Distributed intelligent microgrid fault diagnosis method and system based on edge computing

Abstract

The present invention relates to the field of microgrid fault diagnosis technology, and in particular to a distributed intelligent microgrid fault diagnosis method and system based on edge computing. The present invention proposes the following scheme to construct a regional microgrid model including a spatial topology map, a power generation unit model and a load demand model, and calculate the fault risk assessment matrix of the power generation unit based on this. By deploying an optimization strategy, the edge computing node is configured to collect and process the operating information of the power generation unit in real time, extract fault feature information, and use the fault identification network for diagnosis, and output the fault result. According to the fault diagnosis result, the edge computing node collaboratively calculates and locates the fault point, generates a fault response plan, and dynamically updates the fault risk assessment matrix. The present invention improves the accuracy and real-time performance of fault diagnosis and ensures the stable operation of the microgrid.

Description

Technical Field

[0001] The present invention relates to the technical field of microgrid fault diagnosis, and in particular to a distributed intelligent microgrid fault diagnosis method and system based on edge computing. Background Art

[0002] With the widespread application of distributed energy, microgrids play an increasingly important role in modern power systems. Microgrids provide flexible and reliable power supply by integrating multiple energy types (such as solar energy, wind energy) and multiple loads. However, due to the distributed structure and complexity of microgrids, various faults (such as voltage anomalies, line short circuits, equipment failures, etc.) are prone to occur during operation. Traditional centralized fault diagnosis methods are often difficult to detect and handle these faults in a timely and effective manner, resulting in unstable system operation and even fault diffusion. Therefore, a distributed intelligent microgrid fault diagnosis method and system based on edge computing is proposed to improve the speed, accuracy and responsiveness of fault detection.

[0003] For example, a Chinese patent with publication number CN117269848A discloses a method, device and equipment for fault diagnosis of a microgrid, including obtaining a voltage and current data set of the microgrid, and using a preset wavelet basis function to perform a five-layer decomposition of the voltage and current data set to determine a five-layer high-frequency coefficient set; based on a preset random matrix function, determining a learning sample set and an initial smoothing factor population; using the five-layer high-frequency coefficient set and the learning sample set, performing error iteration on the initial smoothing factor population through a preset neural network model and an error function to determine the optimal smoothing factor; when the number of iterations reaches a preset number, a fault diagnosis is performed based on the five-layer high-frequency coefficients to be detected, a group of learning sample groups in the learning sample set and the optimal smoothing factor through a preset neural network model, and a target fault result is output; the technical problem that the prior art is easily interfered with in actual application, causing the microgrid to be unreliable and untimely when judging the fault type, resulting in low accuracy in diagnosing microgrid faults.

[0004] The above patents all have the problems raised by this background technology: the traditional centralized fault diagnosis system relies on the central control center for data processing, and the data transmission delay and processing time are long, resulting in delayed fault detection. To solve the above problems, this application designs a distributed smart microgrid fault diagnosis method and system based on edge computing. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies of the prior art, and provide a distributed intelligent microgrid fault diagnosis method and system based on edge computing, construct a regional microgrid model including a spatial topology map, a power generation unit model and a load demand model, and calculate the fault risk assessment matrix of the power generation unit based on this. By deploying an optimization strategy, the edge computing nodes are configured to collect and process the working operation information of the power generation unit in real time, extract fault feature information, and use the fault identification network for diagnosis, and output the fault diagnosis results. According to the fault diagnosis results, the edge computing nodes are used to collaboratively calculate and locate the fault point, generate a fault response plan, and dynamically update the fault risk assessment matrix.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A distributed intelligent microgrid fault diagnosis method based on edge computing, the method comprising:

[0008] Constructing a regional microgrid model, the regional microgrid model including a spatial topology map, a power generation unit model and a load demand model, and calculating a fault risk assessment matrix of the power generation unit according to the regional microgrid model;

[0009] Deploy edge computing nodes in the power generation unit by deploying an optimization strategy, wherein the edge computing nodes are configured according to the characteristics of the power generation unit and the fault risk assessment matrix;

[0010] The edge computing node collects the working operation information of the power generation unit in real time, processes the working operation information, and calculates the fault characteristic information;

[0011] Constructing a fault identification network, taking the fault characteristic information as input parameters of the fault identification network, training the input parameters through the fault identification network, and outputting a fault diagnosis result;

[0012] According to the fault diagnosis result, the fault source is traced and the fault point location is calculated through collaborative computing and data sharing of edge computing nodes;

[0013] A fault response plan is generated according to the fault point location, and a fault risk assessment matrix of the power generation unit is updated.

[0014] The calculating of the fault risk assessment matrix of the power generation unit according to the regional microgrid model comprises:

[0015] For each power generation unit model, which includes a photovoltaic power generation unit, a wind power generation unit, and a gas turbine power generation unit, the fault type is identified through knowledge graph technology, and the main risk factors are determined;

[0016] The influence of the main risk factors on the fault type is calculated through a multivariate regression model, and a specific risk calculation model is constructed for the fault type of each power generation unit;

[0017] According to a specific risk calculation model, the spatial topology map and the load demand model are substituted into the corresponding risk calculation model, the risk score is calculated, the risk score is aggregated, and a fault risk assessment matrix is ​​constructed, in which each element of the matrix represents the risk score of power generation unit i under fault type j.

[0018] The deployment optimization strategy includes:

[0019] According to the spatial topology graph, the degree centrality, betweenness centrality and closeness centrality of each power generation unit node in the spatial topology graph are calculated, and the node graph theory weight is calculated according to the degree centrality, betweenness centrality and closeness centrality. The calculation formula of the node graph theory weight is:

[0020] ,

[0021] in, represents the node graph weight of node v, represents the weight of degree centrality, L represents the total load of the area where node v is located, represents the degree of node v, n represents the number of adjacent edges of node v, represents a constant greater than 0, represents the weight of the betweenness centrality, P represents the maximum impact range of the fault propagation of node v, s, t and u represent any node in the area where node v is located, represents the total number of paths from node s to node t, represents the total number of nodes v in the path from node s to node t. represents the weight of proximity centrality, T represents the power transmission efficiency of node v, Represents the length of the shortest path between node u and node v;

[0022] According to the node graph theory weights and the fault risk assessment matrix, cluster analysis is performed on the power generation nodes by using a cluster analysis algorithm, and the power generation node sub-areas are output according to the clustering results;

[0023] Performing a quantitative demand analysis on the power generation node sub-areas to calculate the edge computing node demand of each power generation node sub-area, wherein the constraints of the quantitative demand analysis include the real-time data processing capability, fault tolerance, redundancy, data transmission delay requirements and operation and maintenance costs of the edge computing nodes, and the demand includes the number, location, computing power and communication bandwidth of the edge computing nodes;

[0024] Deploy edge computing nodes in the power generation unit according to the edge computing node requirements.

