Power distribution auxiliary monitoring and early warning system

By collecting and processing the operating parameters and temperature strain data of power grid nodes in real time in the distribution system, and using advanced algorithms and models, the problems of data lag and insufficient accuracy of fault identification in traditional distribution system monitoring methods are solved, and the problem of data lag and fault identification is insufficient. Efficient fault diagnosis and early warning improves the operating reliability and maintenance efficiency of the power grid.

CN119051264APending Publication Date: 2024-11-29ZHEJIANG WELLSUN INTELLIGENT TECH CO LTD
View PDF 0 Cites 23 Cited by

Patent Information

Application Number
CN202411168001.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Traditional distribution system monitoring methods have problems such as data acquisition and processing lag, insufficient accuracy of fault identification, slow early warning response, and low model training and optimization efficiency, which are difficult to meet the real-time and dynamic needs of power grid operation.

Method used

A distribution auxiliary monitoring and early warning system was designed, including a data acquisition module, a multi-point temperature strain monitoring module, a data processing module, a power grid topology analysis module and a communication module. By collecting and processing the operating parameters and temperature strain data of the power grid nodes in real time, Using local anomaly factor algorithm, fast Fourier transform, support vector machine and graph convolution network, the generation and transmission of fault diagnosis and early warning signals are realized.

Benefits of technology

It improves the accuracy and response speed of fault diagnosis, ensures real-time monitoring of the operating status of the power grid and accurate evaluation of fault propagation paths, enhances the reliability and maintenance efficiency of the power grid, and reduces the risk of fault expansion.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119051264A_ABST
    Figure CN119051264A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power distribution monitoring, in particular to a power distribution auxiliary monitoring and early warning system, which comprises the steps of acquiring operation parameters of each node in a power distribution system in real time, acquiring temperature and strain data in real time by utilizing sensors arranged at a plurality of nodes of the power distribution system; and recognizing a complex temperature and strain mode, carrying out fault diagnosis on the acquired operation parameters and temperature and strain data, generating an early warning signal according to a fault diagnosis result, establishing a topological structure model of the power grid, and providing real-time topological structure change and fault influence evaluation. The method can efficiently identify and classify fault modes, thereby improving the accuracy and response speed of fault diagnosis, ensuring that the model keeps high performance and reliability in a complex and dynamic power distribution environment, improving the accuracy of power grid operation state monitoring and the accuracy of fault propagation path simulation, and improving the reliability of fault diagnosis. And powerful technical support is provided for reliable operation of a power grid.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution monitoring, and in particular to a power distribution auxiliary monitoring and early warning system. Background Art

[0002] The distribution system is an important part of the power system, and the safety and reliability of its operation are directly related to the stability of power supply. However, the traditional distribution system monitoring and early warning methods have the following problems and technical challenges:

[0003] However, the inventors have discovered that this technical solution still has at least the following defects:

[0004] First, data collection and processing lag: Traditional distribution system monitoring usually relies on data collection from a small number of key nodes. This method has limited coverage and is difficult to fully reflect the operating status of the entire distribution system. In addition, data processing speed lags and cannot reflect system changes in real time, making it difficult to detect and handle potential faults in a timely manner.

[0005] Second, the accuracy of fault identification is insufficient: Traditional fault identification methods mostly rely on simple statistical analysis and rule matching, which show obvious limitations when dealing with complex fault modes and small abnormal changes. Therefore, the accuracy and timeliness of fault diagnosis often cannot meet actual needs, which can easily lead to the spread of faults and affect the overall operation of the system.

[0006] Third, the early warning system is slow to respond: Traditional early warning systems usually rely on preset fixed thresholds and manual judgment, and lack the ability to adjust and optimize in real time. When an abnormal situation occurs in the power grid, the timeliness and accuracy of the early warning signal are low, making it difficult to notify maintenance personnel in time for processing, resulting in untimely fault handling.

[0007] Low efficiency of model training and optimization;

[0008] Fourth, the training and optimization of traditional models usually rely on fixed optimization algorithms and manual parameter adjustment, lacking adaptive and intelligent tuning mechanisms. As the structure and operating conditions of power grids continue to change, traditional models often find it difficult to maintain efficient fault detection and early warning capabilities, resulting in unstable model performance and insufficient generalization capabilities. Summary of the invention

[0009] Based on the above purpose, the present invention provides a power distribution auxiliary monitoring and early warning system, a power distribution auxiliary monitoring and early warning system, including the following steps.

[0010] Optionally, it includes a data acquisition module, a multi-point temperature strain monitoring module, a data processing module, a power grid topology analysis module and a communication module, wherein;

[0011] The data acquisition module collects the operating parameters of each node in the power distribution system in real time, including current and voltage values;

[0012] The multi-point temperature and strain monitoring module uses sensors arranged at multiple nodes of the power distribution system to collect temperature and strain data in real time and identify complex temperature and strain patterns;

[0013] The data processing module performs fault diagnosis on the collected operating parameters and temperature and strain data, and generates a warning signal according to the result of the fault diagnosis;

[0014] The power grid topology analysis module establishes a topology model of the power grid, analyzes the operation status and fault propagation path of the power grid, and provides real-time topology changes and fault impact assessment;

[0015] The communication module is used to transmit the warning signal to the remote monitoring center and the mobile device of the maintenance personnel.

