A method and system for detecting abnormal data in distribution network based on edge computing

By building an abnormal feature extraction model at edge nodes and applying federated learning algorithms, the problems of large data transmission volume, high delay and privacy leakage in traditional centralized detection methods are solved, and efficient processing and accurate abnormality detection of distribution network data are realized. , improves the security and stability of the system.

CN118193954BActive Publication Date: 2024-11-15江苏华网融智科技有限公司
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410422160.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-09
Publication Date
2024-11-15
Estimated Expiration
2044-04-09

AI Technical Summary

Technical Problem

The traditional centralized distribution network abnormal data detection method has the risk of large data transmission volume, high delay and privacy leakage, and insufficient large-scale data processing and analysis computing resources, making model training difficult.

Method used

Distribution network data is collected through edge devices, an abnormal feature extraction model is built at edge nodes, an abnormal feature extraction model is extracted using convolutional neural networks and recurrent neural networks, and a global abnormal data detection model is established using federated learning algorithms to realize data near-end processing and multiple data Data sharing and model training between edge nodes.

Benefits of technology

Reduce data transmission volume and network delay, improve data processing efficiency, protect data privacy, improve the overall performance of the detection model, and more comprehensively and accurately identify abnormal situations in the distribution network, improving the security of the system and stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118193954B_ABST
    Figure CN118193954B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for detecting abnormal data in a distribution network based on edge computing, including: collecting distribution network data through edge devices, and forming a sample data set after data preprocessing of the distribution network data; constructing an abnormal feature extraction model in an edge node based on a convolutional neural network and a recurrent neural network, inputting the sample data set into the abnormal feature extraction model for processing, and extracting abnormal features; establishing a global abnormal data detection model based on a federated learning algorithm, aggregating and analyzing multiple abnormal features, and outputting detection results. The present invention realizes data sharing and model training between multiple edge nodes through the application of a federated learning algorithm, while protecting data privacy, improving the overall performance of the detection model, aggregating and analyzing multiple abnormal features, and outputting global abnormal data detection results, which can more comprehensively and accurately identify abnormal situations in the distribution network and improve the safety and stability of the distribution network system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network anomaly detection, and in particular to a distribution network abnormal data detection method and system based on edge computing. Background Art

[0002] With the development of smart grids and the continuous expansion of distribution networks, the requirements for the security and stability of distribution networks are becoming increasingly higher. Abnormal data detection plays a vital role in the operation of distribution networks. It can identify potential problems in a timely manner and take corresponding measures to ensure the normal operation of the power grid. In traditional abnormal data detection methods, it mainly relies on centralized data processing and analysis, which has problems such as large data transmission volume, high latency, and high risk of privacy leakage.

[0003] In recent years, edge computing technology has been widely used and developed. Edge computing places computing resources and data storage functions on edge devices that are close to the data source, which can realize local processing and analysis of data, reduce data transmission volume, reduce network latency, and improve data processing efficiency. By combining deep learning technologies such as convolutional neural networks (CNN) and recurrent neural networks (RNN), an abnormal feature extraction model is built in the edge node, which can effectively extract abnormal features from distribution network data and provide strong support for subsequent global abnormal data detection.

[0004] However, the existing technology still has some shortcomings in the detection of abnormal data in distribution networks. The traditional centralized data processing method may lead to an increase in data transmission volume and high data processing latency, and there is also a risk of privacy leakage. In addition, for the processing and analysis of large-scale distribution network data, traditional methods may face challenges such as insufficient computing resources and difficulty in model training. The distribution network abnormal data detection method based on edge computing invented by us can effectively overcome these shortcomings by constructing an abnormal feature extraction model in the edge node and applying the federated learning algorithm to establish a global abnormal data detection model.

[0005] By collecting distribution network data through edge devices and extracting abnormal features in edge nodes, local data processing can be achieved, data transmission volume and network latency can be reduced, and data processing efficiency can be improved. The application of federated learning algorithms can realize data sharing and model training among multiple edge nodes, protect data privacy while improving the overall performance of the detection model, aggregate and analyze multiple abnormal features, and output global abnormal data detection results, which can more comprehensively and accurately identify abnormal situations in the distribution network and improve the security and stability of the distribution network system. Summary of the invention

[0006] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0007] In view of the above existing problems, the present invention is proposed.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: collecting distribution network data through edge devices, and forming a sample data set after data preprocessing;

[0009] Based on convolutional neural networks and recurrent neural networks, an abnormal feature extraction model is constructed in the edge node, and the sample data set is input into the abnormal feature extraction model for processing to extract abnormal features;

[0010] A global abnormal data detection model is established based on the federated learning algorithm, multiple abnormal features are aggregated and analyzed, and the detection results are output.