[0025] The processing of the work operation information includes:

[0026] Dividing the Fourier spectrum of the work operation information into continuous intervals through boundary lines to obtain segmented intervals;

[0027] Calculating the low-frequency component of the segmented interval by a scaling function, and calculating the high-frequency component of the segmented interval by a wavelet function;

[0028] Calculating the detail correlation coefficient of the low-frequency component, calculating the approximate approximation coefficient of the high-frequency component, reconstructing the segmented interval by the detail correlation coefficient and the approximate approximation coefficient, and calculating the optimal solution of the segmented interval under the condition of minimum total bandwidth;

[0029] The optimal solutions of all segmented intervals are energy iterated through information entropy theory, the energy entropy of the working operation information is calculated, and the energy entropy is used as the fault feature information.

[0030] The calculation formula of the optimal solution of the segmentation interval when the total bandwidth is the smallest is:

[0031] ,

[0032] in, It represents the optimal solution of the segmented interval when the total bandwidth is the smallest. Min represents the minimum value function. B represents the bandwidth constraint condition. represents the inverse Fourier transform function, Indicates the current bandwidth value of the split interval. Indicates the maximum bandwidth value of the segmentation interval. Indicates the minimum bandwidth value of the segmentation interval. Indicated in The detail correlation coefficient of the low-frequency component within the interval, Indicated in The approximate coefficients of the high-frequency components in the interval, represents the low-frequency component, Represents high frequency components.

[0033] The fault identification network comprises:

[0034] An input layer, which establishes a feature sequence according to input parameters and calculates the fitting value of the feature sequence;

[0035] A hidden layer converts the fitting value of the feature sequence of the input layer into a high-dimensional space, and calculates the output value of the hidden layer in the high-dimensional space;

[0036] The output layer is used to output the fault diagnosis result.

[0037] The training of the input parameters by the fault identification network includes:

[0038] Initializing the fault identification network according to the number of nodes in the input layer, the number of nodes in the hidden layer and the number of nodes in the output layer of the fault identification network, and determining the connection weights between the input layer and the hidden layer and the node thresholds between the hidden layer and the output layer;

[0039] The input parameters are biased by the connection weights of the input layer and the hidden layer to obtain the output of the hidden layer, and the output of the hidden layer is updated by the node threshold to obtain the output layer input;

[0040] Calculating the error of the output layer input according to the error function of the output layer, determining whether the error of the output layer input meets the convergence requirement of the output layer, and if not, updating the connection weight and the node threshold;

[0041] If the convergence requirement is met, the fault diagnosis result is output.

[0042] The method of tracing the fault source and calculating the fault point location through collaborative computing and data sharing of edge computing nodes includes:

[0043] The edge computing node receives a fault diagnosis result, wherein the fault diagnosis result includes a fault type, a timestamp of the fault occurrence, and a fault impact range;

[0044] According to the fault impact range, the fault diagnosis result is broadcasted within the fault impact range through the point-to-point communication protocol between edge computing nodes to build a node loop communication status table;

[0045] If the communication status between the edge computing node and another edge computing node is normal, the status in the node loop communication status table of the two nodes is set to 1;

[0046] If the communication status between an edge computing node and another edge computing node is in an abnormal state, it is judged as faulty communication, and the status in the node loop communication status table of the two nodes is set to 0;

[0047] Traverse all nodes until all node states in the node loop communication state table are set, extract nodes with a state of 0, and take the propagation paths that overlap or cross between nodes as the fault locations.

[0048] Generating a fault response plan according to the fault point location and updating the fault risk assessment matrix of the power generation unit includes:

[0049] According to the specific location and type of the fault point, the severity of the fault is preliminarily classified into: minor fault, moderate fault and severe fault. Combined with historical fault data and current environmental conditions, the fault impact assessment is carried out to predict the potential impact of the fault on the operation of the microgrid;

[0050] Generate targeted fault response plans based on fault classification and impact assessment;

[0051] Through the collaborative work of edge computing nodes, power data is continuously collected and analyzed to determine the execution of fault response. If the execution effect is not good, the fault response plan is adjusted;

[0052] Archive the data during the execution of the fault response plan, analyze the data, and extract the factors that affect the risk assessment;

[0053] According to the factors, the failure risk factors of the power generation units are updated, and the failure risk value of each power generation unit under the updated environmental conditions is calculated;

[0054] According to the fault risk value, the risk score of the power generation unit under different fault types is calculated, and the fault risk assessment matrix of the power generation unit is updated.

[0055] A distributed intelligent microgrid fault diagnosis system based on edge computing, the system comprising a microgrid modeling and deployment module, a data processing and fault identification module, and a fault response module;

[0056] The microgrid modeling and deployment module is configured with a node deployment strategy, which is used to integrate the geographical information, topological structure, power generation unit model and load demand of the microgrid, build a digital model of the microgrid, and optimize the deployment of edge computing nodes;

[0057] The data processing and fault identification module is configured with a data analysis strategy, which is used to collect the working operation information of the power generation unit, process the data using Fourier transform and wavelet analysis methods, extract fault feature information, and use the trained fault identification network to output the fault diagnosis result according to the input fault feature information;

[0058] The fault response module determines the fault location and type according to the results of the fault identification network, generates and executes a fault response plan, analyzes the data during the fault response process, and updates the fault risk factor and fault risk assessment matrix of the power generation unit.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. The present invention uses edge computing technology to deploy computing resources in a distributed manner at each node of the microgrid to achieve real-time detection and processing of faults, significantly reducing data transmission delays and processing time;

[0061] 2. Through distributed computing and intelligent analysis algorithms, the system of the present invention can process multi-source data, improve the accuracy of fault diagnosis, reduce false alarms and missed alarms, quickly and accurately locate the fault point, and promptly start isolation and repair procedures to ensure that the fault does not spread and minimize the impact on the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0063] Figure 1 This is a flow chart of a distributed intelligent microgrid fault diagnosis method based on edge computing according to Embodiment 1 of the present invention;

[0064] Figure 2 This is a schematic diagram of the microgrid model construction process in Example 1 of the present invention;

[0065] Figure 3 This is a flow chart of the steps of deploying the optimization strategy in Example 1 of the present invention;

[0066] Figure 4 This is a cluster analysis flow chart of Example 1 of the present invention;

[0067] Figure 5 This is a fault identification network structure diagram of Example 1 of the present invention;

[0068] Figure 6 This is a module diagram of a distributed intelligent microgrid fault diagnosis system based on edge computing according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0070] Embodiment 1:

[0071] See also Figure 1 , an embodiment provided by the present invention: a distributed intelligent microgrid fault diagnosis method based on edge computing, the specific steps of the method are as follows:

[0072] S1: Build a regional microgrid model;

[0073] In this step, a digital model of the microgrid is constructed based on the geographical location of the regional microgrid, the type of power generation unit (such as photovoltaic, wind power, gas turbine, etc.), the power load distribution, and the grid topology. The model is used to calculate the fault risk assessment matrix of each power generation unit under different operating conditions. The matrix evaluates the possibility of different types of faults (such as short circuit, open circuit, voltage fluctuation, etc.) in each power generation unit based on multiple factors such as historical operating data, equipment status parameters, environmental factors (such as temperature, humidity, wind speed, etc.) and equipment operating life.