[0016] Optionally, the data acquisition module includes:

[0017] Sensor deployment: Multiple sensor units are deployed at each node of the power distribution system, and each sensor unit collects the operating parameters of the node in real time;

[0018] Data preprocessing: preprocess the real-time collected operating parameters, including data denoising, data smoothing and data standardization, including:

[0019] Data denoising uses filters or noise removal algorithms to remove high-frequency noise and random interference from operating parameters;

[0020] Data transmission: real-time transmission of pre-processed operating parameters;

[0021] Through sensor layout, data preprocessing and real-time data transmission, efficient, accurate and real-time monitoring of the operating parameters of each node is ensured.

[0022] Optionally, the multi-point temperature strain monitoring module includes:

[0023] Temperature and strain data collection: Temperature sensors and strain sensors are arranged at multiple nodes of the power distribution system to collect temperature and strain data at the nodes;

[0024] Preliminary data processing: Preliminary processing of the collected temperature and strain data, including removal of noise and outliers;

[0025] Temperature and strain data analysis: The temperature and strain data after preliminary processing are analyzed using the Local Outlier Factor (LOF) algorithm to identify complex temperature and strain patterns.

[0026] Optionally, the local outlier factor (LOF) algorithm includes:

[0027] Data preprocessing: standardize the temperature and strain data and form temperature and strain data sets;

[0028] Select parameter k: The parameter k in the local outlier factor (LOF) algorithm, i.e. the number of neighbors of each point, is determined by cross-validation.

[0029] Calculate k-distance and k-distance neighbors: For each data point p in the temperature and strain dataset, find its k-distance k_dist(p) and k-distance neighbor set N k (p), expressed as:

[0030] k_dist(p)=min{r:|N k (p)|=k};

[0031] Among them, N k (p) is the set of k nearest neighbors of point p;

[0032] Calculate the reachable distance: For each point o in the temperature and strain data set relative to point p, calculate the reachable distance, expressed as: reach_dist k (p,o)=max(k_dist(o),d(p,o)), where d(p,o) is the Euclidean distance between points p and o;

[0033] Calculate the local reachable density (LRD): For each data point p, calculate its local reachable density, expressed as:

[0034]

[0035] Calculate the local anomaly factor: For each data point p, calculate its local anomaly factor, expressed as:

[0036] According to LOF k The value of (p) determines whether point p is an outlier;

[0037] Identify complex patterns and extract key features: Based on the calculated LOF values, identify points with high LOF values ​​as abnormal points, analyze the distribution and patterns of abnormal points, extract temperature and strain features, including the extraction of temperature sharp rise points and strain mutation points, perform pattern recognition on the extracted temperature sharp rise points and strain mutation points, identify frequently occurring abnormal patterns and features, and determine their causes and impact ranges;

[0038] The pattern recognition specifically includes:

[0039] Feature vector construction: The data of each temperature rise point and strain mutation point are expressed as a feature vector x, x = [ΔT i ,Δ i ];

[0040] Where, ΔT i is the temperature change at the i-th moment, Δ i is the strain change at the i-th moment;

[0041] K-means cluster analysis: Use the K-means clustering algorithm to cluster feature vectors and identify frequently occurring abnormal patterns and features, expressed as:

[0042] Among them, C k is the kth cluster, μ k is the center point of the kth cluster;

[0043] Abnormal pattern recognition: Analyze each clustering result and identify frequently occurring abnormal patterns, expressed as:

[0044]

[0045] Among them, P(C k |x) is the feature vector x belongs to cluster C k The probability of N k is cluster C k The number of data points;

[0046] Pattern feature extraction: Extract key features in each cluster and analyze the frequency, amplitude and distribution of temperature rise points and strain mutation points, expressed as:

[0047]

[0048] Among them, μ k is cluster C k The mean eigenvector of reflects the average characteristics of the cluster;

[0049] Cause analysis and impact assessment: By analyzing the distribution of cluster centers and eigenvectors, the causes and impact of abnormal patterns are determined, expressed as:

[0050]

[0051] Among them, R k is cluster C k The average impact range of the cluster reflects the breadth of abnormal distribution of the cluster.

[0052] Optionally, the data processing module includes:

[0053] Feature extraction: Utilize power fluctuation detection and use fast Fourier transform (FFT) to analyze the spectrum of the current signal and extract the current fluctuation characteristics, which are expressed as:

[0054]

[0055] Where X(f) is the spectrum, x[n] is the time domain signal, and N is the number of sampling points.

[0056] Pattern recognition: The extracted features are classified using the support vector machine (SVM) algorithm to preliminarily identify potential failure modes, expressed as:

[0057]

[0058] subjecttoy i (w T φ(x i )+b)≥1-ξ i ,ξ i ≥0;

[0059] Where w is the weight vector, b is the bias, ξ i is the slack variable, C is the penalty parameter, φ(x i ) is the kernel function;

[0060] Fault type judgment: The identified fault mode is judged using the naive Bayes classifier to determine the specific type of fault (such as short circuit, overload, equipment aging, etc.), which is expressed as:

[0061]

[0062] Among them, P(C k |x) is the given data x belongs to category C k The probability of P(x|C k ) is in Category C k The likelihood of the data x, P(C k ) is Category C k The prior probability of , P(x) is the marginal probability of data x;

[0063] Fault severity assessment: The severity of the fault is assessed based on the fuzzy logic algorithm to quantify the impact of the fault on the distribution system, expressed as:

[0064]

[0065] Where S is the fault severity, w i is the weight of the i-th fuzzy rule, μ i (x) is the i-th fuzzy membership function;

[0066] Warning signal generation: Based on the results of fault diagnosis, a warning signal is generated. The warning signal includes the fault type, location, severity and recommended maintenance measures, expressed as:

[0067] Alert=(T fault ,L fault ,S fault, ,M suggest );

[0068] Among them, T fault is the fault type, L fault is the fault location, S fault is the fault severity, M suggest is a recommended maintenance action.

