss compensation algorithm, it estimates and fills in missing data based on data change trends to ensure integrity and consistency.

[0032] An improved Kalman filter algorithm optimizes the sampling frequency. Relying on the Kalman filter's dynamic sampling optimization formula, it makes real-time adjustments based on data fluctuations and system resource usage: When data is stable, the sampling frequency is reduced to alleviate transmission and processing pressure, while when data fluctuates sharply, the sampling frequency is increased to ensure that critical data is not missed. This dynamic optimization, unlike static sampling strategies, forms a standard time series, providing high-quality data support for subsequent modules.

[0033] The edge intelligent anomaly detection module uses a sliding window adaptive threshold algorithm and a lightweight edge support vector machine (SVM) algorithm to detect anomalies based on preprocessed time series data. The former dynamically calculates statistical features such as the mean and variance of the data within a set window to perform preliminary screening for anomalies. The latter trains a classification model based on historical normal and abnormal data to accurately identify abnormal power conditions such as leakage, overload, and instantaneous power outages. The two algorithms work together to achieve efficient detection.

[0034] A sliding window anomaly detection threshold adjustment formula dynamically optimizes anomaly thresholds based on historical and current data trends. This formula adapts to fluctuations in on-site power consumption in real time, avoiding misjudgments or missed detections caused by environmental changes. This dynamic threshold adjustment, coupled with dual algorithms, creates an edge-based self-learning anomaly detection system that efficiently completes detection tasks locally, reduces cloud computing load, and effectively improves the accuracy and timeliness of anomaly detection.

[0035] The Edge Priority Alarm Module, based on the output of the Edge Intelligent Anomaly Detection Module, automatically categorizes abnormal power usage according to a pre-set three-level criteria: emergency, minor, and warning. Leakages, which could lead to serious consequences like fire, are classified as emergency, minor overloads as minor, and degraded communication quality as warning, laying the foundation for subsequent alarm processing.

[0036] During the data transmission phase, the module monitors the latency, bandwidth, signal strength, and other status parameters of communication links such as 4G, NB-IoT, and LoRa in real time. Using an alarm priority assignment formula, it quantitatively correlates anomaly levels with link parameters, assigning precise priority values ​​to alarm data. This is used as a basis for prioritization, prioritizing high-priority alarm data over the link with the lowest latency. This formula provides quantitative support for prioritization, enabling local alarm prioritization and dynamic link selection at the edge node, effectively improving the timeliness of alarm transmission while reducing cloud processing pressure.

[0037] Among them, the edge dynamic routing and load balancing module receives the alarms and normal data of the edge priority alarm module, and monitors the core parameters such as delay, packet loss rate, signal-to-noise ratio of multiple communication links such as 4G, NB-IoT, and LoRa in real time. By continuously collecting and analyzing link status information, it provides accurate data support for subsequent communication path optimization and data traffic allocation, and realizes dynamic grasp of link status.

[0038] Based on monitored parameters, the module uses a genetic algorithm to simulate biological evolution, selecting, crossing, and mutating potential communication paths. Combined with a link-optimized path cost function, it quantifies factors like link latency and bandwidth utilization as path costs, continuously iterating and optimizing path selection. During data transmission, traffic is dynamically allocated based on the load of each link to achieve load balancing, driving edge nodes to dynamically select the optimal communication link. This overcomes the limitations of traditional fixed links, ensures low-latency, highly reliable data transmission, and significantly improves the system's adaptability in complex network environments.

[0039] The cloud-based deep prediction and analysis module receives power consumption and alarm data transmitted from the edge and performs deep processing using an LSTM model. Leveraging the LSTM power consumption trend prediction formula, it conducts time series analysis on historical power consumption data, exploring its temporal dependencies and accurately predicting future load trends at the meter. It also categorizes and organizes abnormal data, identifying different abnormal patterns such as leakage and overload, providing fundamental data support for power system operations and maintenance.

[0040] Combining massive amounts of historical data with user electricity usage behavior, this system constructs intuitive electricity usage maps, clearly illustrating usage habits and peak and off-peak periods, facilitating large-scale grid load forecasting and regional fault warnings. It also innovatively employs a cloud-based multi-node collaborative learning mechanism, integrating computing and storage resources across nodes. This system leverages electricity usage behavior maps to deepen predictive analysis, significantly improving forecast accuracy and reliability, and providing a solid data foundation for power system planning and dispatch.