[0074] S2: Configure edge computing nodes according to the deployment optimization strategy;

[0075] In this step, the optimal location and number of edge computing nodes around each power generation unit are determined by optimizing the deployment strategy. On this basis, the computing power, data transmission rate, storage capacity and operation algorithm of the edge computing nodes are configured according to the operating characteristics of the power generation unit and the fault risk assessment matrix. Ensure that the edge computing nodes can quickly process data and provide real-time diagnosis and feedback capabilities when receiving the operating data of the power generation unit.

[0076] S3: The edge computing node collects the working operation information of the power generation unit in real time;

[0077] In this step, the edge computing node obtains the operating parameters of the power generation unit in real time through the sensor component, including voltage, current, power, frequency and temperature.

[0078] S4: Calculate fault characteristic information;

[0079] Based on the preprocessed data, the edge computing node uses a time-frequency analysis algorithm to calculate the fault characteristic information of the power generation unit. This characteristic information includes the abnormal fluctuation pattern of the electrical signal. The extraction of characteristic information helps to identify potential fault types and further improve the accuracy of fault identification.

[0080] S5: Build a fault identification network and train the model;

[0081] Using deep learning technology, a multi-layer neural network or other machine learning model is constructed as a fault identification network. The fault feature information obtained from the edge computing node is input into the fault identification network for network training. During the training process, the model continuously optimizes parameters and weights to improve the diagnostic accuracy and rapid response capabilities for different types of faults. After the model training is completed, the fault identification network can output fault diagnosis results in real time.

[0082] S6: Fault diagnosis and fault location calculation;

[0083] Based on the diagnosis results of the fault identification network, the edge computing nodes perform collaborative computing and data sharing, and combine the grid topology information to perform fault tracing analysis. By calculating the changes in parameters such as current, voltage, and phase, the location and possible causes of the fault are determined. Collaborative computing can effectively reduce the false alarm rate and improve the accuracy of fault location.

[0084] S7: Generate a fault response plan and update the fault risk assessment matrix;

[0085] According to the location of the fault point and the diagnosis results, the corresponding fault response plan is automatically generated, including isolating the fault area, switching the backup power supply, adjusting the load distribution, etc. At the same time, the new fault diagnosis results and response process are fed back to the regional microgrid model to update the fault risk assessment matrix of each power generation unit. This can improve the model's prediction ability and diagnostic accuracy, and provide stronger support for future fault diagnosis and risk management.

[0086] See also Figure 2 , a schematic diagram of the process of constructing a microgrid model according to an embodiment of the present invention, wherein the specific steps of constructing a regional microgrid model are as follows:

[0087] S1.1: Based on the coverage of the microgrid, the area is divided into multiple sub-areas using GIS technology, and each sub-area is further subdivided into micro-units based on the power generation units;

[0088] S1.2: Collect basic data of each micro unit, and construct a micro grid model according to the basic data, wherein the basic data includes geographic location data, power generation unit data, power load data and grid topology data, and the micro grid model includes a spatial topology map, a power generation unit model and a load demand model;

[0089] The geographic location data includes the geographical distribution, meteorological conditions and terrain characteristics of the microgrid, and the specific collection content includes:

[0090] Geographic coordinates (latitude and longitude): The specific location of each microcell, which is used to determine the relative position of the microcell in the entire microgrid;

[0091] Meteorological conditions: such as temperature, humidity, wind speed, wind direction, sunshine intensity, etc. These factors will affect the power generation capacity and efficiency of the power generation unit;

[0092] Topographic features: such as altitude, surrounding obstacles (trees, buildings, etc.), which can affect the performance of wind and photovoltaic power generation units;

[0093] The power generation unit data includes the type, capacity, technical parameters and historical operation data of the power generation unit, wherein the type of the power generation unit includes photovoltaic power generation unit, wind power generation unit and gas turbine power generation unit, and the technical parameters include the output power and efficiency of the power generation unit;

[0094] The power load data includes load demand and load change rules;

[0095] The grid topology data includes the topology structure of the microgrid, line connection information, transformer location and switch station configuration.

[0096] The spatial topology diagram includes:

[0097] Import geographic location data into micro-units, establish basic geographic information models for micro-units, and visualize the location relationships among power generation units, load centers, lines, and switch stations on maps;

[0098] Based on the grid topology data, a topological structure diagram of a micro-unit is drawn to show the connection relationship between power transmission lines and equipment. The topological structure diagram includes node representation and edge representation. Nodes can represent power generation units, load centers, transformers and switch stations. Edges represent power line or cable connections, including information such as voltage level, capacity, type (overhead line or underground cable), etc.;

[0099] Integrate the basic geographic information model and topological structure map to generate a spatial topological map.

[0100] S1.3: Determine the fault type of the power generation unit according to the power generation unit model, and determine the main risk factor according to the fault type;

[0101] Among them, the fault types of photovoltaic power generation units include:

[0102] Partial shading (such as shading by trees or buildings causing partial panel failure) is related to changes in light intensity, panel clarity, and panel output current and voltage;

[0103] Hot spot effect (due to local overheating of PV modules), which is related to the rate of temperature change and abnormal changes in local temperature of the panel;

[0104] Inverter failure (failure of the device that converts DC power to AC power), which is related to voltage fluctuations, output power fluctuations and inverter operating temperature;

[0105] Panel aging (long-term sun exposure causes panel performance degradation), which is related to long-term light accumulation and the degradation rate of panel materials;

[0106] Failure types of wind power generation units include:

[0107] Blade damage (damage to wind turbine blades caused by wind or impact from external objects) is related to wind speed, wind direction changes, blade vibration frequency, and material fatigue of wind turbine blades;

[0108] Generator failure (mechanical component damage or electrical failure), which is related to generator temperature, lubricating oil quality, electrical short circuit frequency and engine vibration frequency;

[0109] Bearing wear (mechanical wear due to long-term rotation), which is related to the fan operation time, bearing lubrication condition, bearing temperature and mechanical load changes;

[0110] Yaw system failure (failure of wind direction control system), which is related to yaw motor power consumption, yaw system response time and yaw gear wear;

[0111] Types of failures in gas turbine power generation units include:

[0112] Combustion chamber failure (temperature abnormality caused by incomplete combustion or unstable combustion), which is related to combustion chamber temperature, gas flow, flame stability and exhaust emissions;

[0113] Blade fracture (blade damage caused by long-term high temperature and high pressure), which is related to blade temperature, stress concentration, number of fatigue cycles and blade material aging;

[0114] Lubrication system failure (lubricant oil quality degradation or leakage) is related to lubricant oil pressure, lubricant oil temperature, lubricant oil contamination and lubricant oil replacement cycle.