[0069] Optionally, the power grid topology analysis module uses a graph convolutional network (GCN) to model the topological structure of the power grid, including:

[0070] Data collection: Collect topological data of the distribution system, including the connection relationship between nodes (such as substations, distribution equipment) and edges (such as power lines) and their attributes (such as resistance, capacitance, load);

[0071] Graph structure construction: The collected topological data is represented as a graph structure G = (V, E), where V is a set of nodes, E is a set of edges, and each node and edge contains its specific attribute information;

[0072] Adjacency matrix construction: Construct the adjacency matrix A of the graph, where A ij Represents the connection relationship between node i and node j. If there is a connection between node i and node j, then A ij ≠0; otherwise, A ij =0;

[0073] Node feature matrix construction: Construct the node feature matrix X. The feature vector of each node contains its attribute information, including current, voltage, and load.

[0074] Graph convolution layer design: Design a graph convolution layer to perform convolution operations on the node features in the graph structure, expressed as:

[0075] in, is the adjacency matrix with self-loops added, yes The degree matrix, H (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the activation function;

[0076] Model building and debugging: Use the collected historical running data to train the graph convolutional network model, and use the validation set data to evaluate the trained GCN model;

[0077] Real-time topology modeling: Real-time monitoring data is input into the trained GCN model, node characteristics and topology structure are updated in real time, and the current topological status and operating characteristics of the power grid are provided.

[0078] Optionally, the model building and debugging includes:

[0079] Model training: Use the training set data to train the GCN model and adjust the model parameters through the optimizer Adam;

[0080] Hyperparameter tuning: Use validation set data to perform hyperparameter tuning and optimize the model’s hyperparameters (such as learning rate, number of layers, and number of hidden units);

[0081] Model validation: Use validation set data to evaluate the trained GCN model to ensure its accuracy in topological structure modeling and calculate performance indicators on the validation set, including accuracy, recall, and F1 score;

[0082] Accuracy:

[0083] Among them, TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example;

[0084] Precision:

[0085] Recall:

[0086] F1 Score:

[0087] Loss function: For graph convolutional networks, the commonly used loss function is cross entropy loss:

[0088]

[0089] Among them, y i is the actual label, is the predicted probability, N is the total number of samples;

[0090] Model testing: Use the test set data to evaluate the performance of the final model and verify its performance on unseen data;

[0091] Model deployment and monitoring: Deploy the trained and verified GCN model to the real-time monitoring platform of the power distribution system to analyze the power grid topology and operating status in real time;

[0092] Continuously monitor the performance of the model, and regularly update and optimize the model.

[0093] Optionally, model training with the Adam optimizer includes:

[0094] Initialize the first-order moment estimate and the second-order moment estimate, expressed as:

[0095] m t =β1m t-1 +(1-β1)g t ;

[0096]

[0097] Among them, m t and v t are the first-order and second-order moment estimates, g t is the current gradient, β1 and β2 are the decay rates;

[0098] Bias correction, expressed as:

[0099]

[0100] Parameter update, expressed as:

[0101] Among them, θ t is the parameter for the tth iteration, α is the learning rate, and is a small constant to prevent division by zero.

[0102] Optionally, analyze the operating status of the power grid and the fault propagation path, provide real-time topology changes and fault impact assessment, including:

[0103] Operation status evaluation: Based on the results of graph convolution operations, the feature vector of each node is analyzed to evaluate the operation status of the node, detect abnormal changes in operation parameters such as current, voltage, and load, and identify potential faulty nodes;

[0104] Fault propagation path simulation: After identifying the faulty node, the GCN model is used to simulate the fault propagation path in the power grid. By dynamically updating the adjacency matrix and node features, the possible path of the fault propagating from the initial node to other nodes is predicted.

[0105] Propagation probability calculation: Calculate the probability of each node being affected during the fault propagation process. Let the initial fault node be i and the probability of propagating to node j be P ij , expressed as:

[0106]

[0107] Among them, N(j) is the set of neighbor nodes of node j, Ajk is the connection relationship between nodes j and k, H k is the feature vector of node k;

[0108] Fault impact scope assessment: Based on the fault propagation path and propagation probability, the fault impact scope is assessed, the affected nodes and edges are determined, and the degree of their impact is calculated.

[0109] Optionally, the communication module includes outputting the results of operation status assessment and fault propagation path simulation through the communication module; generating a detailed fault impact report, including fault nodes, propagation paths, impact range and recommended countermeasures, and generating an early warning signal to notify the remote monitoring center and maintenance personnel for timely processing.

[0110] Beneficial effects of the present invention:

[0111] The present invention arranges temperature and strain sensors at multiple nodes of the power distribution system, collects temperature and strain data of each node in real time, and removes noise and outliers through data preprocessing to ensure the accuracy of the data. The local outlier factor (LOF) algorithm is used to analyze the temperature and strain data after preliminary processing, which can identify complex temperature and strain patterns and accurately locate abnormal points. Combining the fast Fourier transform (FFT) spectrum analysis of the current signal and the support vector machine (SVM) algorithm to classify the features, it can efficiently identify and classify fault modes, determine the type, location and severity of the fault, thereby improving the accuracy and response speed of fault diagnosis.

