Method, system and storage medium for power grid-communication network coupled node state assessment

Abstract

The application discloses a power grid-communication network coupling node state evaluation method and system, and a storage medium. The topological relationship between the nodes in the network is considered, an internal weighted correlation matrix is constructed, and an improved PageRank algorithm is combined to realize the calculation of the topological importance of the power and communication nodes, thereby providing an effective basis for subsequent node state evaluation. The coupling relationship between the power grid and the communication network is considered, a cross-network coupling matrix is constructed, and the matrix elements are iteratively updated in combination with the node state of the two networks, thereby realizing the capture of the cross-network coupling relationship and making up for the vacancy of the traditional method in the coupling evaluation of the node state. The multi-dimensional quantization and the graph convolutional neural network method are combined to calculate the node state score of the power grid and the communication network, and finally the weak nodes of the network are screened out. The multi-dimensional features can be comprehensively used to realize the network node state evaluation, and the accuracy and reliability of the node state evaluation are improved.

Description

Technical Field

[0001] This invention belongs to the field of power grid-communication network collaborative analysis technology. It relates to a method, system and storage medium for evaluating the node status of power grid-communication network coupling. Background Technology

[0002] Power grids and communication networks are core and critical infrastructures supporting the operation of modern society, and their safe and stable operation plays an irreplaceable role in national economic development and the maintenance of social order. As the carriers of these two networks, power grid nodes (including distribution transformers, automated switches, and other equipment) and communication network nodes (including core routers, base stations, and other equipment) respectively undertake the core functions of power transmission and distribution and information exchange and transmission. Currently, the perception of the operational status of these nodes mainly relies on the collection and transmission of data from various monitoring systems. Accurate and real-time monitoring of the operational status of these two types of nodes is a prerequisite for achieving network scheduling optimization, rapid fault response, and efficient resource allocation.

[0003] With the rapid development of smart grids and next-generation communication technologies, the coupling relationship between power grids and communication networks is increasingly strengthened, gradually forming a two-way interdependence pattern of "power supporting communication and communication empowering power": the power system relies on communication networks to achieve remote control, protection, and data acquisition, while the operation of communication equipment depends on a stable power supply. However, current methods for assessing node status are still limited to the internal workings of a single network. For example, power grids assess node status based on electrical measurement data through state estimation and power flow calculation, while communication networks rely on performance index weighting and fault tree analysis to quantify node health. These methods generally ignore the deep coupling characteristics between power and communication nodes at the physical and functional levels, making it difficult for assessment results to reflect the true status of nodes in cross-network collaborative scenarios. This fails to provide effective support for the collaborative operation and maintenance and risk warning of power-communication coupled systems. Summary of the Invention

[0004] To address the complex coupling effects of power grids and communication networks, traditional methods struggle to comprehensively consider the multi-dimensional characteristics of both networks for network node status assessment. This invention aims to provide a method, system, and storage medium for assessing the state of nodes in a power grid-communication network coupling scenario. By considering the coupling effects of the power grid and communication network and comprehensively considering multi-dimensional characteristics, the accuracy and reliability of node status assessment can be improved.

[0005] The technical solution to achieve the purpose of this invention is as follows:

[0006] A method for evaluating the state of nodes in a power grid-communication network coupling system includes the following steps:

[0007] Construct weighted correlation matrices within the power grid and within the communication network to quantify the strength of topological correlations among nodes in the same network;

[0008] Based on the weighted correlation matrix and the real-time status of the nodes, the topological importance scores of power nodes and communication nodes are calculated respectively.

[0009] A cross-network coupling matrix is ​​constructed based on business characteristics and topological importance, and the elements of the cross-network coupling matrix are dynamically updated in combination with the states of power nodes and communication nodes. The cross-network coupling matrix and the states of power and communication nodes form a bidirectional iterative update relationship, so that the cross-network coupling matrix and the states of power and communication nodes remain dynamically adapted.

[0010] The communication node status score is calculated by weighted summation based on the communication node's own performance score, intra-network topology impact score, cross-network power feedback score, and topology importance score.

[0011] Using the weighted correlation matrix within the power grid as a graph structure, an input vector is constructed by fusing the communication node status, topological importance score, and power node features. The output power node status score is then optimized through unsupervised iterative optimization using a graph convolutional neural network.

[0012] In the preferred technical solution, constructing the weighted correlation matrix within the power grid and the weighted correlation matrix within the communication network includes:

[0013] Weighted correlation matrix within the power grid :

[0014]

[0015] in, Y [ p , q [Power node] p and q Inter-normalized electrical coupling degree, Γ p for p The set of adjacent power nodes, Z [ p , q ]for p and q mutual impedance, Z [ p , p ]、 Z [ q , q ]for p and q Self-impedance;

[0016] Weighted correlation matrix within the communication network :

[0017]

[0018] in, B[ c , k [Communication node] c and k Inter-link bandwidth, Ω c for c The set of adjacent communication nodes.