[0115] S1.4: construct a risk calculation model according to the fault type and the main risk factor, calculate the risk score of the fault type through the risk calculation model, aggregate the risk scores of all fault types, and obtain a fault risk assessment matrix;

[0116] The specific steps of S1.4 are as follows:

[0117] S1.4.1: Based on the general risk calculation model and the main risk factors, a specific risk calculation model is constructed for each failure type of the power generation unit. The risk calculation model can be expressed as:

[0118] ,

[0119] in, represents the risk score of power generation unit i under fault type j, represents the risk calculation function for fault type j, , and represents the main risk factor of power generation unit i under fault type j;

[0120] The risk calculation function calculates the influence of the main risk factors on the fault type through a multivariate regression model, and uses historical fault data and risk factor data to train the risk calculation function of each fault type through cross-validation and grid search optimization methods;

[0121] S1.4.2: According to a specific risk calculation model, the input data is substituted into the corresponding risk calculation model, the risk score is calculated, the risk score is aggregated, and a fault risk assessment matrix is ​​constructed, where each element of the matrix represents the risk score of power generation unit i under fault type j;

[0122] S1.4.3: Use the latest operating data and failure data to regularly retrain and optimize the risk calculation model to improve the accuracy and real-time adaptability of the model.

[0123] See also Figure 3 , a flow chart of the steps of the deployment optimization strategy of an embodiment of the present invention, the deployment optimization strategy is intended to optimize the deployment strategy of edge computing nodes in a regional microgrid to improve the real-time and accuracy of microgrid fault diagnosis and response. By analyzing the topological structure of the microgrid, the type of power generation unit and its distribution, load demand and fault risk assessment results, the optimal location and number of edge computing nodes around each power generation unit are determined, and the computing power, data transmission rate, storage capacity and operation algorithm of the nodes are configured to ensure that the edge computing nodes can quickly process and provide real-time diagnosis and feedback capabilities when receiving the operation data of the power generation unit. The specific steps are as follows:

[0124] S2.1: According to the regional microgrid model, the power network topology of the microgrid is analyzed, including the geographical location of the power generation unit, the distribution of the power load center, the direction of the trunk power transmission line, the connection relationship between nodes, and the location of the transformer and switch station. Through the topological structure analysis, the degree centrality, betweenness centrality and closeness centrality of each power generation unit node in the spatial topological graph are calculated, and the node graph theory weight is calculated according to the degree centrality, betweenness centrality and closeness centrality. The calculation formula of the node graph theory weight is:

[0125] ,

[0126] in, represents the node graph weight of node v, represents the weight of degree centrality, L represents the total load of the area where node v is located, represents the degree of node v, n represents the number of adjacent edges of node v, represents a constant greater than 0, represents the weight of the betweenness centrality, P represents the maximum impact range of the fault propagation of node v, s, t and u represent any node in the area where node v is located, represents the total number of paths from node s to node t, represents the total number of nodes v in the path from node s to node t. represents the weight of proximity centrality, T represents the power transmission efficiency of node v, Represents the length of the shortest path between node u and node v;

[0127] Degree centrality is used to measure the number of nodes directly connected to a node. In the power network, nodes with high degree centrality usually undertake more power transmission tasks and are key hubs connecting multiple power generation units and load centers. Nodes with high degree centrality represent important power generation units in regional power transmission. Failures of these nodes will directly affect multiple load centers, so it is necessary to focus on deploying edge computing nodes;

[0128] Betweenness centrality is used to measure the intermediary role of a node in the network, that is, the number of shortest paths to other nodes on which the node is located. Nodes with high betweenness centrality are usually key hubs in the power transmission path, and their failures may affect the flow of power between multiple power generation units and load centers. Power generation unit nodes with high betweenness centrality need to be equipped with high-performance edge computing nodes to quickly identify and handle failures and ensure the stability of the power transmission path;

[0129] Closeness centrality is used to measure the average shortest path length between a node and all other nodes in the network. Nodes with high closeness centrality indicate that they can quickly access other nodes. They are usually important hubs for power transmission in the power grid. Edge computing devices equipped with power generation unit nodes with high closeness centrality can ensure rapid response and fault diagnosis on the power transmission path.

[0130] The node graph weights can dynamically reflect the importance of nodes in the network and be optimized based on the topological structure and operation requirements of the actual power network. By calculating the node graph weights, key nodes in the network can be identified more accurately, providing guidance for subsequent edge computing node deployment.

[0131] S2.2: performing cluster analysis on the power generation nodes by using a cluster analysis algorithm according to the node graph theory weights and the fault risk assessment matrix, and outputting the power generation node sub-areas according to the clustering results;

[0132] The power generation unit nodes are partitioned, and the nodes with similar graph weights and fault risk assessment characteristics are divided into sub-areas. This step aims to optimize the deployment strategy of edge computing nodes so that they can efficiently cover multiple power generation units. Through the clustering algorithm, the power generation unit nodes are automatically divided into several sub-areas. The nodes in each sub-area have similar power network locations and fault risk assessment characteristics, thereby reducing costs while ensuring the stability of the power system.

[0133] See also Figure 4 , the cluster analysis flow chart of the embodiment of the present invention, the specific steps of S2.2 are as follows:

[0134] S2.2.1: Calculate the rank of the fault risk assessment matrix, perform preliminary clustering on the rank of the fault risk assessment matrix according to the DBSCAN clustering algorithm, obtain core points, and determine the number of cluster center clusters according to the number of core points;

[0135] Calculate the rank of the fault risk assessment matrix and identify the independence and dimensionality characteristics of the matrix. The rank reflects the amount of independent information in the matrix. Determining the rank helps identify the core elements in the fault risk assessment matrix, which in turn affects the initial selection of cluster centers.