[0112] The present invention evaluates the severity of the fault through a fuzzy logic algorithm, quantifies the impact of the fault on the power distribution system, and generates an early warning signal, which contains detailed fault type, location, severity and recommended maintenance measures. The early warning signal can be transmitted to the remote monitoring center and maintenance personnel in a timely manner to ensure rapid response and processing. At the same time, the Adam optimizer is used for model training, and efficient optimization of the adaptive learning rate is achieved by initializing first-order and second-order moment estimation, deviation correction and parameter updating. Hyperparameter tuning is performed using validation set data to optimize the model's hyperparameters (such as learning rate, number of layers, number of hidden units), improve the model's performance and generalization ability, thereby ensuring that the model maintains high performance and reliability in a complex and dynamic power distribution environment.

[0113] The present invention deploys the trained and verified graph convolutional network (GCN) model to the real-time monitoring platform of the power distribution system to analyze the topology and operating status of the power grid in real time. By continuously monitoring the operating effect of the model, the model is regularly updated and optimized to adapt to changes in the power grid structure and operating conditions, ensuring the long-term stable operation of the system. The system can provide real-time topology changes and fault impact assessments, improve the accuracy of power grid operating status monitoring and the accuracy of fault propagation path simulation, and provide strong technical support for the reliable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0115] Figure 1 This is a schematic diagram of the functional modules of the early warning system according to an embodiment of the present invention;

[0116] Figure 2 Schematic diagram of a multi-point temperature strain monitoring module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0117] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0118] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0119] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0120] like Figure 1-Figure 2 As shown, a power distribution auxiliary monitoring and early warning system includes a data acquisition module, a multi-point temperature strain monitoring module, a data processing module, a power grid topology analysis module and a communication module, wherein;

[0121] The data acquisition module collects the operating parameters of each node in the power distribution system in real time, including current and voltage values;

[0122] The multi-point temperature and strain monitoring module uses sensors arranged at multiple nodes of the power distribution system to collect temperature and strain data in real time and identify complex temperature and strain patterns;

[0123] The data processing module performs fault diagnosis on the collected operating parameters, temperature and strain data, and generates an early warning signal based on the results of the fault diagnosis;

[0124] The power grid topology analysis module establishes a topological structure model of the power grid, analyzes the operation status and fault propagation path of the power grid, and provides real-time topological structure changes and fault impact assessment;

[0125] The communication module is used to transmit the warning signal to the remote monitoring center and the mobile devices of the maintenance personnel.

[0126] Through the above content, real-time monitoring of the power grid operation status and fault warning are realized. The system can comprehensively collect and analyze the operating parameters of each node, improve the reliability and maintenance efficiency of the power grid, reduce the risk of fault occurrence and expansion, and ensure the safe and stable operation of the power system.

[0127] The data acquisition module includes:

[0128] Sensor deployment: Multiple sensor units are deployed at each node of the power distribution system, and each sensor unit collects the operating parameters of the node in real time;

[0129] Data preprocessing: preprocess the real-time collected operating parameters, including data denoising, data smoothing and data standardization, including:

[0130] Data denoising uses filters or noise removal algorithms to remove high-frequency noise and random interference from operating parameters;

[0131] Data smoothing smoothes the operating parameters to reduce data fluctuations and extract more stable signals;

[0132] Data standardization: Standardize the operating parameters to facilitate subsequent processing and analysis;

[0133] Data transmission: real-time transmission of pre-processed operating parameters;

[0134] Through sensor layout, data preprocessing and real-time data transmission, efficient, accurate and real-time monitoring of the operating parameters of each node is ensured.

[0135] The multi-point temperature strain monitoring module includes:

[0136] Temperature and strain data collection: Temperature sensors and strain sensors are arranged at multiple nodes of the power distribution system to collect temperature and strain data at the nodes;

[0137] Preliminary data processing: Preliminary processing of the collected temperature and strain data, including removal of noise and outliers;

[0138] Temperature and strain data analysis: Analyze the initially processed temperature and strain data using the Local Outlier Factor (LOF) algorithm to identify complex temperature and strain patterns;

[0139] The benefit of this system is that it can accurately collect and process multi-point temperature and strain data, identify abnormal patterns, thereby improving the sensitivity and accuracy of fault detection, effectively warning of potential problems, and ensuring the reliable operation of the distribution system.

[0140] The Local Outlier Factor (LOF) algorithm includes:

[0141] Data preprocessing: The temperature and strain data are standardized and formed into temperature and strain data sets, which are expressed as:

[0142]

[0143] Among them, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data;

[0144] Select parameter k: The parameter k in the local outlier factor (LOF) algorithm, i.e. the number of neighbors of each point, is determined by cross-validation.