[0019] In the preferred technical solution, the calculation of topology importance scores for power nodes and communication nodes includes:

[0020] An improved PageRank algorithm iteratively calculates the topological importance score of power nodes:

[0021]

[0022] in, t For the number of iterations, I P [ p ]、 I P [ q [These are power nodes] p、q Topological importance score, d P [ p [ ] represents the dynamic damping coefficient. U p This is a normalized value for voltage stability; the more stable the voltage, the better. d P [ p The larger the [], n This represents the number of power nodes.

[0023] initial value I P (0) [ p ]=1 / n Iterate to max| I P (t+1) [ p ]- I P (t) [ p If the convergence is less than a given threshold, then the power node is obtained. p Topological importance score I P [ p ];

[0024] The topological importance score of communication nodes is calculated iteratively using an improved Pagerank algorithm.

[0025]

[0026] in, I C [ c ]、 I C [ k [These are communication nodes] c、k Topological importance score, d C [ c [ ] represents the dynamic damping coefficient. , L c The packet loss rate is the highest possible packet loss rate. d C [ c The larger the [], m Number of communication nodes;

[0027] initial value I C (0) [c]=1 / m Iterate to max| I C (t+1) [c]- I C (t) [c]|Convergence occurs when the value is less than a given threshold, thus obtaining the communication node. c Topological importance score I C [ c ].

[0028] In the preferred technical solution, a cross-network coupling matrix is ​​constructed based on business characteristics and topological importance, and the elements of the cross-network coupling matrix are dynamically updated in conjunction with the states of power nodes and communication nodes, including:

[0029] Constructing a cross-network coupling matrix ;

[0030] Calculate the initial cross-network coupling matrix:

[0031]

[0032] in, e ( c ) is the direct power supply node for obtaining communication node c; R[ p , c [Power node] p Through communication nodes c The amount of data transmitted accounts for p The proportion of total transmitted power data; I P [ p ]、 I C[ c These are the topological importance scores for power nodes and communication nodes, respectively. K This is a normalization constant;

[0033] Dynamically updated formula:

[0034]

[0035] Where: σ is the normalized activation function. Q c ( t ) is a communication node c No. t Round of scoring, Q ref This serves as a reference value for the health status of communication nodes. S p ( t ) is a power node p No. t Round of scoring, S ref This serves as a reference value for the health status of power nodes.

[0036] In the preferred technical solution, the communication node c Status score Q c for:

[0037]

[0038] Among them, weight α 1+ α 2 + α 3+ α 4=1, I C [ c [This represents the topological importance score of the communication nodes;]

[0039] Self-performance score Q self,c :

[0040]

[0041] in, D c For the service latency of node c, D th For business latency threshold, L c For packet loss rate, T c Measuring temperature for communication equipment, T nom , Tmax These are the rated temperature and maximum temperature of the communication equipment, respectively.

[0042] Network topology affects score Q com,c :

[0043]

[0044] Cross-grid power feedback score Q elec,c :

[0045]

[0046] Among them, Ψ c To communicate with nodes c The set of power nodes that transmit data For power nodes p pass c Measurement distortion of transmitted data L p,c for c transmission p Data packet loss rate D p,c for c transmission p Data latency, D req,p for p Business latency requirements.

[0047] In the preferred technical solution, the unsupervised iterative optimization of the output power node state score using a graph convolutional neural network includes:

[0048] Construct the input feature vector x p :

[0049]

[0050] in, , For normalized electrical quantities, U 0 represents the power node. p Rated voltage, U p For power nodes p Measure voltage, P max For power nodes p Rated maximum power P p For power nodes p Measure power; Weighted values ​​for the quality of associated communication nodes. H p For power transmission nodesp Data communication node set;

[0051] The power node state is calculated using a graph convolutional neural network, including the first layer of in-network feature aggregation:

[0052]

[0053] Where ReLU is the corrected linear unit. W (1) This is the first layer weight matrix. b (1) For the first layer bias;

[0054] Second layer output score:

[0055]

[0056] The sigmoid function is the activation function. W (2) This is the weight matrix for the second layer. b (2) For the second layer bias;

[0057] Unsupervised iterative optimization of the objective function:

[0058]

[0059] Wherein, parameter λ1+λ2+λ3=1;

[0060] Topological smoothness constraints L topo :

[0061]

[0062] in, , Power nodes p、q Status score;

[0063] Physical property constraints L phys :

[0064]

[0065] in, P p,q For directional power flow. For nodes p The load power, For inflow node p Total power;

[0066] Cross-network coupling consistency constraints L cross:

[0067]

[0068] R ( i ) is a regular expression term. or The regularization coefficient is... i The complexity of the model parameters;

[0069] Iterative process: The Adam optimizer is used to update the parameters until the loss converges, and the final power node state score is output. S p .

[0070] In the preferred technical solution, after obtaining the node status score, the method further includes screening weak nodes based on the status scores and topological importance of communication nodes and power nodes, locating weak links in cross-network connections by combining the cross-network coupling matrix, and outputting the causes and relationships of the weak points.