[0136] S2.2.2: The node graph weights of the core points after DBSCAN clustering are used as the initial center clusters of the K-means clustering algorithm, and the Euclidean distances from the node graph weights of other nodes to each initial center cluster are calculated;

[0137] S2.2.3: Perform clustering according to the Euclidean distance, calculate the average clustering error of the initial center cluster, and update the center cluster according to the average clustering error;

[0138] S2.2.4: Define the convergence condition and determine whether the updated center cluster meets the convergence condition. If it does not meet the convergence condition, continue to update the center cluster. If it meets the convergence condition, output the clustering result.

[0139] Specifically, by combining the two clustering algorithms, DBSCAN and K-means, the nodes are clustered and analyzed, and more accurate clustering results are obtained by iteratively optimizing the central cluster, so as to better understand the working status of the nodes, which helps to improve the operating efficiency of the system and reduce energy consumption. DBSCAN is a density clustering algorithm used to identify closely connected data points in space and divide them into clusters. DBSCAN can be used for preliminary clustering to obtain core points. Core points refer to the core area containing at least a specified number of data points within a given radius. Since each core point may represent a cluster, the number of core points can be used to preliminarily determine the number of central clusters of clustering.

[0140] The kernel points obtained by preliminary clustering through DBSCAN are used as the initial center cluster of the K-means clustering algorithm, providing an initial cluster center position for the K-means algorithm; the Euclidean distance is used for clustering, and each node is assigned to the nearest center cluster by calculating the Euclidean distance to each initial center cluster. The clustering average error refers to the average value of the Euclidean distance of all data points to the center. The clustering average error can be used to update the initial center cluster to improve the clustering effect. The position of the center cluster can be adjusted according to the calculated clustering average error, so as to update the center cluster, and the convergence condition is defined to determine whether the updated center cluster meets the condition to determine whether to end the iteration. If the convergence condition is met, the clustering result is output, and the clustering result is used to divide nodes with similar characteristics into sub-regions.

[0141] S2.3: Perform a quantitative demand analysis on the power generation node sub-area, and calculate the edge computing node demand of each power generation node sub-area. The constraints of the quantitative demand analysis include the real-time data processing capability, fault tolerance, redundancy, data transmission delay requirements and operation and maintenance costs of the edge computing nodes. The requirements include the number, location, computing power and communication bandwidth of the edge computing nodes, including:

[0142] Construct a mathematical model based on multi-objective optimization to determine the optimal deployment location and number of edge computing nodes. The optimization objectives include: minimizing data transmission delay, maximizing the number of power generation units covered by computing nodes, balancing the load between nodes, and minimizing node deployment and operation costs;

[0143] By using an intelligent optimization algorithm, the mathematical model of the multi-objective optimization is solved according to the constraints, and the solution result is calculated, wherein the constraints include node computing capacity limitation, network bandwidth limitation, and distance limitation between the node and the power generation unit;

[0144] Based on the solution of the optimization model, determine the number and coverage of edge computing nodes in each sub-area. The number and coverage of nodes should meet the following requirements: be able to cover all key nodes and high-risk areas; ensure the efficiency of real-time data collection and processing; and avoid blind spots and overlapping areas of node coverage.

[0145] S2.4: Deploy edge computing nodes in the power generation unit according to the edge computing node requirements;

[0146] The fault feature information is divided by Fourier spectrum, and the scaling function and wavelet function are combined to calculate the low-frequency and high-frequency components of the segmented interval, and reconstructed by detail correlation coefficient and approximate approximation coefficient. Finally, the energy entropy is calculated and used as the fault feature information. When processing complex working operation information, the fault features can be accurately extracted, and it has high noise resistance and computational efficiency.

[0147] The specific steps of S4 are as follows:

[0148] S4.1: Perform Fourier transform on the work operation information to obtain its spectrum representation. The Fourier spectrum is divided into several continuous intervals through preset boundary lines, and these intervals represent the energy distribution in different frequency ranges;

[0149] The Fourier spectrum is divided by boundary lines to ensure that the signal energy in different frequency bands can be analyzed independently, avoiding information loss or overlap in mixed signals, thereby more effectively extracting features from work operation information.

[0150] S4.2: Use the scaling function to extract the low-frequency components of the segmented intervals. The scaling function is used to extract the low-frequency characteristics of the signal in multi-resolution analysis, which can reflect the global characteristics of the signal. Use the wavelet function to extract the high-frequency components of the segmented intervals. The wavelet function can effectively analyze the local change characteristics of the signal and is suitable for detecting transient changes and mutations in the signal.

[0151] By calculating the low-frequency components, the global information of the signal can be effectively extracted, the influence of high-frequency noise can be reduced, and the stability and accuracy of signal analysis can be improved;

[0152] By calculating the high-frequency components, the sudden changes and anomalies in the working operation information can be detected. In particular, wavelet analysis has unique advantages for abnormal changes with high frequency components in fault detection.

[0153] S4.3: Calculate the detail correlation coefficient of the low-frequency component, calculate the approximate approximation coefficient of the high-frequency component, reconstruct the segmented interval by the detail correlation coefficient and the approximate approximation coefficient, and calculate the optimal solution of the segmented interval when the total bandwidth is minimized;

[0154] The detail correlation coefficient is used to quantify the correlation between low-frequency components and high-frequency components, reflecting the coordination and consistency of each frequency component in the segmentation interval. By comparing the changing trends between low-frequency components and high-frequency components, their correlation is calculated. A high correlation means that there is a strong linear relationship between low-frequency and high-frequency components, which usually indicates that the high-frequency details in the signal are coordinated with the overall trend;

[0155] The approximation coefficient is used to measure the degree of approximation of the high-frequency component to the low-frequency component. The approximation error of the signal is calculated by approximating the high-frequency component in the segmentation interval.

[0156] Use low-frequency components to approximate high-frequency components, compare the approximation values ​​generated by low-frequency components with the actual high-frequency components, calculate the error between each pair of values, summarize the approximation errors of all frequency components, and calculate the average error. This average error is the approximate correlation coefficient.

[0157] S4.4: Based on the information entropy theory, the optimal solution of the reconstructed segmented interval is iterated, and the energy of each segmented interval is gradually accumulated to form the energy distribution of the overall working operation information. By calculating the energy entropy, it is used as an important part of the fault feature information. Energy entropy reflects the distribution complexity and uncertainty of signal energy and is an important indicator to characterize the characteristics of working operation information;

[0158] The introduction of energy entropy enhances the expressiveness of fault feature information, can more accurately identify abnormal changes in working operation information, and improve the reliability of fault detection;

[0159] The calculation formula of the optimal solution of the segmentation interval when the total bandwidth is the smallest is:

[0160] ,

[0161] in, It represents the optimal solution of the segmented interval when the total bandwidth is the smallest. Min represents the minimum value function. B represents the bandwidth constraint condition. represents the inverse Fourier transform function, Indicates the current bandwidth value of the split interval. Indicates the maximum bandwidth value of the segmentation interval. Indicates the minimum bandwidth value of the segmentation interval. Indicated in The detail correlation coefficient of the low-frequency component within the interval, Indicated in The approximate coefficients of the high-frequency components in the interval, represents the low-frequency component, Represents high frequency components.