[0145] Calculate k-distance and k-distance neighbors: For each data point p in the temperature and strain dataset, find its k-distance k_dist(p) and k-distance neighbor set N k (p), expressed as:

[0146] k_dist(p)=min{r:|N k (p)|=k};

[0147] Among them, N k (p) is the set of k nearest neighbors of point p;

[0148] Calculate the reachable distance: For each point o in the temperature and strain data set relative to point p, calculate the reachable distance, expressed as: reach_dist k (p,o)=max(k_dist(o),d(p,o)), where d(p,o) is the Euclidean distance between points p and o;

[0149] Calculate the local reachable density (LRD): For each data point p, calculate its local reachable density, expressed as:

[0150]

[0151] Calculate the local anomaly factor: For each data point p, calculate its local anomaly factor, expressed as:

[0152] According to LOF k The value of (p) determines whether point p is an outlier;

[0153] If LOF k (p) value is significantly greater than 1, then the point p may be an outlier;

[0154] Identify complex patterns and extract key features: Based on the calculated LOF values, identify points with high LOF values ​​as abnormal points, analyze the distribution and patterns of abnormal points, extract temperature and strain features, including the extraction of temperature sharp rise points and strain mutation points, perform pattern recognition on the extracted temperature sharp rise points and strain mutation points, identify frequently occurring abnormal patterns and features, and determine their causes and impact ranges;

[0155] Pattern recognition specifically includes:

[0156] Feature vector construction: The data of each temperature rise point and strain mutation point are expressed as a feature vector x, x = [ΔT i ,Δ i ];

[0157] Where, ΔT i is the temperature change at the i-th moment, Δ i is the strain change at the i-th moment;

[0158] K-means cluster analysis: Use the K-means clustering algorithm to cluster feature vectors and identify frequently occurring abnormal patterns and features, expressed as:

[0159] Among them, C k is the kth cluster, μ k is the center point of the kth cluster;

[0160] Abnormal pattern recognition: Analyze each clustering result and identify frequently occurring abnormal patterns, expressed as:

[0161]

[0162] Among them, P(C k |x) is the feature vector x belongs to cluster C k The probability of N k is cluster C k The number of data points;

[0163] Pattern feature extraction: Extract key features in each cluster and analyze the frequency, amplitude and distribution of temperature rise points and strain mutation points, expressed as:

[0164]

[0165] Among them, μ k is cluster C k The mean eigenvector of reflects the average characteristics of the cluster;

[0166] Cause analysis and impact assessment: By analyzing the distribution of cluster centers and eigenvectors, the causes and impact of abnormal patterns are determined, expressed as:

[0167]

[0168] Among them, R k is cluster C k The average impact range of the cluster reflects the breadth of abnormal distribution of the cluster;

[0169] a. Extraction of sharp temperature rise points: Use the first-order difference method to detect the mutation point of temperature data, expressed as:

[0170] ΔT i =T i+1 -T i ;

[0171] Among them, T i is the temperature value at the i-th moment, ΔT i is the temperature change, set the threshold ΔT i , if ΔT i >ΔΔT th , it is determined as the point where the temperature rises sharply;

[0172] b. Extraction of strain mutation points: The mutation points of strain data are detected using the first-order difference method, expressed as:

[0173] Δ i = i+1 - i ;

[0174] in, i is the strain value at the i-th moment, Δ i is the strain change, set the threshold Δ th , if Δ i >Δ th , it is determined to be a strain mutation point;

[0175] Through the above content, it is possible to accurately identify and locate abnormal data points, deeply analyze complex failure modes, and extract key features, thereby improving the sensitivity and accuracy of temperature and strain monitoring and effectively warning of potential failures.

[0176] The data processing module includes:

[0177] Feature extraction: Utilize power fluctuation detection and use fast Fourier transform (FFT) to analyze the spectrum of the current signal and extract the current fluctuation characteristics, which are expressed as:

[0178]

[0179] Where X(f) is the spectrum, x[n] is the time domain signal, and N is the number of sampling points.

[0180] Pattern recognition: The extracted features are classified using the support vector machine (SVM) algorithm to preliminarily identify potential failure modes, expressed as:

[0181]

[0182] subjecttoy i (w T φ(x i )+b)≥1-ξ i ,ξ i ≥0;

[0183] Where w is the weight vector, b is the bias, ξ i is the slack variable, C is the penalty parameter, φ(x i ) is the kernel function;

[0184] Fault type judgment: The identified fault mode is judged using the naive Bayes classifier to determine the specific type of fault (such as short circuit, overload, equipment aging, etc.), which is expressed as:

[0185]

[0186] Among them, P(C k |x) is the given data x belongs to category Ck The probability of P(x|C k ) is in Category C k The likelihood of the data x, P(C k ) is Category C k The prior probability of , P(x) is the marginal probability of data x;

[0187] Fault severity assessment: The severity of the fault is assessed based on the fuzzy logic algorithm to quantify the impact of the fault on the distribution system, expressed as:

[0188]

[0189] Where S is the fault severity, w i is the weight of the i-th fuzzy rule, μ i (x) is the i-th fuzzy membership function;

[0190] Warning signal generation: Based on the results of fault diagnosis, a warning signal is generated. The warning signal includes the fault type, location, severity and recommended maintenance measures, expressed as:

[0191] Alert=(T fault ,L fault ,S fault, ,M suggest );

[0192] Among them, T fault is the fault type, L fault is the fault location, S fault is the severity of the fault,

[0193] M suggest is a recommended maintenance action;

[0194] Through the synergy of multiple algorithms, accurate detection, type judgment and severity assessment of distribution system faults are achieved, and early warning signals are generated, providing strong support for timely maintenance and repairs, and improving the reliability and safety of the system.