[0071] In the preferred technical solution, the method of locating weak links in cross-network connections by combining the cross-network coupling matrix includes:

[0072] Nodes with a status score below the health status threshold and a topological importance above the importance threshold are identified as critical weak nodes. The remaining nodes below the health status threshold are identified as general weak nodes, and all are sorted in descending order of their scores.

[0073] Based on the cross-network coupling matrix, if the coupling strength is higher than the tight coupling threshold and the corresponding power node and communication node are both key weak nodes, it is determined to be a core weak link in the cross-network; if the coupling strength is higher than the tight coupling threshold and at least one of them is a general weak node, it is determined to be a common weak link in the cross-network.

[0074] This invention also discloses a node state assessment system coupled to a power grid and a communication network, used to implement the above-mentioned node state assessment method coupled to a power grid and a communication network, comprising:

[0075] The network structure modeling module constructs weighted correlation matrices within the power grid and within the communication network to quantify the strength of topological associations between nodes in the same network.

[0076] The topology importance assessment module calculates the topology importance scores for power nodes and communication nodes based on the weighted correlation matrix and the real-time status of the nodes.

[0077] The cross-network coupling matrix construction module constructs a cross-network coupling matrix based on business characteristics and topological importance, and dynamically updates the elements of the cross-network coupling matrix in combination with the status of power nodes and communication nodes. The cross-network coupling matrix and the status of power and communication nodes form a bidirectional iterative update relationship, so that the cross-network coupling matrix and the status of power and communication nodes remain dynamically adapted.

[0078] The communication node status quantification and evaluation module calculates the communication node status score by weighted summation based on the communication node's own performance score, intra-network topology impact score, cross-network power feedback score, and topology importance score.

[0079] The power node state deep evaluation module uses the weighted correlation matrix within the power grid as a graph structure. It integrates the communication node state, topological importance score, and power node features to construct an input vector, and outputs the power node state score through unsupervised iterative optimization using a graph convolutional neural network.

[0080] The present invention also discloses a computer storage medium storing a computer program thereon, wherein when the computer executes the computer program, it implements the node state assessment method for power grid-communication network coupling described in any of the above claims.

[0081] The present invention also discloses an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program stored in the memory. When the computer program is executed, it implements the node state assessment method for power grid-communication network coupling described in any of the above claims.

[0082] Compared with the prior art, the significant advantages of this invention are:

[0083] (1) This invention considers the topological relationship between nodes within the network. By constructing an internal weighted correlation matrix and combining it with an improved PageRank algorithm, it realizes the calculation of the topological importance of power and communication nodes, providing a valid basis for subsequent node status assessment.

[0084] (2) This invention considers the coupling relationship between the power grid and the communication network. By constructing a cross-network coupling matrix and combining the state of the nodes of the two networks to iteratively update the matrix elements, the cross-network coupling relationship is captured, which makes up for the gap in the traditional method in evaluating the state of the nodes in the coupling.

[0085] (3) This invention combines multidimensional quantization and graph convolutional neural network methods to calculate the state scores of power grid and communication network nodes, and finally screens out weak nodes in the network. It can assist professionals in carrying out various business analyses and decisions, and has good application prospects. Attached Figure Description

[0086] Figure 1 This is a flowchart of the node state assessment method for power grid-communication network coupling in this embodiment;

[0087] Figure 2 This is a complete flowchart of the node status assessment system coupled with the power grid and communication network in this embodiment. Detailed Implementation

[0088] The principle of this invention is as follows: Considering the topological relationships between nodes within a network, this invention constructs an internal weighted association matrix and combines it with an improved PageRank algorithm to calculate the topological importance of power and communication nodes, providing a valid basis for subsequent node state assessment. Considering the coupling relationship between the power grid and the communication network, this invention constructs a cross-network coupling matrix and iteratively updates the matrix elements based on the node states of both networks, thus capturing the cross-network coupling relationship and filling the gap in traditional methods for evaluating node states through coupling. Finally, multidimensional quantization and graph convolutional neural network methods are combined to calculate the node state scores of the power grid and the communication network, ultimately identifying weak nodes in the network.

[0089] Example:

[0090] like Figure 1 As shown, a node state assessment method for power grid-communication network coupling includes the following steps:

[0091] Construct weighted correlation matrices within the power grid and within the communication network to quantify the strength of topological correlations among nodes in the same network;

[0092] Based on the weighted correlation matrix and the real-time status of the nodes, the topological importance scores of power nodes and communication nodes are calculated respectively.

[0093] A cross-network coupling matrix is ​​constructed based on business characteristics and topological importance, and the elements of the cross-network coupling matrix are dynamically updated in combination with the states of power nodes and communication nodes. The cross-network coupling matrix and the states of power and communication nodes form a bidirectional iterative update relationship, so that the cross-network coupling matrix and the states of power and communication nodes remain dynamically adapted.

[0094] The communication node status score is calculated by weighted summation based on the communication node's own performance score, intra-network topology impact score, cross-network power feedback score, and topology importance score.