[0162] The fault identification network achieves efficient diagnosis of complex faults in microgrids through the collaboration of input layer, hidden layer and output layer, specifically including:

[0163] Input layer: responsible for receiving characteristic parameters of microgrid operation;

[0164] Hidden layer: Use activation function to map input data to high-dimensional space to improve data separability. The number of nodes and layers in the hidden layer directly affects the network's expressiveness and training speed;

[0165] Output layer: converts high-dimensional features into diagnostic results, which may include multi-class classification nodes (for identification of multiple fault types) or regression nodes (for quantification of fault severity);

[0166] See also Figure 5 , the fault identification network structure diagram of the embodiment of the present invention, specifically, the fault identification network is provided with an input layer, a hidden layer and an output layer, the input layer is used to establish a feature sequence according to the input parameters of the low-voltage distribution network loss analysis network, accumulate the feature sequence, establish a grey differential equation, calculate the least squares parameter of the grey differential equation, and calculate the fitting value of the feature sequence by discretizing the least squares parameter solution. The hidden layer is used to convert the feature sequence fitting value of the input layer into a high-dimensional space by mapping according to the connection weight between the input layer and the hidden layer, sum the feature sequence fitting value in the high-dimensional space after nonlinear weighting, and calculate the output value of the hidden layer by bias. The output layer is used to obtain the output layer input, calculate the average error of the input value of the output layer, and determine whether the average error meets the convergence condition. If it does not meet the convergence requirement, the connection weight and the node threshold are updated. If the convergence requirement is met, the fault identification result is output;

[0167] First, the number of nodes in each layer is determined according to the actual application scenario of the microgrid. The number of nodes in the input layer is consistent with the dimension of the fault feature information, the number of nodes in the hidden layer is determined by the complexity requirements and computing resources of the network, and the number of nodes in the output layer corresponds to the fault category or value that needs to be classified or regressed;

[0168] During the initialization process, the connection weights between the input layer and the hidden layer, and between the hidden layer and the output layer are initialized by random distribution, usually using uniform distribution or normal distribution. The node threshold is used to adjust the output of the node so that the network can converge more easily during the training process;

[0169] The microgrid operation data received by the input layer is first weighted by the connection weight between the input layer and the hidden layer, and then biased and passed to the hidden layer after the activation function. The output of the hidden layer reflects the representation of the input parameters in the high-dimensional space;

[0170] The output value calculated by the hidden layer is then adjusted through the node threshold. This process can be regarded as a further correction of the output of the hidden layer to ensure that the input of the output layer is closer to the real fault characteristics;

[0171] In the output layer, the high-dimensional features passed from the hidden layer are further processed to obtain the final fault diagnosis result. Then, the error between the diagnosis result and the true label is calculated, and the network is judged whether it has reached the convergence standard according to the error value of the output layer. If the error meets the preset convergence condition, the final fault diagnosis result is output; otherwise, the connection weights and node thresholds are continuously updated until the error meets the convergence requirements. By gradually reducing the error and repeatedly adjusting the weights, the network can eventually achieve the best fault recognition effect on the training data and show good generalization ability in practical applications.

[0172] Specifically, the present invention uses a neural network to perform loss analysis, and provides outputs of different loss types and loss rates of low-voltage distribution networks by training the network. First, the network structure is initialized, the connection weights and node thresholds are determined, and initial parameters can be provided for subsequent training. The input parameters are biased by the connection weights. After the input parameters are processed by the connection weights and biases, they can enter the hidden layer of the neural network to form outputs; then, through error calculation and back propagation algorithms, the network parameters are continuously optimized to improve the fitting ability of the network. It can be determined whether the convergence requirements of the output layer are met. If not, the connection weights and node thresholds can be updated. If the convergence requirements of the output layer are met, the fault identification results of the trained network can be output.

[0173] By detecting the communication status between nodes and analyzing the propagation path, the fault point in the microgrid is accurately located. The specific steps of S6 are as follows:

[0174] S6.1: After receiving the output of the fault identification network, the edge computing node first analyzes the fault diagnosis result. The fault diagnosis result usually includes the fault type, the timestamp of the fault occurrence, and the scope of the fault impact;

[0175] By receiving and processing fault diagnosis results, edge computing nodes can quickly identify basic fault information and provide necessary data support for subsequent fault tracing operations;

[0176] S6.2: Based on the fault impact range, the fault diagnosis results are broadcasted within the fault impact range through the point-to-point communication protocol between edge computing nodes. Each edge computing node that receives the broadcast information will automatically join the fault tracing process and generate a node loop communication status table;

[0177] By broadcasting fault information, all relevant edge computing nodes are ensured to participate in the fault tracing process, thereby improving the coverage and accuracy of fault location.

[0178] S6.3: Each edge computing node generates and updates a local node loop communication status table based on the received fault diagnosis results. This table records the communication status of each node with its neighboring nodes for subsequent fault tracing analysis;

[0179] Each record entry in the table contains the communication status between a node and its adjacent nodes, and the initial status is undetermined. The communication status can be set to normal (1) or abnormal (0). As the fault diagnosis results are propagated, the node loop communication status table can be dynamically updated to ensure the real-time and accuracy of the communication status.

[0180] S6.4: In the node loop communication status table, by regularly sending heartbeat packets or performing handshake signal detection, it is determined whether the communication status between nodes is normal. If the communication status between an edge computing node and another edge computing node is normal, it indicates that the communication link between the two nodes is intact, and the status is set to 1; if the communication status is abnormal, it is determined to be faulty communication, and the status is set to 0;

[0181] S6.5: Traverse the loop communication status table of all edge computing nodes in the network until the communication status of all nodes is set. Extract the nodes with status 0, and identify the overlapping or intersecting propagation paths between these nodes as potential fault locations;

[0182] By traversing the node status table and analyzing the propagation path, the fault point in the network can be accurately located. This method is particularly suitable for complex microgrid systems and can effectively identify the critical path of fault propagation;

[0183] S6.6: By analyzing the propagation paths between nodes with status 0, identify the overlapped or intersecting parts in these paths. Since these parts are the concentrated areas of communication anomalies of multiple nodes, they are usually determined to be high-probability areas of fault locations;

[0184] The analysis of overlapping or crossing paths can accurately identify the areas where faults may occur in the network, significantly reduce the scope and time of fault point location, and improve the efficiency of fault handling;

[0185] S6.7: Combined with the above analysis results, the location of the fault point is finally determined, and a fault location report is output. The report lists in detail the geographical location of the fault point, the scope of impact, the fault type and the recommended repair measures. By generating a fault location report, a basis for fault handling is provided, which significantly shortens the response time of fault handling and improves the reliability and stability of the microgrid system.