[0195] The power grid topology analysis module uses graph convolutional networks (GCN) to model the topology of the power grid, including:

[0196] Data collection: Collect topological data of the distribution system, including the connection relationship between nodes (such as substations, distribution equipment) and edges (such as power lines) and their attributes (such as resistance, capacitance, load);

[0197] Graph structure construction: The collected topological data is represented as a graph structure G = (V, E), where V is a set of nodes, E is a set of edges, and each node and edge contains its specific attribute information;

[0198] Adjacency matrix construction: Construct the adjacency matrix A of the graph, where A ij Represents the connection relationship between node i and node j. If there is a connection between node i and node j, then A ij ≠0; otherwise, A ij =0;

[0199] Node feature matrix construction: Construct the node feature matrix X. The feature vector of each node contains its attribute information, including current, voltage, and load.

[0200] Graph convolution layer design: Design a graph convolution layer to perform convolution operations on the node features in the graph structure, expressed as:

[0201] in, is the adjacency matrix with self-loops added, yes The degree matrix, H (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the activation function;

[0202] Model building and debugging: Use the collected historical running data to train the graph convolutional network model, and use the validation set data to evaluate the trained GCN model to ensure its accuracy and generalization ability in topological structure modeling.

[0203] Real-time topology modeling: Real-time monitoring data is input into the trained GCN model, node characteristics and topology structure are updated in real time, and the current topological status and operating characteristics of the power grid are provided.

[0204] Through the above steps, the power grid topology analysis module uses the graph convolutional network to accurately model the topological structure of the power grid, which can effectively capture the complex relationship between power grid nodes and edges and improve the monitoring and analysis capabilities of the distribution system.

[0205] Model building and debugging includes:

[0206] Model training: Use the training set data to train the GCN model, and adjust the model parameters through the optimizer Adam so that the model can accurately fit the node features and topological structure relationships;

[0207] Hyperparameter tuning: Use validation set data to perform hyperparameter tuning and optimize the model’s hyperparameters (such as learning rate, number of layers, number of hidden units) to improve the model’s performance and generalization ability.

[0208] Model validation: Use validation set data to evaluate the trained GCN model to ensure its accuracy in topological structure modeling and calculate performance indicators on the validation set, including accuracy, recall, and F1 score;

[0209] Accuracy:

[0210] Among them, TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example.

[0211] Precision:

[0212] Recall:

[0213] F1 Score:

[0214] Loss function: For graph convolutional networks, the commonly used loss function is cross entropy loss:

[0215]

[0216] Among them, y i is the actual label, is the predicted probability, N is the total number of samples;

[0217] Model testing: Use the test set data to evaluate the performance of the final model, verify its performance on unseen data, and ensure the generalization ability and reliability of the model;

[0218] Model deployment and monitoring: Deploy the trained and verified GCN model to the real-time monitoring platform of the power distribution system to analyze the power grid topology and operating status in real time;

[0219] Continuously monitor the operating results of the model, and regularly update and optimize the model to adapt to changes in grid structure and operating conditions.

[0220] Through the above steps, the power grid topology analysis module can use the graph convolutional network to train and verify the topological structure of the power grid, and ensure the accuracy and reliability of the model through a scientific verification algorithm, thereby improving the monitoring and analysis capabilities of the distribution system.

[0221] Model training through the optimizer Adam includes:

[0222] Initialize the first-order moment estimate and the second-order moment estimate, expressed as:

[0223] m t =β1m t-1 +(1-β1)g t ;

[0224]

[0225] Among them, m t and vt are the first-order and second-order moment estimates, g t is the current gradient, β1 and β2 are the decay rates;

[0226] Bias correction, expressed as:

[0227]

[0228] Parameter update, expressed as:

[0229] Among them, θ t is the parameter of the tth iteration, α is the learning rate, and is a small constant to prevent division by zero;

[0230] The Adam optimizer combines the advantages of first-order and second-order moment estimation. Through bias correction and adaptive learning rate, it makes parameter updates more robust and efficient. It can converge quickly and adapt to different gradient changes, thereby improving the effect and performance of model training. It is particularly suitable for data with noisy and sparse gradients.

[0231] Analyze the operation status and fault propagation path of the power grid, provide real-time topology changes and fault impact assessment, including:

[0232] Operation status evaluation: Based on the results of graph convolution operations, the feature vector of each node is analyzed to evaluate the operation status of the node, detect abnormal changes in operation parameters such as current, voltage, and load, and identify potential faulty nodes;

[0233] Fault propagation path simulation: After identifying the faulty node, the GCN model is used to simulate the fault propagation path in the power grid. By dynamically updating the adjacency matrix and node features, the possible path of the fault propagating from the initial node to other nodes is predicted.

[0234] Propagation probability calculation: Calculate the probability of each node being affected during the fault propagation process. Assume that the initial fault node is i, and the probability of propagating to node j is P ij , expressed as:

[0235]

[0236] Among them, N(j) is the set of neighbor nodes of node j, A jk is the connection relationship between nodes j and k, H k is the feature vector of node k;

[0237] Fault impact scope assessment: Based on the fault propagation path and propagation probability, the impact scope of the fault is assessed. The affected nodes and edges are determined, and the degree of their impact is calculated.

[0238] Through the above steps, the method realizes real-time monitoring of the power grid operation status and fault warning through the application of the graph convolutional network (GCN) model, improves the timeliness and accuracy of power grid fault handling, and enhances the reliability and safety of the power grid.