[0011] As a preferred solution of the method for detecting abnormal data in a distribution network based on edge computing described in the present invention, the edge device includes a sensor, a smart meter, a smart circuit breaker and a data collector;

[0012] The distribution network data includes power parameters, environmental parameters, equipment status data and communication data.

[0013] As a preferred solution of the abnormal data detection method for distribution network based on edge computing described in the present invention, the data preprocessing includes data cleaning, format conversion, feature extraction and dimensionality reduction processing.

[0014] As a preferred solution of the method for detecting abnormal data in a distribution network based on edge computing according to the present invention, forming the sample data set includes:

[0015]

[0016] Among them, X is the sample data set, m is the number of samples, n is the number of features, S is the data after data and processing, i is the sample index, j is the feature index, S ij is the value of the jth feature of the ith sample in the distribution network data, t is the integral variable, and its value range is (0, T), e t is an exponential function, represents the sum of all eigenvalues ​​of the i-th sample, It means that the eigenvalues ​​of each sample are summed and then multiplied by the exponential function for integral calculation.

[0017] As a preferred solution of the method for detecting abnormal data in a distribution network based on edge computing described in the present invention, the abnormal feature extraction model is constructed, including:

[0018]

[0019] Among them, F is the output abnormal feature, M is the abnormal feature extraction model, X is the input sample data set, σ is the activation function, m represents the number of samples, and x i represents the i-th sample, h i-1 represents the recurrent neural network output at the previous moment, Conv(·) represents the convolutional neural network processing, LSTM(·) represents the long short-term memory network processing, Indicates that the output of the convolutional neural network and the recurrent neural network is integrated.

[0020] As a preferred solution of the method for detecting abnormal data in a distribution network based on edge computing according to the present invention, extracting the abnormal features includes:

[0021] Inputting the sample data set through the input layer of the abnormal feature extraction model;

[0022] Extracting spatial features from the sample data set using a convolutional neural network and capturing temporal correlations and sequence patterns in the sample data set using a recurrent neural network;

[0023] The abnormal feature extraction model gradually extracts the abnormal features in the sample data set through multiple iterative learning of the abnormal features.

[0024] As a preferred solution of the method for detecting abnormal data in a distribution network based on edge computing described in the present invention, the global abnormal data detection model is established to aggregate and analyze multiple abnormal features, including:

[0025] y=f(Z)

[0026]

[0027] Where Z is the global anomaly feature, f(·) is the classification function, y is the detection result, which is a probability vector, that is, each element represents the probability of the corresponding anomaly category, which can be calculated by the softmax function, K represents the number of anomaly categories, and q represents the dimension of the feature vector.

[0028] The detection result y is judged as abnormal by setting a threshold. When the probability of the abnormal category is ≥ the threshold, it is judged as abnormal.

[0029] As a preferred solution of the distribution network abnormal data detection system based on edge computing described in the present invention, it includes:

[0030] one or more processors;

[0031] A memory stores operable instructions, which, when executed by the one or more processors, cause the one or more processors to perform operations, wherein the operations include the process of the aforementioned method for detecting abnormal data in a distribution network based on edge computing.

[0032] The beneficial effects of the present invention are as follows: the present invention collects distribution network data through edge devices and extracts abnormal features in edge nodes, which can realize proximal processing of data, reduce data transmission volume and network delay, and improve data processing efficiency; and through the application of federated learning algorithms, realizes data sharing and model training among multiple edge nodes, protects data privacy while improving the overall performance of the detection model, aggregates and analyzes multiple abnormal features, and outputs global abnormal data detection results, which can more comprehensively and accurately identify abnormal situations in the distribution network and improve the safety and stability of the distribution network system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0034] Figure 1 Schematic diagram of the process of detecting abnormal data in a distribution network based on edge computing shown in the present invention. DETAILED DESCRIPTION

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0036] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0038] Example 1

[0039] According to an embodiment of the present invention, Figure 1 The flowchart shown is a method for detecting abnormal data in a distribution network based on edge computing, which specifically includes the following steps:

[0040] S1: Collect distribution network data through edge devices, and form a sample data set after data preprocessing. Among them, what needs to be explained in this step is:

[0041] Edge devices include sensors, smart meters, smart circuit breakers, and data collectors.