[0095] Using the weighted correlation matrix within the power grid as a graph structure, an input vector is constructed by fusing the communication node status, topological importance score, and power node features. The output power node status score is then optimized through unsupervised iterative optimization using a graph convolutional neural network.

[0096] In a preferred embodiment, constructing the weighted correlation matrix within the power grid and the weighted correlation matrix within the communication network includes:

[0097] Weighted correlation matrix within the power grid :

[0098]

[0099] in, Y [ p , q [Power node] p andq Inter-normalized electrical coupling degree, Γ p for p The set of adjacent power nodes, Z [ p , q ]for p and q mutual impedance, Z [ p , p ]、 Z [ q , q ]for p and q Self-impedance;

[0100] Weighted correlation matrix within the communication network :

[0101]

[0102] in, B [ c , k [Communication node] c and k Inter-link bandwidth, Ω c for c The set of adjacent communication nodes.

[0103] In a preferred embodiment, the calculation of topology importance scores for power nodes and communication nodes includes:

[0104] An improved PageRank algorithm iteratively calculates the topological importance score of power nodes:

[0105]

[0106] in, t For the number of iterations, I P [ p ]、 I P [ q [These are power nodes] p、q Topological importance score, d P [ p [ ] represents the dynamic damping coefficient. U p This is a normalized value for voltage stability; the more stable the voltage, the better. d P [ p The larger the [], n This represents the number of power nodes.

[0107] initial valueI P (0) [ p ]=1 / n Iterate to max| I P (t+1) [ p ]- I P (t) [ p If the convergence is less than a given threshold, then the power node is obtained. p Topological importance score I P [ p ];

[0108] The topological importance score of communication nodes is calculated iteratively using an improved Pagerank algorithm.

[0109]

[0110] in, I C [ c ]、 I C [ k [These are communication nodes] c、k Topological importance score, d C [ c [ ] represents the dynamic damping coefficient. , L c The packet loss rate is the highest possible packet loss rate. d C [ c The larger the [], m Number of communication nodes;

[0111] initial value I C (0) [c]=1 / m Iterate to max| I C (t+1) [c]- I C (t) [c]|Convergence occurs when the value is less than a given threshold, thus obtaining the communication node. c Topological importance score I C [ c ].

[0112] A preferred embodiment involves constructing a cross-network coupling matrix based on service characteristics and topological importance, and dynamically updating the elements of the cross-network coupling matrix by combining the states of power nodes and communication nodes, including:

[0113] Constructing a cross-network coupling matrix ;

[0114] Calculate the initial cross-network coupling matrix:

[0115]

[0116] in, e ( c ) is the direct power supply node for obtaining communication node c; R[ p , c [Power node] p Through communication nodes c The amount of data transmitted accounts for p The proportion of total transmitted power data; I P [ p ]、 I C [ c These are the topological importance scores for power nodes and communication nodes, respectively. K This is a normalization constant;

[0117] Dynamically updated formula:

[0118]

[0119] Where: σ is the normalized activation function. Q c ( t ) is a communication node c No. t Round of scoring, Q ref This serves as a reference value for the health status of communication nodes. S p ( t ) is a power node p No. t Round of scoring, S ref This serves as a reference value for the health status of power nodes.

[0120] In a preferred embodiment, the communication node c Status score Q c for:

[0121]

[0122] Among them, weight α1+ α 2 + α 3+ α 4=1, I C [ c [This represents the topological importance score of the communication nodes;]

[0123] Self-performance score Q self,c :

[0124]

[0125] in, D c For the service latency of node c, D th For business latency threshold, L c For packet loss rate, T c Measuring temperature for communication equipment, T nom , T max These are the rated temperature and maximum temperature of the communication equipment, respectively.

[0126] Network topology affects score Q com,c :

[0127]

[0128] Cross-grid power feedback score Q elec,c :

[0129]

[0130] Among them, Ψ c To communicate with nodes c The set of power nodes that transmit data For power nodes p pass c Measurement distortion of transmitted data L p,c for c transmission p Data packet loss rate D p,c for c transmission p Data latency, D req,p for p Business latency requirements.

[0131] In a preferred embodiment, the unsupervised iterative optimization of the output power node state score using a graph convolutional neural network includes:

[0132] Construct the input feature vector x p :

[0133]

[0134] in, , For normalized electrical quantities, U 0 represents the power node. p Rated voltage, U p For power nodes p Measure voltage, P max For power nodes p Rated maximum power P p For power nodes p Measure power; Weighted values ​​for the quality of associated communication nodes. H p For power transmission nodes p Data communication node set;

[0135] The power node state is calculated using a graph convolutional neural network, including the first layer of in-network feature aggregation:

[0136]

[0137] Where ReLU is the corrected linear unit. W (1) This is the first layer weight matrix. b (1) For the first layer bias;

[0138] Second layer output score:

[0139]

[0140] The sigmoid function is the activation function. W (2) This is the weight matrix for the second layer. b (2) For the second layer bias;

[0141] Unsupervised iterative optimization of the objective function:

[0142]

[0143] Wherein, parameter λ1+λ2+λ3=1;

[0144] Topological smoothness constraintsL topo :

[0145]

[0146] in, , Power nodes p、q Status score;

[0147] Physical property constraints L phys :

[0148]

[0149] in, P p,q For directional power flow. For inflow node p Total power;

[0150] Cross-network coupling consistency constraints L cross :

[0151]

[0152] R ( i ) is a regular expression term. or The regularization coefficient is... i The complexity of the model parameters;

[0153] Iterative process: The Adam optimizer is used to update the parameters until the loss converges, and the final power node state score is output. S p .