[0239] The communication module includes outputting the results of operation status evaluation and fault propagation path simulation through the communication module; generating a detailed fault impact report, including fault nodes, propagation paths, impact range and recommended countermeasures, and generating early warning signals to notify the remote monitoring center and maintenance personnel for timely processing.

[0240] Through the above steps, the method can provide comprehensive and accurate fault information and early warning, ensure that faults can be quickly identified and responded to, and improve the maintenance efficiency and operation safety of the power grid.

[0241] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0242] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A power distribution auxiliary monitoring and early warning system, It is characterized in that It includes data acquisition module, multi-point temperature strain monitoring module, data processing module, power grid topology analysis module and communication module, among which; The data acquisition module collects the operating parameters of each node in the power distribution system in real time, including current and voltage values; The multi-point temperature and strain monitoring module uses sensors arranged at multiple nodes of the power distribution system to collect temperature and strain data in real time and identify complex temperature and strain patterns; The data processing module performs fault diagnosis on the collected operating parameters and temperature and strain data, and generates a warning signal according to the result of the fault diagnosis; The power grid topology analysis module establishes a topology model of the power grid, analyzes the operation status and fault propagation path of the power grid, and provides real-time topology changes and fault impact assessment; The communication module is used to transmit the warning signal to the remote monitoring center and the mobile device of the maintenance personnel.

2. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that The data acquisition module comprises: Sensor deployment: Multiple sensor units are deployed at each node of the power distribution system, and each sensor unit collects the operating parameters of the node in real time; Data preprocessing: preprocess the real-time collected operating parameters, including data denoising, data smoothing and data standardization, including: Data denoising uses filters or noise removal algorithms to remove high-frequency noise and random interference from operating parameters; Data transmission: real-time transmission of pre-processed operating parameters; Through sensor layout, data preprocessing and real-time data transmission, efficient, accurate and real-time monitoring of the operating parameters of each node is ensured.

3. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that The multi-point temperature strain monitoring module comprises: Temperature and strain data collection: Temperature sensors and strain sensors are arranged at multiple nodes of the power distribution system to collect temperature and strain data at the nodes; Preliminary data processing: Preliminary processing of the collected temperature and strain data, including removal of noise and outliers; Temperature and strain data analysis: The temperature and strain data after preliminary processing are analyzed using the local anomaly factor algorithm to identify complex temperature and strain patterns.

4. A distribution auxiliary monitoring and early warning system according to claim 3, It is characterized in that The local anomaly factor algorithm includes: Data preprocessing: standardize the temperature and strain data and form temperature and strain data sets; Select parameter k: The parameter k in the local outlier factor algorithm, i.e. the number of neighbors of each point, is determined by cross-validation. Calculate k-distance and k-distance neighbors: For each data point p in the temperature and strain dataset, find its k-distance k_dist(p) and k-distance neighbor set N k (p), expressed as: k_dist(p)=min{r:|N k (p)|=k}, where N k (p) is the set of k nearest neighbors of point p; Calculate the reachable distance: For each point o in the temperature and strain data set relative to point p, calculate the reachable distance, expressed as: reach_dist k (p,o)=max(k_dist(o),d(p,o)), where d(p,o) is the Euclidean distance between points p and o; Calculate the local reachable density: For each data point p, calculate its local reachable density, expressed as: Calculate the local anomaly factor: For each data point p, calculate its local anomaly factor, expressed as: according to The value of determines whether point p is an outlier; Identify complex patterns and extract key features: Based on the calculated LOF values, identify points with high LOF values ​​as abnormal points, analyze the distribution and patterns of the abnormal points, extract temperature and strain features, including the extraction of points where temperature rises sharply and strain mutation points, perform pattern recognition on the extracted points where temperature rises sharply and strain mutation points, identify frequently occurring abnormal patterns and features, and determine their causes and scope of influence.

5. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that The data processing module comprises: Feature extraction: Utilize power fluctuation detection, use fast Fourier transform to analyze the spectrum of current signal, and extract the current fluctuation feature, which is expressed as: Where X(f) is the spectrum, x[n] is the time domain signal, and N is the number of sampling points. Pattern recognition: The extracted features are classified using the support vector machine algorithm to preliminarily identify potential failure modes, expressed as: Where w is the weight vector, b is the bias, ξ i is the slack variable, C is the penalty parameter, φ(x i ) is the kernel function; Fault type judgment: The identified fault mode is judged using the naive Bayes classifier to determine the specific type of fault, which is expressed as: Among them, P(C k |x) is the given data x belongs to category C k The probability of P(x|C k ) is in Category C k The likelihood of the data x, P(C k ) is Category C k The prior probability of , P(x) is the marginal probability of data x; Fault severity assessment: The severity of the fault is assessed based on the fuzzy logic algorithm to quantify the impact of the fault on the distribution system, expressed as: Where S is the fault severity, w i is the weight of the i-th fuzzy rule, μ i (x) is the i-th fuzzy membership function; Warning signal generation: Based on the results of fault diagnosis, a warning signal is generated. The warning signal includes the fault type, location, severity and recommended maintenance measures, expressed as: Alert=(T fault ,L fault ,S fault, ,M suggest ); Among them, T fault Is the fault type, L fault is the fault location, S fault is the fault severity, M suggest is a recommended maintenance action.

6. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that The power grid topology analysis module uses a graph convolutional network model to model the topological structure of the power grid, including: Data collection: Collect the topological structure data of the distribution system, including the connection relationship between nodes and edges and their attributes; Graph structure construction: The collected topological data is represented as a graph structure G = (V, E), where V is a set of nodes, E is a set of edges, and each node and edge contains its specific attribute information; Adjacency matrix construction: Construct the adjacency matrix A of the graph, where A ij Represents the connection relationship between node i and node j. If there is a connection between node i and node j, then A ij ≠0; otherwise, A ij =0; Node feature matrix construction: Construct the node feature matrix X. The feature vector of each node contains its attribute information, including current, voltage, and load. Graph convolution layer design: Design a graph convolution layer to perform convolution operations on the node features in the graph structure, expressed as: in, is the adjacency matrix with self-loops added, yes The degree matrix, H (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the activation function; Model building and debugging: Use the collected historical running data to train the graph convolutional network model, and use the validation set data to evaluate the trained graph convolutional network model; Real-time topology modeling: Real-time monitoring data is input into the trained graph convolutional network model to update node features and topological structure in real time, providing the current topological status and operating characteristics of the power grid.

7. A power distribution auxiliary monitoring and early warning system according to claim 6, It is characterized in that The model building and debugging includes: Model training: Use the training set data to train the graph convolutional network model and adjust the model parameters through the optimizer Adam; Hyperparameter tuning: Use validation set data to perform hyperparameter tuning and optimize the model’s hyperparameters; Model validation: Use validation set data to evaluate the trained graph convolutional network model to ensure its accuracy in topological structure modeling and calculate performance indicators on the validation set, including accuracy, recall, and F1 score; Accuracy: Among them, TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example; Accuracy: Recall: F1 score: Loss function: For graph convolutional networks, the loss function uses cross entropy loss: Among them, y i is the actual label, is the predicted probability, N is the total number of samples; Model testing: Use the test set data to evaluate the performance of the final model and verify its performance on unseen data; Model deployment and monitoring: Deploy the trained and verified GCN model to the real-time monitoring platform of the power distribution system to analyze the power grid topology and operating status in real time; Continuously monitor the performance of the model, and regularly update and optimize the model.

8. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that Model training through the optimizer Adam includes: Initialize the first-order moment estimate and the second-order moment estimate, expressed as: m t =b 1 m t-1 +(1-β 1 )g t ; Among them, m t and v t are the first-order and second-order moment estimates, g t is the current gradient, β 1 and β 2 is the decay rate; Bias correction, expressed as: Parameter update, expressed as: Among them, θ t is the parameter for the tth iteration, α is the learning rate, and is a small constant to prevent division by zero.

9. A power distribution auxiliary monitoring and early warning system according to claim 1, It is characterized in that Analyze the operation status and fault propagation path of the power grid, provide real-time topology changes and fault impact assessment, including: Operation status evaluation: Based on the results of graph convolution operations, the feature vector of each node is analyzed to evaluate the operation status of the node, detect abnormal changes in current, voltage, and load operation parameters, and identify potential faulty nodes; Fault propagation path simulation: After identifying the faulty node, the graph convolutional network model is used to simulate the fault propagation path in the power grid. By dynamically updating the adjacency matrix and node features, the path of the fault propagation from the initial node to other nodes is predicted. Propagation probability calculation: Calculate the probability of each node being affected during the fault propagation process. Let the initial fault node be i and the probability of propagating to node j be P ij , expressed as: Among them, N(j) is the set of neighbor nodes of node j, A jk is the connection relationship between nodes j and k, H k is the feature vector of node k; Fault impact scope assessment: Based on the fault propagation path and propagation probability, the fault impact scope is assessed, the affected nodes and edges are determined, and the degree of their impact is calculated.

10. A power distribution auxiliary monitoring and early warning system according to claim 4, It is characterized in that The pattern recognition specifically includes: Feature vector construction: The data of each temperature rise point and strain mutation point are expressed as a feature vector x, x = [ΔT i ,Δ i ]; Where, ΔT i is the temperature change at the i-th moment, Δ i is the strain change at the i-th moment; K-means cluster analysis: Use the K-means clustering algorithm to cluster feature vectors and identify frequently occurring abnormal patterns and features, expressed as: Among them, C k is the kth cluster, μ k is the center point of the kth cluster; Abnormal pattern recognition: Analyze each clustering result and identify the frequently occurring abnormal patterns, expressed as: Among them, P(C k |x) is the feature vector x belongs to cluster C k The probability of N k is cluster C k The number of data points; Pattern feature extraction: Extract key features in each cluster and analyze the frequency, amplitude and distribution of temperature rise points and strain mutation points, expressed as: Among them, μ k is cluster C k The mean eigenvector of reflects the average characteristics of the cluster; Cause analysis and impact assessment: By analyzing the distribution of cluster centers and eigenvectors, the causes and impact of abnormal patterns are determined, expressed as: Among them, R k is cluster C k The average impact range of the cluster reflects the breadth of abnormal distribution of the cluster.

Citation Information

Cited By

  • Operation and maintenance management method and system for railway power supply and distribution

    CN119313317A

  • Intelligent monitoring system for sewage treatment pipeline

    CN119511921A

  • Feeder equipment and method of intelligent monitoring power distribution system based on AI algorithm

    CN119602482A

  • Gas detection method and device based on MEMS gas sensor array

    CN119691628A

  • Gas detection method and device based on MEMS gas sensor array

    CN119691628B