[0042] As an example, smart meters are used to monitor electricity usage in real time, such as power consumption and power demand data; sensors are used to monitor environmental parameters, such as temperature sensors and humidity sensors; and smart circuit breakers are used to monitor power parameters such as current and voltage.

[0043] Distribution network data includes power parameters, environmental parameters, equipment status data and communication data.

[0044] As an example, power parameter data includes at least current, voltage, and power; environmental parameter data includes at least temperature and humidity; device status data includes at least switch status and device operating status; communication data includes at least communication data between devices, which is used to monitor the communication status between devices.

[0045] It is not difficult to understand that this embodiment realizes real-time monitoring and data acquisition through data collection, that is, collecting distribution network data through edge devices, which helps to timely understand the operating status of the distribution network and improve the efficiency of fault diagnosis.

[0046] In an optional embodiment, data preprocessing includes data cleaning, format conversion, feature extraction and dimensionality reduction.

[0047] It should be noted that through data preprocessing technology, the collected distribution network data can be noise-removed, the data format can be unified, key features can be extracted, and the data dimension can be reduced, providing a high-quality data foundation for subsequent model establishment.

[0048] It should be further explained in this embodiment that the sample data set can be formed by the following formula:

[0049]

[0050] Among them, X is the sample data set, m is the number of samples, n is the number of features, S is the data after data and processing, i is the sample index, j is the feature index, S ij is the value of the jth feature of the ith sample in the distribution network data, t is the integral variable, and its value range is (0, T), e t is an exponential function, represents the sum of all eigenvalues ​​of the i-th sample, It means that the eigenvalues ​​of each sample are summed and then multiplied by the exponential function for integral calculation.

[0051] This embodiment processes and integrates the distribution network data to form a sample data set, which includes the processed data and the integral calculation of the characteristic values, which helps to build an anomaly detection model and perform data analysis, and improve the intelligence level of system operation.

[0052] Preferably, this step can better monitor the operating status of the distribution network, predict potential faults and improve the response speed of the system by utilizing edge devices to collect data, perform data preprocessing and form a sample data set, thereby optimizing the management and operation efficiency of the distribution network.

[0053] Preferably, this step realizes real-time data collection and processing, reduces the amount of data transmission and processing delay, is conducive to improving data processing efficiency and reducing network burden, and provides a high-quality data foundation for subsequent abnormal feature extraction.

[0054] It should be noted that the above-mentioned data cleaning, format conversion, feature extraction and dimensionality reduction processing can be achieved by means or methods in the prior art, and will not be described in detail in this embodiment.

[0055] S2: Based on convolutional neural networks and recurrent neural networks, an abnormal feature extraction model is constructed in the edge node, and the sample data set is input into the abnormal feature extraction model for processing to extract abnormal features. It should be noted that the mathematical expression formula of the abnormal feature extraction model is as follows:

[0056]

[0057] Among them, F is the output abnormal feature, M is the abnormal feature extraction model, X is the input sample data set, σ is the activation function, m represents the number of samples, and x i represents the i-th sample, h i-1 represents the recurrent neural network output at the previous moment, Conv(·) represents the convolutional neural network processing, LSTM(·) represents the long short-term memory network processing, Indicates that the output of the convolutional neural network and the recurrent neural network is integrated.

[0058] Furthermore, abnormal features are extracted, including:

[0059] Input the sample data set through the input layer of the anomaly feature extraction model;

[0060] Use convolutional neural networks to extract spatial features from sample datasets, and use recurrent neural networks to capture temporal correlations and sequence patterns in sample datasets;

[0061] The abnormal feature extraction model gradually extracts abnormal features from the sample data set through multiple iterative learning of abnormal features.

[0062] As an example, the anomaly feature extraction model learns the inherent laws of the data, identifies abnormal patterns that are different from the normal data distribution, and then identifies abnormal features.

[0063] It is not difficult to understand that the abnormal feature extraction model built based on convolutional neural networks and recurrent neural networks can effectively extract abnormal features from distribution network data, help detect and identify abnormal data, and improve the safety and stability of the distribution network system.

[0064] This implementation builds an abnormal feature extraction model by combining convolutional neural networks and recurrent neural networks. Convolutional neural networks are used to extract spatial features and recurrent neural networks are used to capture temporal correlations and sequence patterns. This helps to comprehensively mine abnormal features in sample data sets and improve the accuracy and efficiency of anomaly detection.