[0154] In a preferred embodiment, after obtaining the node status score, the method further includes screening weak nodes based on the status scores and topological importance of communication nodes and power nodes, locating weak links in cross-network associations by combining the cross-network coupling matrix, and outputting the causes and relationships of the weak points.

[0155] A preferred embodiment, combining cross-network coupling matrix to locate weak links in cross-network connections, includes:

[0156] Nodes with a status score below the health status threshold and a topological importance above the importance threshold are identified as critical weak nodes. The remaining nodes below the health status threshold are identified as general weak nodes, and all are sorted in descending order of their scores.

[0157] Based on cross-network coupling matrix C [ p , cIf the coupling strength is higher than the tight coupling threshold and both the corresponding power node and communication node are critical weak nodes, it is determined to be a core weak link in the cross-network; if the coupling strength is higher than the tight coupling threshold and at least one of them is a general weak node, it is determined to be a common weak link in the cross-network.

[0158] Another embodiment provides a node state assessment system coupled to a power grid and a communication network, used to implement the above-described node state assessment method coupled to a power grid and a communication network, comprising:

[0159] The network structure modeling module constructs weighted correlation matrices within the power grid and within the communication network to quantify the strength of topological associations between nodes in the same network.

[0160] The topology importance assessment module calculates the topology importance scores for power nodes and communication nodes based on the weighted correlation matrix and the real-time status of the nodes.

[0161] The cross-network coupling matrix construction module constructs a cross-network coupling matrix based on business characteristics and topological importance, and dynamically updates the elements of the cross-network coupling matrix in combination with the status of power nodes and communication nodes. The cross-network coupling matrix and the status of power and communication nodes form a bidirectional iterative update relationship, so that the cross-network coupling matrix and the status of power and communication nodes remain dynamically adapted.

[0162] The communication node status quantification and evaluation module calculates the communication node status score by weighted summation based on the communication node's own performance score, intra-network topology impact score, cross-network power feedback score, and topology importance score.

[0163] The power node state deep evaluation module uses the weighted correlation matrix within the power grid as a graph structure. It integrates the communication node state, topological importance score, and power node features to construct an input vector, and outputs the power node state score through unsupervised iterative optimization using a graph convolutional neural network.

[0164] The following example illustrates the workflow of a node state assessment system coupled with a power grid and communication network. Figure 2 As shown, it includes the following steps:

[0165] S1. Multi-source data acquisition and preprocessing: Collect ledger-type and measurement-type data from power grids and communication networks, and form an evaluation data source after spatiotemporal alignment and normalization;

[0166] S2. Network structure modeling: Construct weighted correlation matrices within the power grid and the communication network to quantify the strength of topological associations between nodes in the same network;

[0167] S3. Topological Importance Assessment: Based on the improved PageRank algorithm, combined with the weighted correlation matrix and the real-time status of nodes, the topological importance scores of power nodes and communication nodes are calculated respectively.

[0168] S4. Construction of cross-network coupling matrix: Initialize the cross-network coupling matrix based on business characteristics and topological importance, and dynamically update the matrix elements in combination with the status of power nodes and communication nodes to reflect the bidirectional influence between the two networks.

[0169] S5. Quantitative evaluation of communication node status: Based on four dimensions, namely “self-performance - intra-network topology impact - cross-network power feedback - topology importance”, combined with the weighted correlation matrix within the communication network, the cross-network coupling matrix and the topology importance score, the communication node status score is calculated by weighted summation.

[0170] S6. Deep evaluation of power node status: Using the weighted correlation matrix within the power grid as a graph structure, the input vector is constructed by fusing the status of communication nodes, topological importance scores, and power node features. The output power node status score is then optimized through unsupervised iterative optimization using a graph convolutional neural network.

[0171] S7. Weakness Analysis: Based on the state scores and topological importance of communication nodes and power nodes, weak nodes are screened, and the cross-network coupling matrix is ​​used to locate the weak links in the cross-network connection, outputting the causes and relationships of the weaknesses.

[0172] Specifically, the multi-source data acquisition and preprocessing steps include:

[0173] The data collected from power grids and communication networks includes ledger-type data such as:

[0174] Power grid ledger: node type, voltage level, rated capacity, topology connection relationship, business priority, reasonable voltage range, etc., imported through power GIS system and PMS system.

[0175] Communication network ledger: Node type, communication standard, link bandwidth, service latency threshold, equipment rated temperature, maximum tolerable temperature, etc., are imported through the communication resource management system.