[0186] The specific steps of S7 are as follows:

[0187] S7.1: Based on the specific location and type of the fault point, make a preliminary classification of the severity of the fault, including: minor fault, moderate fault and severe fault. Combine historical fault data and current environmental conditions to conduct fault impact assessment and predict the potential impact of the fault on the operation of the microgrid;

[0188] S7.2: Generate targeted fault response plans based on fault classification and impact assessment. The response plans include:

[0189] Fault isolation: Isolate the fault point electrically to prevent the fault from spreading to other areas. Isolation measures include cutting off the corresponding power lines or disconnecting the relevant power generation units.

[0190] Load redistribution: Redistribute the power load in the microgrid to ensure that important load centers continue to receive power supply. Dispatch other power generation units to increase output to make up for the power supply gap caused by the fault point.

[0191] Backup power supply activation: Start the backup power supply system (such as diesel generators, battery energy storage systems, etc.) to provide temporary power supply for critical loads and ensure continuous power supply for important loads;

[0192] S7.3: Monitor the operation status of the power grid in real time based on the execution of fault response. Through the collaborative work of edge computing nodes, continuously collect and analyze power data, adjust fault response strategies in time, and ensure the stable operation of the entire microgrid;

[0193] S7.4: Archive all data during the execution of the fault response plan, including the time of fault occurrence, fault type, fault point location, fault handling steps and effects. Perform a detailed analysis of these data to extract key factors affecting risk assessment, such as fault frequency, load characteristics of the fault point, and historical environmental conditions;

[0194] S7.5: Re-evaluate and update the failure risk factors of the power generation units based on the archived data. Calculate the failure risk value of each power generation unit under the updated environmental conditions. For power generation units with historical failure records, increase the weight of their failure risk factors; for units that have not failed, make adjustments based on the risk conditions of their adjacent nodes;

[0195] S7.6: Use the updated failure risk factors to retrain or adjust the original multivariate regression model. The new model takes into account the latest failure data and its impact, recalculates the failure risk of the power generation unit, calculates the risk score of the power generation unit under different failure types, and updates the failure risk assessment matrix of the power generation unit.

[0196] Embodiment 2:

[0197] See also Figure 6, the present invention provides an embodiment: a distributed intelligent microgrid fault diagnosis system based on edge computing, the system comprising a microgrid modeling and deployment module, a data processing and fault identification module and a fault response module;

[0198] The microgrid modeling and deployment module is configured with a node deployment strategy, which is used to integrate the geographical information, topological structure, power generation unit model and load demand of the microgrid, build a digital model of the microgrid, and optimize the deployment of edge computing nodes;

[0199] The data processing and fault identification module is configured with a data analysis strategy, which is used to collect the working operation information of the power generation unit, process the data using Fourier transform and wavelet analysis methods, extract fault feature information, and use the trained fault identification network to output the fault diagnosis result according to the input fault feature information;

[0200] The fault response module determines the fault location and type according to the results of the fault identification network, generates and executes the fault response plan, analyzes the data in the fault response process, and updates the fault risk factor and fault risk assessment matrix of the power generation unit;

[0201] The microgrid modeling and deployment module includes:

[0202] A modeling unit, which is used to integrate the geographical information, topological structure, power generation unit model and load demand of the microgrid to build a digital model of the microgrid;

[0203] A deployment unit is used to calculate the fault risk assessment matrix of the power generation unit and deploy edge computing nodes in the power generation unit through a deployment optimization strategy;

[0204] The node deployment strategy includes a modeling logic and a deployment logic, the modeling logic is configured in the modeling unit, and the deployment logic is configured in the deployment unit;

[0205] The data processing and fault identification module comprises:

[0206] A data processing unit, used for the edge computing node to collect the working operation information of the power generation unit in real time, process the working operation information, and calculate the fault characteristic information;

[0207] A fault identification unit, used to construct a fault identification network, use the fault characteristic information as an input parameter of the fault identification network, train the input parameter through the fault identification network, and output a fault diagnosis result;

[0208] The fault response module comprises:

[0209] A fault location unit, configured to perform fault tracing and calculate the location of the fault point through collaborative computing and data sharing of edge computing nodes according to the fault diagnosis result;

[0210] A scheme generating unit generates a fault response scheme according to the fault point location and updates the fault risk assessment matrix of the power generation unit.

[0211] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A distributed intelligent microgrid fault diagnosis method based on edge computing, characterized in that: The method comprises: Constructing a regional microgrid model, the regional microgrid model including a spatial topology map, a power generation unit model and a load demand model, and calculating a fault risk assessment matrix of the power generation unit according to the regional microgrid model; Deploy edge computing nodes in the power generation unit by deploying an optimization strategy, wherein the edge computing nodes are configured according to the characteristics of the power generation unit and the fault risk assessment matrix; The edge computing node collects the working operation information of the power generation unit in real time, processes the working operation information, and calculates the fault characteristic information; Constructing a fault identification network, taking the fault characteristic information as input parameters of the fault identification network, training the input parameters through the fault identification network, and outputting a fault diagnosis result; According to the fault diagnosis result, the fault source is traced and the fault point location is calculated through collaborative computing and data sharing of edge computing nodes; Generate a fault response plan according to the fault point location, and update the fault risk assessment matrix of the power generation unit; The calculating of the fault risk assessment matrix of the power generation unit according to the regional microgrid model comprises: For each power generation unit model, which includes a photovoltaic power generation unit, a wind power generation unit, and a gas turbine power generation unit, the fault type is identified through knowledge graph technology, and the main risk factors are determined; The influence of the main risk factors on the fault type is calculated through a multivariate regression model, and a specific risk calculation model is constructed for the fault type of each power generation unit; According to a specific risk calculation model, the spatial topology map and the load demand model are substituted into the corresponding risk calculation model, the risk score is calculated, the risk score is aggregated, and a fault risk assessment matrix is ​​constructed, where each element of the matrix represents the risk score of power generation unit i under fault type j; The deployment optimization strategy includes: According to the spatial topology graph, the degree centrality, betweenness centrality and closeness centrality of each power generation unit node in the spatial topology graph are calculated, and the node graph theory weight is calculated according to the degree centrality, betweenness centrality and closeness centrality. The calculation formula of the node graph theory weight is: ， in, represents the node graph weight of node v, represents the weight of degree centrality, L represents the total load of the area where node v is located, represents the degree of node v, n represents the number of adjacent edges of node v, represents a constant greater than 0, represents the weight of the betweenness centrality, P represents the maximum impact range of the fault propagation of node v, s, t and u represent any node in the area where node v is located, represents the total number of paths from node s to node t, represents the total number of nodes v in the path from node s to node t. represents the weight of proximity centrality, T represents the power transmission efficiency of node v, Represents the length of the shortest path between node u and node v; According to the node graph theory weights and the fault risk assessment matrix, cluster analysis is performed on the power generation nodes by using a cluster analysis algorithm, and the power generation node sub-areas are output according to the clustering results; Performing a quantitative demand analysis on the power generation node sub-areas to calculate the edge computing node demand of each power generation node sub-area, wherein the constraints of the quantitative demand analysis include the real-time data processing capability, fault tolerance, redundancy, data transmission delay requirements and operation and maintenance costs of the edge computing nodes, and the demand includes the number, location, computing power and communication bandwidth of the edge computing nodes; Deploy edge computing nodes in the power generation unit according to the edge computing node requirements.

2. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 1 is characterized in that: The processing of the work operation information includes: Dividing the Fourier spectrum of the work operation information into continuous intervals through boundary lines to obtain segmented intervals; Calculating the low-frequency component of the segmented interval by a scaling function, and calculating the high-frequency component of the segmented interval by a wavelet function; Calculating the detail correlation coefficient of the low-frequency component, calculating the approximate approximation coefficient of the high-frequency component, reconstructing the segmented interval by the detail correlation coefficient and the approximate approximation coefficient, and calculating the optimal solution of the segmented interval under the condition of minimum total bandwidth; The optimal solutions of all segmented intervals are energy iterated through information entropy theory, the energy entropy of the working operation information is calculated, and the energy entropy is used as the fault feature information.

3. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 2 is characterized in that: The calculation formula of the optimal solution of the segmentation interval when the total bandwidth is the smallest is: ， in, It represents the optimal solution of the segmented interval when the total bandwidth is the smallest. Min represents the minimum value function. B represents the bandwidth constraint condition. represents the inverse Fourier transform function, Indicates the current bandwidth value of the split interval. Indicates the maximum bandwidth value of the segmentation interval. Indicates the minimum bandwidth value of the segmentation interval. Indicated in The detail correlation coefficient of the low-frequency component within the interval, Indicated in The approximate coefficients of the high-frequency components in the interval, represents the low-frequency component, Represents high frequency components.

4. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 1 is characterized in that: The fault identification network comprises: An input layer, which establishes a feature sequence according to input parameters and calculates the fitting value of the feature sequence; A hidden layer converts the fitting value of the feature sequence of the input layer into a high-dimensional space, and calculates the output value of the hidden layer in the high-dimensional space; The output layer is used to output the fault diagnosis result.

5. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 4 is characterized in that: The training of the input parameters by the fault identification network includes: Initializing the fault identification network according to the number of nodes in the input layer, the number of nodes in the hidden layer and the number of nodes in the output layer of the fault identification network, and determining the connection weights between the input layer and the hidden layer and the node thresholds between the hidden layer and the output layer; The input parameters are biased by the connection weights of the input layer and the hidden layer to obtain the output of the hidden layer, and the output of the hidden layer is updated by the node threshold to obtain the output layer input; Calculating the error of the output layer input according to the error function of the output layer, determining whether the error of the output layer input meets the convergence requirement of the output layer, and if not, updating the connection weight and the node threshold; If the convergence requirement is met, the fault diagnosis result is output.

6. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 1 is characterized in that: The method of tracing the fault source and calculating the fault point location through collaborative computing and data sharing of edge computing nodes includes: The edge computing node receives a fault diagnosis result, wherein the fault diagnosis result includes a fault type, a timestamp of the fault occurrence, and a fault impact range; According to the fault impact range, the fault diagnosis result is broadcasted within the fault impact range through the point-to-point communication protocol between edge computing nodes to build a node loop communication status table; If the communication status between the edge computing node and another edge computing node is normal, the status in the node loop communication status table of the two nodes is set to 1; If the communication status between an edge computing node and another edge computing node is in an abnormal state, it is judged as faulty communication, and the status in the node loop communication status table of the two nodes is set to 0; Traverse all nodes until all node states in the node loop communication state table are set, extract nodes with a state of 0, and take the propagation paths that overlap or cross between nodes as the fault locations.

7. The distributed intelligent microgrid fault diagnosis method based on edge computing according to claim 1 is characterized in that: Generating a fault response plan according to the fault point location and updating the fault risk assessment matrix of the power generation unit includes: According to the specific location and type of the fault point, the severity of the fault is preliminarily classified into: minor fault, moderate fault and severe fault. Combined with historical fault data and current environmental conditions, the fault impact assessment is carried out to predict the potential impact of the fault on the operation of the microgrid; Generate targeted fault response plans based on fault classification and impact assessment; Through the collaborative work of edge computing nodes, power data is continuously collected and analyzed to determine the execution of fault response. If the execution effect is not good, the fault response plan is adjusted; Archive the data during the execution of the fault response plan, analyze the data, and extract the factors that affect the risk assessment; According to the factors, the failure risk factors of the power generation units are updated, and the failure risk value of each power generation unit under the updated environmental conditions is calculated; According to the fault risk value, the risk score of the power generation unit under different fault types is calculated, and the fault risk assessment matrix of the power generation unit is updated.

8. A distributed intelligent microgrid fault diagnosis system based on edge computing, implemented based on the distributed intelligent microgrid fault diagnosis method based on edge computing as claimed in any one of claims 1 to 7, characterized in that: The system includes a microgrid modeling and deployment module, a data processing and fault identification module, and a fault response module; The microgrid modeling and deployment module is configured with a node deployment strategy, which is used to integrate the geographical information, topological structure, power generation unit model and load demand of the microgrid, build a digital model of the microgrid, and optimize the deployment of edge computing nodes; The data processing and fault identification module is configured with a data analysis strategy, which is used to collect the working operation information of the power generation unit, process the data using Fourier transform and wavelet analysis methods, extract fault feature information, and use the trained fault identification network to output the fault diagnosis result according to the input fault feature information; The fault response module determines the fault location and type according to the results of the fault identification network, generates and executes a fault response plan, analyzes the data during the fault response process, and updates the fault risk factor and fault risk assessment matrix of the power generation unit.

Images