[0065] Preferably, this step uses an abnormal feature extraction model, inputs a sample data set, and uses a convolutional neural network and a recurrent neural network to extract spatial features and time correlations respectively. The abnormal features are gradually extracted through multiple iterative learning, which helps to discover abnormal patterns and abnormal behaviors in the data. By combining spatial features and time series information, the data can be analyzed more comprehensively, thereby improving the accuracy and robustness of anomaly detection.

[0066] S3: Establish a global abnormal data detection model based on the federated learning algorithm, aggregate and analyze multiple abnormal features, and output the detection results. Among them, what needs to be explained in this step is:

[0067] Define that there are N edge nodes, and the abnormal feature extracted from each node is F pq , where p∈{1,2,…,N} represents the node number, q represents the dimension of the feature vector, and each abnormal feature is aggregated and analyzed to obtain the global abnormal feature Z q , its mathematical expression formula is as follows:

[0068]

[0069] Among them, w p represents the weight of the pth edge node;

[0070] In order to ensure that the sum of the weights is 1, normalization is required, namely:

[0071]

[0072] Among them, r p It represents the score of the pth edge node, and its mathematical expression formula is as follows:

[0073]

[0074] Among them, θ represents the parameter vector of the global model, represents the sigmoid function, F p represents the parameter vector of the local model of the p-th edge node,

[0075] Output test results, including:

[0076] y=f(Z)

[0077]

[0078] Where Z is the global anomaly feature, f(·) is the classification function, y is the detection result, which is a probability vector, that is, each element represents the probability of the corresponding anomaly category, which can be calculated by the softmax function, K represents the number of anomaly categories, and q represents the dimension of the feature vector.

[0079] The detection result y is judged as abnormal by setting a threshold. When the probability of the abnormal category is ≥ the threshold, it is judged as abnormal.

[0080] It should be noted that the threshold value can be set according to the actual application situation on site, and this embodiment does not impose a unique restriction on it.

[0081] In an optional implementation, a federated learning framework is established to conduct federated training on the abnormal feature extraction models of multiple edge nodes and establish a global abnormal data detection model.

[0082] During the federated learning process, each edge node only shares model parameters but not data to protect data privacy and security.

[0083] In an optional implementation, aggregate analysis may be performed by weighted averaging or voting methods to improve the accuracy of the detection results, such as aggregating the abnormal features of each edge node according to its weight to obtain a global abnormal feature.

[0084] As an example, the detection result includes at least the type, degree, and location information of the abnormality.

[0085] As an example, the type of abnormality: such as voltage abnormality, current abnormality.

[0086] As an example, the degree of abnormality: such as slight abnormality, moderate abnormality, severe abnormality.

[0087] As an example, the location of the anomaly is enclosed: such as the node or area where the anomaly occurred.

[0088] It should be noted that this step uses the federated learning algorithm to establish a global anomaly data detection model, which can learn and detect anomalies from the data of multiple participants, help to comprehensively consider the information of different data sources, and improve the global vision and accuracy of anomaly detection; at the same time, aggregating and analyzing multiple abnormal features helps to comprehensively consider the information of different features, improve the robustness and accuracy of anomaly detection, and can more comprehensively analyze abnormal patterns in the data.

[0089] It should be further explained that this step judges the abnormality of the detection result by setting a threshold. When the probability of the abnormal category exceeds the threshold, it is judged as an abnormality. This method can automatically identify abnormalities based on probability information, thereby improving the efficiency and practicality of detection.

[0090] Preferably, this step applies a federated learning algorithm to establish a global abnormal data detection model, aggregates and analyzes multiple abnormal features, and finally outputs the abnormal detection results. Federated learning can realize data sharing and model training among multiple edge nodes, protect data privacy, and improve the overall performance of the detection model; through the establishment of a global model and comprehensive analysis of abnormal features, more accurate and comprehensive detection of abnormal situations in the distribution network is achieved, thereby improving the safety and stability of the distribution network system.

[0091] Preferably, this embodiment improves the efficiency and accuracy of abnormal data detection in the distribution network through the combination of edge computing technology and deep learning algorithm, which is beneficial to timely discover and handle abnormal situations in the distribution network, thereby improving the safety and stability of the distribution network system.

[0092] It should be noted that other aspects disclosed in the embodiments of the present invention also propose a distribution network abnormal data detection system based on edge computing, including: one or more processors and a memory.