[0176] The measurement data collected from power grids and communication networks include:

[0177] Power grid measurements: node voltage, current, active power, reactive power, switch status, insulation resistance, and line power flow are collected in real time through the power monitoring system.

[0178] Communication network measurements: node transmission delay, packet loss rate, signal strength, equipment temperature, and bit error rate are collected in real time through the communication network performance monitoring system.

[0179] For example, the power grid ledger and measurement data can be obtained by selecting equipment through the electricity information collection system (a system that exists in the prior art), exporting ledger data such as node type, voltage level, and rated capacity, and measurement data such as equipment voltage, current, active power, and reactive power. The ledger data is updated daily, and the measurement data sampling interval is 15 minutes. The sampling frequency can be modified according to the actual situation.

[0180] Data preprocessing includes:

[0181] Spatiotemporal alignment: Based on timestamps, power grid and communication network measurement data are synchronized in time. For data with out-of-sync clocks, a sliding window method is used to achieve time alignment. Power supply and communication ledgers are used to realize spatial association of nodes, ensuring that multi-source data in the same time and related areas can be matched.

[0182] For example, based on the boundaries of urban communities, the spatial association between power grid nodes and communication network nodes can be achieved through the power supply information of distribution transformers and base station coverage information within the same community.

[0183] Normalization: Using the rated value, maximum value, etc. as a benchmark, the measurement data is mapped to the [0,1] interval to eliminate dimensional differences.

[0184] For example, for node voltage data, reasonable upper and lower ranges can be defined based on the rated voltage, such as the reasonable upper and lower ranges of 220V voltage level being [180V, 260V], and normalization processing can be achieved through Min-Max standardization.

[0185] Network structure modeling steps include:

[0186] Weighted correlation matrix within the power grid ,in, For real numbers, n Number of power nodes:

[0187]

[0188] in, Y [ p , q [Power node] p and q Inter-normalized electrical coupling degree, Γ p for p The set of adjacent power nodes, Z [ p , q ]for p and q mutual impedance, Z [ p , p ]、Z [ q , q ]for p and q Self-impedance.

[0189] Weighted correlation matrix within the communication network ,in m Number of communication nodes:

[0190]

[0191] in, B [ c , k [Communication node] c and k Inter-link bandwidth, Ω c for c The set of adjacent communication nodes.

[0192] For example, consider the number of power nodes in a regional power distribution network. n =50, Number of communication nodes m =60. The weighted correlation matrix within the power grid is calculated using the node impedance matrix. The impedance data comes from the line parameter table exported from the PMS system. The weighted correlation matrix within the communication network is constructed based on the actual bandwidth configuration of the links in the communication resource management system.

[0193] The weighted association matrix is ​​used to quantify the association strength between nodes, providing weight inputs for the improved PageRank algorithm.

[0194] Specifically, the steps for topological importance assessment include:

[0195] Power node topology importance score I P [ p The results were obtained through iterative calculation using the improved Pagerank algorithm:

[0196]

[0197] in, d P [ p [This refers to the dynamic damping coefficient.] , (U p This is a voltage stability normalization value, ranging from [0,1], where the more stable the voltage, the better. d P [ p The larger the value, the stronger the node's ability to transmit importance to its neighbors.

[0198] initial value I P(0) [ p ]=1 / n Iterate to max| I P (t+1) [ p ]- I P (t) [ p If the convergence is less than a given threshold, then the power node is obtained. p Topological importance score I P [ p ].

[0199] Communication node topology importance score I C [ c The results were obtained through iterative calculation using the improved Pagerank algorithm:

[0200]

[0201] in d C [ c [This refers to the dynamic damping coefficient.] , L c The packet loss rate is in the range [0,1], where the lower the packet loss rate, the better. d C [ c The larger the value, the stronger the node's ability to transmit importance to its neighbors.

[0202] initial value I C (0) [c]=1 / m Iterate to max| I C (t+1) [c]- I C (t) [c]|Convergence occurs when the value is less than a given threshold, thus obtaining the communication node. c Topological importance score I C [ c ].

[0203] Topology importance assessment targets nodes within a single network, using a weighted correlation matrix and dynamic damping coefficients to reflect the influence of node states on the topology.

[0204] Specifically, the steps for constructing the cross-network coupling matrix include:

[0205] Initial cross-network coupling matrix :

[0206]

[0207] in:

[0208] e ( c () is the direct power supply node for obtaining communication node c;

[0209] R[ p , c [Power node] p Through communication nodes c The amount of data transmitted accounts for p The proportion of total transmitted power data;

[0210] I P [ p ]、 I C [ c These are the topological importance scores for power nodes and communication nodes, respectively.

[0211] K As a normalization constant, ensure C(0)[ p , c ]∈[0,1];

[0212] Dynamically updated formula:

[0213]

[0214] in:

[0215] σ is the normalized activation function. Q c ( t ) is a communication node c No. t Round state score (initialized to 100). Q ref This serves as a reference value for the health status of communication nodes. S p ( t ) is a power node p No. t Round state score (initialized to 100). S ref This serves as a reference value for the health status of power nodes.