[0041] Among them, the cloud-edge collaborative strategy delivery module is based on the prediction results of the cloud-side deep prediction and analysis module. Through in-depth analysis of data such as power consumption trends and abnormal patterns, it intelligently generates parameter adjustment instructions, abnormal response strategies, and communication link optimization strategies, etc., to provide strategic support for system optimization operation.

[0042] During the command distribution phase, the module utilizes a differentiated edge command distribution mechanism, combined with a cloud-edge policy prioritization formula. This dynamically prioritizes commands based on factors such as edge node functionality, performance, and operating status, as well as command importance and urgency. This formula accurately prioritizes multi-node cloud policies, ensuring that critical and urgent commands are prioritized and general commands are flexibly scheduled based on node load. This intelligently optimizes multi-node concurrent scheduling, effectively achieving real-time cloud-edge collaborative control and significantly improving system operational efficiency and responsiveness.

[0043] The remote user interaction and policy configuration module provides users with multiple access channels, including mobile terminal applications and web platforms, to build an efficient remote management and control system. Through these interfaces, users can monitor electricity usage data in real time, comprehensively view operating parameters such as meter voltage and current, subscribe to alarm messages to promptly understand abnormal power usage trends, and conveniently perform management operations such as edge node policy adjustments, parameter configuration, and firmware upgrades.

[0044] The user policy influencing factor adjustment formula comprehensively considers multiple influencing factors such as node type, alarm level, and power usage scenario, enabling personalized dynamic adjustment of remote policies. Users can flexibly adjust management policies for specific edge nodes or alarm types. By quantifying the weights of each factor, precise adaptation to different needs is achieved, greatly improving management flexibility, ensuring efficient system operation in complex scenarios, and meeting diverse remote meter management needs.

[0045] The edge adaptive learning optimization module integrates user configuration, historical operational data, and cloud-edge collaborative information. It relies on an edge reinforcement learning parameter update formula (based on Q-Learning) and uses reinforcement learning algorithms as its core driving force to automatically optimize edge node parameters and policies. This formula continuously adjusts the optimal solution for each parameter through quantitative environmental feedback and reward mechanisms, providing precise mathematical guidance for edge node self-optimization and building a basic framework for dynamic optimization.

[0046] Based on this formula, the module deeply optimizes edge nodes throughout their lifecycle: dynamically adjusting anomaly detection thresholds to improve accuracy, optimizing communication strategies based on network conditions to enhance transmission efficiency, and adjusting sampling periods based on data characteristics and resource usage to balance quality and consumption. Through continuous self-optimization of multi-dimensional parameters, the system continuously adapts to complex power consumption environments, achieving both enhanced intelligence and long-term operational stability.

[0047] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0048] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A remote monitoring and management system for electric energy meters based on edge computing, characterized in that: The system consists of a multi-protocol power data acquisition module, an edge data preprocessing module, an edge intelligent anomaly detection module, an edge priority alarm module, an edge dynamic routing and load balancing module, a cloud-based deep prediction and analysis module, a cloud-edge collaborative strategy delivery module, a remote user interaction and strategy configuration module, and an edge adaptive learning optimization module. The multi-protocol power data acquisition module is equipped with multiple communication interfaces, connects to different power meters, collects power consumption data, automatically identifies protocols and standardizes processing, and supports protocol self-adaptation and format conversion. After receiving the collected data, the edge data preprocessing module filters, denoises, restructures the time series, and compensates for packet loss. It uses an improved Kalman filter to dynamically optimize the sampling frequency to form a standard time series. The edge intelligent anomaly detection module uses a sliding window and lightweight SVM algorithm for joint detection based on preprocessed time series data, dynamically adjusts the threshold based on historical data, and quickly identifies abnormal power consumption status. The edge priority alarm module receives anomaly detection results, classifies them by level, monitors the communication link status, uses a local algorithm to sort the alarm data, and selects a low-latency link for transmission. The edge dynamic routing and load balancing module uses a genetic algorithm to optimize the communication path based on link status parameters, dynamically allocates traffic, and ensures reliable data transmission to the cloud. After receiving the uploaded data, the cloud-based deep prediction and analysis module uses the LSTM model to predict electricity consumption trends, classify abnormal patterns, and form an electricity consumption behavior map to provide a decision-making basis for power grid management; the cloud-edge collaborative strategy delivery module generates various policy instructions based on the cloud-based analysis results, and uses a differentiated distribution mechanism to send them to edge nodes according to priority to achieve cloud-edge collaborative control; the remote user interaction and policy configuration module provides users with a multi-terminal access interface, supports data monitoring, alarm subscription and policy configuration functions; the edge adaptive learning optimization module combines user configuration and operation data, uses reinforcement learning algorithms to optimize edge node parameters and strategies, and supports dynamic optimization throughout the entire life cycle.

2. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The multi-protocol electric energy data acquisition module includes: (1) Multi-source data acquisition: Through the locally deployed multi-protocol acquisition unit, a physical connection with the energy meter is achieved through the communication interface. During the acquisition process, not only the voltage and current consumption data are obtained in real time, but also the operation status of the energy meter, communication quality, and ambient temperature and humidity information are collected in an all-round manner; (2) Intelligent protocol processing: Built-in protocol feature library, automatically identifies the communication protocol of the electric energy meter, sends a detection signal, accurately determines the protocol type based on the matching results of the returned data and the feature library, and uses the unified standardization formula of electric energy sampling to format data in different protocol formats; Unified standardized formula for electric energy sampling: ； Where: is the standardized electric energy data; It is the original sampling data of the electric energy meter; The maximum value of the protocol sampling data corresponding to the electric energy meter; It is the minimum value of the protocol sampling data corresponding to the electric energy meter.

3. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The edge data preprocessing module includes: (1) Data purification and regularization: Based on the output data of the multi-protocol power data acquisition module, a digital filtering algorithm is used to filter out noise interference, and then a denoising algorithm is used to accurately remove abnormal data points, reorganize the data in time series, and use a packet loss compensation algorithm to estimate and fill in lost data based on the data change trend; (2) Dynamic sampling frequency optimization: Relying on the Kalman filter dynamic sampling optimization formula, dynamic optimization of sampling noise is achieved, and the sampling frequency is adjusted in real time according to the data change characteristics and system resource usage; Kalman filter dynamic sampling optimization formula: ； Where: The best estimate at the current moment; The estimated value at the previous moment; Kalman gain at the current moment; The actual observation value at the current moment; Observation matrix.

4. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The edge intelligent anomaly detection module includes: (1) Dual-algorithm collaborative anomaly detection: Based on pre-processed time series data, a sliding window adaptive threshold algorithm and an edge lightweight SVM algorithm are used to work together. The former dynamically calculates the mean and variance characteristics of the data according to the set window to preliminarily screen anomalies; the latter trains a classification model based on historical data to accurately identify abnormal power consumption status; (2) Dynamic threshold optimization mechanism: With the help of the sliding window anomaly detection threshold adjustment formula, the anomaly judgment threshold is dynamically optimized according to the trend of historical data and current data changes. Through the dynamic adjustment of the threshold and the linkage of the dual algorithm, an edge self-learning anomaly detection system is formed; Sliding window anomaly detection threshold adjustment formula: ； Where: Adaptively adjusted anomaly detection threshold; Current anomaly detection threshold; Threshold adjustment factor; The number of abnormal points in the current window; The total number of sampling points in the current window.

5. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The edge priority alarm module includes: (1) Automatic classification of abnormal alarms: After receiving the results of the edge intelligent abnormality detection module, the abnormal power usage is automatically classified according to the preset three-level standards of emergency, minor, and prompt; (2) Priority transmission strategy optimization: Real-time monitoring of the delay and bandwidth status parameters of 4G, NB-IoT, and LoRa communication links. Using the alarm priority assignment formula, the abnormality level is quantitatively associated with the link parameters, and the alarm data is assigned an accurate priority value. Alarm priority assignment formula: ； Where: No. The priority score of the alarm; Alarm weight coefficient; Alarm severity level Frequency of alarms; Alarm duration.

6. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The edge dynamic routing and load balancing module includes: (1) Multi-link data monitoring and processing: Receive alarms and normal data from the edge priority alarm module, and monitor the core parameters of 4G, NB-IoT, and LoRa multi-communication links, such as delay, packet loss rate, and signal-to-noise ratio, in real time; (2) Genetic algorithm-driven path optimization: Based on the link parameters obtained from monitoring, the genetic algorithm is used to simulate the biological evolution mechanism to perform selection, crossover, and mutation operations on potential communication paths. Combined with the link optimization path cost function, the link delay and bandwidth utilization factors are quantified as path costs, and the path selection scheme is continuously iterated and optimized. Link optimization path cost function: ； Where: No. The path cost of the links; Link real-time delay; Link packet loss rate; Link signal-to-noise ratio; Path weight parameter.

7. The electric energy meter remote monitoring and management system based on edge computing according to claim 1 is characterized in that: The cloud-based deep prediction and analysis module includes: (1) Data in-depth analysis and trend prediction: After receiving edge transmission data, the LSTM power consumption trend prediction formula is used to perform time series analysis on historical power consumption data, explore the time dependency of the data, classify and organize abnormal data, and identify different abnormal patterns; LSTM electricity consumption trend prediction formula: ； Where: The current hidden layer state; Activation Function Input weight matrix; Current moment input; State transition weight matrix; Bias term; (2) Graph construction and collaborative learning optimization: Combine historical data with user electricity consumption behavior to construct an electricity consumption behavior graph to present electricity consumption habits and peak and valley periods. Use a cloud-based multi-node collaborative learning mechanism to integrate the computing and storage resources of each node, and deepen predictive analysis based on the electricity consumption behavior graph.

8. The electric energy meter remote monitoring and management system based on edge computing according to claim 1, characterized in that: The cloud-edge collaborative strategy delivery module includes: (1) Intelligent strategy generation: Based on the prediction results of the cloud-based deep prediction and analysis module, intelligently generate parameter adjustment instructions, abnormal response strategies, and communication link optimization strategies. Through in-depth analysis of power consumption trends and abnormal pattern data, various strategies that meet actual needs are formulated. (2) Differentiated instruction distribution: A differentiated edge instruction distribution mechanism is adopted, combined with a cloud-edge policy delivery priority formula. This formula comprehensively considers the edge node function, performance, working status, and instruction importance and urgency factors to assign precise priorities to cloud multi-node policies; Cloud-edge policy delivery priority formula: ； Where: No. Priority scoring for issuing policies; Current edge node utilization; The current number of alarms for the node; Average alarm priority level; Priority weight parameter.

9. The electric energy meter remote monitoring and management system based on edge computing according to claim 1, characterized in that: The remote user interaction and policy configuration module includes: (1) Diversified remote interactive management: Create multiple access channels for users, such as mobile terminal applications and web platforms; through these channels, users can monitor electricity consumption data in real time and comprehensively view the operating parameters of electricity meters; and conveniently perform edge node strategy adjustment, parameter configuration, and firmware upgrade management operations; (2) Dynamic optimization of personalized strategies: With the user strategy influencing factor adjustment formula, which comprehensively considers the node type, alarm level, and power usage scenario influencing factors, users can adjust the management strategy for specific edge nodes or alarm types; User strategy impact factor adjustment formula: ； Where: No. Individual user strategy adjustment weights; User-defined alarm category priority; User-set alarm response time requirements; The number of functions enabled set by the user.

10. The electric energy meter remote monitoring and management system based on edge computing according to claim 1, characterized in that: The edge adaptive learning optimization module includes: (1) Reinforcement learning-driven optimization mechanism: Integrating user configuration, historical operation data, and cloud-edge collaborative information, relying on edge reinforcement learning parameter update formulas, and continuously adjusting the optimal solution of each parameter through quantitative environmental feedback and reward mechanisms; Edge reinforcement learning parameter update formula: ； In the formula state Take action Estimated current value of Learning rate Immediate rewards after the current execution; Future reward discount factor; Next state; Next optional action; (2) Dynamic optimization of all-dimensional parameters: In terms of anomaly detection, the detection threshold is dynamically adjusted, the communication strategy is optimized based on data transmission requirements and network conditions, the sampling period is reasonably adjusted based on the data change characteristics and resource usage, and the balance between data quality and resource consumption is achieved.

Images