[0093] The memory is used to store operable instructions, which, when executed by the one or more processors, cause the one or more processors to perform operations, including the process of a method for detecting abnormal data in a distribution network based on edge computing in the aforementioned embodiment, in particular Figure 1 The process of the method shown.

[0094] The detection system disclosed in the embodiment of the present invention also includes a computer-readable medium storing software, wherein the software includes instructions that can be executed by one or more computers, and the instructions enable one or more computers to perform operations through such execution, and the operations include the process of a distribution network abnormal data detection method based on edge computing in the aforementioned embodiment, especially Figure 1 The process of the method shown.

[0095] It should be appreciated that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory.

[0096] The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner.

[0097] Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system.

[0098] However, the program can be implemented in assembly or machine language, if desired.

[0099] In any case, the language may be a compiled or interpreted language.

[0100] Furthermore, the program can be run on an application specific integrated circuit programmed for this purpose.

[0101] The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that executes collectively on one or more processors, by hardware, or a combination thereof.

[0102] The computer program includes a plurality of instructions executable by one or more processors.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting abnormal data in distribution network based on edge computing. It is characterized in that include: The distribution network data is collected through edge devices, and the distribution network data is formed into a sample data set after data preprocessing; include: Among them, X is the sample data set, m is the number of samples, n is the number of features, S is the data after data and processing, i is the sample index, j is the feature index, S ij is the value of the jth feature of the ith sample in the distribution network data, t is the integral variable, and its value range is (0, T), e t is an exponential function, represents the sum of all eigenvalues ​​of the i-th sample, It means that the eigenvalues ​​of each sample are summed and then multiplied by the exponential function for integral calculation; Based on convolutional neural networks and recurrent neural networks, an abnormal feature extraction model is constructed in the edge node, the sample data set is input into the abnormal feature extraction model for processing, and abnormal features are extracted; including: Inputting the sample data set through the input layer of the abnormal feature extraction model; Extracting spatial features from the sample data set using a convolutional neural network and capturing temporal correlations and sequence patterns in the sample data set using a recurrent neural network; The abnormal feature extraction model gradually extracts the abnormal features in the sample data set through multiple iterative learning of the abnormal features; Constructing the abnormal feature extraction model includes: Among them, F is the output abnormal feature, M is the abnormal feature extraction model, X is the input sample data set, σ is the activation function, m represents the number of samples, and x i represents the i-th sample, h i-1 represents the recurrent neural network output at the previous moment, Conv(·) represents the convolutional neural network processing, LSTM(·) represents the long short-term memory network processing, Indicates that the output of the convolutional neural network and the recurrent neural network is integrated; A global abnormal data detection model is established based on the federated learning algorithm, multiple abnormal features are aggregated and analyzed, and the detection results are output; including: Define that there are N edge nodes, and the abnormal feature extracted from each node is F pq , where p∈1,2,…,N} represents the node number, q represents the dimension of the feature vector, and each abnormal feature is aggregated and analyzed to obtain the global abnormal feature Z q , its mathematical expression formula is as follows: Among them, w p represents the weight of the pth edge node; In order to ensure that the sum of the weights is 1, normalization is required, namely: Among them, r p It represents the score of the pth edge node, and its mathematical expression formula is as follows: Among them, θ represents the parameter vector of the global model, represents the sigmoid function, F p represents the parameter vector of the local model of the p-th edge node, Output test results, including: y=f(Z q Among them, Z q is the global anomaly feature, f(·) is the classification function, y is the detection result, which is a probability vector, that is, each element represents the probability of the corresponding anomaly category, which can be calculated by the softmax function, K represents the number of anomaly categories, and q represents the dimension of the feature vector. The detection result y is judged as abnormal by setting a threshold. When the probability of the abnormal category is ≥ the threshold, it is judged as abnormal.

2. According to the method for detecting abnormal data in distribution network based on edge computing according to claim 1, It is characterized in that The edge devices include sensors, smart meters, smart circuit breakers and data collectors; The distribution network data includes power parameters, environmental parameters, equipment status data and communication data.

3. According to the method for detecting abnormal data in distribution network based on edge computing in claim 1, It is characterized in that The data preprocessing includes data cleaning, format conversion, feature extraction and dimensionality reduction.

4. A distribution network abnormal data detection system based on edge computing, It is characterized in that include: one or more processors; A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors perform operations, wherein the operations include the process of the method for detecting abnormal data in a distribution network based on edge computing as described in any one of claims 1 to 3.

Citation Information

Cited By

  • Industrial Internet of Things anomaly detection system and method for edge nodes

    CN122601436A