[0216] For example, a normalized activation function is chosen as the Sigmoid function, the reference value for the health status of the communication node is set to 85, and the reference value for the health status of the power node is set to 90.

[0217] The cross-network coupling matrix and the state of the two network nodes form a bidirectional driving iterative update relationship, so that the cross-network coupling matrix and the state of the two network nodes always remain dynamically adapted.

[0218] Specifically, the steps for quantitatively evaluating the state of communication nodes include:

[0219] Communication Node c Status score Q c ∈[0,100]:

[0220]

[0221] Among them, weight α 1+ α 2 + α 3+ α 4 = 1, and the individual items are:

[0222] Self-performance score Q self,c :

[0223]

[0224] in, D c For the service latency of node c, D th For business latency threshold, L c For packet loss rate, T c Measuring temperature for communication equipment, T nom , T max The rated and maximum temperature of the communication equipment;

[0225] Network topology affects score Q com,c This reflects the relationship between adjacent communication nodes. c Status effects:

[0226]

[0227] Cross-grid power feedback score Q elec,c :

[0228]

[0229] Among them, Ψ c For a set of power nodes that transmit data through communication node c, L is the measurement distortion of data transmitted by power node p through c. p,cD is the packet loss rate for transmitting data from c to p. p,c D is the delay for C to transmit data to P. req,p Let p be the business latency requirement.

[0230] For example, the state score weighting coefficient of a communication node is selected as follows: α 1 = 0.4 α 2 =0.2, α 3 = 0.3 α 4 = 0.1, set according to expert experience.

[0231] The topology importance item directly incorporates the single-network topology importance of communication nodes into the state assessment. The higher the importance of a node, the greater the weight of its state score on the overall reliability of the system.

[0232] Specifically, the steps for in-depth assessment of power node status include:

[0233] Input feature vector x p :

[0234]

[0235] in:

[0236] , For normalized electrical quantities, U 0 represents the power node. p Rated voltage, U p For power nodes p Measure voltage, P max For power nodes p Rated maximum power P p For power nodes p Measure power.

[0237] Weighted values ​​for the quality of associated communication nodes. H p For power transmission nodes p A set of communication nodes for data.

[0238] The power node state is calculated using a graph convolutional neural network, including the first layer of in-network feature aggregation:

[0239]

[0240] in, W (1) This is the first layer weight matrix. b (1) For the first layer bias;

[0241] The power node state is computed using a graph convolutional neural network, including the output score of the second layer:

[0242]

[0243] in, W (2) This is the weight matrix for the second layer. b (2) For the second layer bias;

[0244] For example, one implementation of a graph convolutional neural network is using the PyTorch framework, with the first layer weight matrix W... (1) The dimension is 64×5, and the initialization uses a uniform Xavier distribution with a bias of b. (1) A 64-dimensional zero vector; the second-layer weight matrix W (2) The dimension is 1×64, and the initialization uses a He normal distribution; the bias b (2) It is a 1-dimensional zero vector.

[0245] Unsupervised iterative optimization of the objective function:

[0246]

[0247] in l 1+ l 2+ l 3=1, and the constraints for each item are:

[0248] Topological smoothness constraints L topo :

[0249]

[0250] The topology smoothing constraint aims to ensure that the scores of neighboring nodes in a power grid are similar. It calculates a penalty on the node state score, so that the node state score of nodes that do not meet the requirement of having similar scores of neighboring nodes is reduced.

[0251] Physical property constraints L phys :

[0252]

[0253] in P p,q For directional power flow, positive indicates outflow and negative indicates inflow. For inflow node p Total power.

[0254] Power grid constraints penalize the state of power nodes based on three indicators: power balance, voltage stability, and load balancing, thereby reducing the state score of nodes that violate physical criteria.

[0255] Cross-network coupling consistency constraints L cross :

[0256]

[0257] Cross-network coupling consistency constraints target coupling strength C [ p , c The larger the [size], the better. S p and Q c The higher the matching degree requirement, the more penalties are imposed on the node's state calculation, resulting in a decrease in the state score of nodes that do not conform to cross-network coupling consistency.

[0258] R ( i ) is a regular expression term. or These are regularization coefficients, intended to constrain model parameters. i This reduces the complexity of the network and prevents overfitting to specific network states during training, thereby improving the robustness and generalization ability of the state evaluation model when facing unseen operating scenarios.

[0259] For example, the coefficients of an unsupervised iterative optimization objective function are selected as follows: l 1 = 0.4 l 2 = 0.4, l 3 = 0.2, regularization coefficient or =0.01, set according to expert experience.

[0260] Iterative process: The Adam optimizer is used to update the parameters until the loss converges, and the final power node state score is output. S p .

[0261] Specifically, the steps for vulnerability analysis include:

[0262] Single-network weak node screening: Combining status score and single-network topology importance, nodes with status scores below the health status threshold and topology importance above the importance threshold are identified as critical weak nodes, while other nodes below the health status threshold are general weak nodes, all sorted in descending order of score.

[0263] Cross-network weak point location: based on cross-network coupling matrix C [ p , cIf the coupling strength is higher than the tight coupling threshold and both the corresponding power node and communication node are critical weak nodes, it is determined to be a core weak link in the cross-network; if the coupling strength is higher than the tight coupling threshold and at least one of them is a general weak node, it is determined to be a common weak link in the cross-network.

[0264] Output of weak points: For weak nodes in communication, output the influencing factors (such as excessive latency, excessive packet loss rate, abnormal equipment temperature, or mismatch between the importance and status of a single network topology); for weak nodes in power, output the influencing factors (such as power imbalance, excessive voltage fluctuation, overload, or mismatch between the importance and status of a single network topology); for weak links across networks, output the correlation between coupling strength, the status of nodes in the two networks, and the importance of a single network topology, and clarify the weak transmission path.

[0265] For example, if a weak node in communication is found through vulnerability analysis, the output will be "Latency exceeds limit: 150ms>100ms"; if a weak link in the cross-network core is found, the output will be "Coupling strength between power node P-001 and communication node C-005 is 0.8, and the status scores of both are below 60". The analysis results are exported in the form of a report for operation and maintenance personnel to make decisions.

[0266] In another embodiment, a computer storage medium stores a computer program thereon, wherein when a computer executes the computer program, it implements the node state assessment method for power grid-communication network coupling described in any of the above embodiments.

[0267] The specific implementation method is the one described above, and will not be repeated here.

[0268] In another embodiment, an electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program stored in the memory, wherein the computer program, when executed, implements the node state assessment method for power grid-communication network coupling as described in any of the preceding embodiments.

[0269] The specific implementation method is the one described above, and will not be repeated here.

[0270] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method of node state assessment for a power-communication network coupling, characterized in that, The method comprises the following steps: constructing a power grid internal weighted association matrix and a communication network internal weighted association matrix to quantify the topological association strength between nodes in the same network; calculating the topological importance scores of power nodes and communication nodes according to the weighted association matrix and the real-time state of the nodes; constructing a cross-network coupling matrix based on the service characteristics and the topological importance, and dynamically updating the elements of the cross-network coupling matrix in combination with the states of the power nodes and the communication nodes; the cross-network coupling matrix and the states of the power nodes and the communication nodes form an iterative updating relationship driven by both, so that the cross-network coupling matrix and the states of the power nodes and the communication nodes remain dynamically adaptive; calculating the state score of a communication node by weighted summation according to the performance score of the communication node itself, the topological influence score, the cross-network power feedback score and the topological importance score; taking the power grid internal weighted association matrix as a graph structure, fusing the state of the communication node, the topological importance score and the power node characteristics to construct an input vector, and outputting the state score of the power node by unsupervised iterative optimization of a graph convolutional neural network.

2. A method of node state assessment for a power-communication network coupling according to claim 1, characterized in that, The constructing of the power grid internal weighted association matrix and the communication network internal weighted association matrix comprises: Weighted incidence matrix within the power grid : wherein, Y [ p , q ] are the normalized electrical coupling between the power nodes p and q , p are the self-impedances of the power nodes p , Z [ p , q ] are the mutual impedances between the power nodes p and q , Z [ p , p ], Z [ q , q ] are the self-impedances of the power nodes p and q . Weighted incidence matrix within a communication network : wherein B [ c , k ] is a communication node c with k a link bandwidth, Ω c is c a set of neighboring communication nodes of 3. A method of node state assessment for a power-communication network coupling according to claim 2, characterized in that, The calculating of the topological importance scores of the power nodes and the communication nodes comprises: iteratively calculating the topological importance score of a power node by an improved Pagerank algorithm: wherein, t is the iteration number, I P [ p ] is the topology importance score of the power node I P [ q ] is the topology importance score of the power node p, q d P [ p ] is the dynamic damping coefficient, , U p is the voltage stability normalized value, the more stable the voltage, d P [ p ] is the voltage stability normalized value, the more stable the voltage, n is the number of power nodes;​ initial value I P (0) [ p ]=1 / n , iterates to max I P (t+1) [ p ]- I P (t) [ p ]| less than a given threshold convergence, resulting in power node p topology importance score I P [ p ] iteratively calculating the topological importance score of a communication node by an improved Pagerank algorithm: wherein, I C [ c ] is a topology importance score of the communication node I C [ k ] respectively, c, k d C [ c ] is a dynamic damping coefficient, , L c d C [ c ] is a packet loss rate, the lower the packet loss rate, m is the number of communication nodes;​​ initial value I C (0) [c]=1 / m , iterating to max I C (t+1) [c]- I C (t) [c]| less than a given threshold converges, resulting in a communication node c topological importance score I C [ c ].

4. A method of node state assessment for a power-communication network coupling according to claim 3, characterized in that, The constructing of the cross-network coupling matrix based on the service characteristics and the topological importance, and the dynamic updating of the elements of the cross-network coupling matrix in combination with the states of the power nodes and the communication nodes comprises: Constructing a cross-network coupling matrix ; calculating an initial cross-network coupling matrix: wherein, e ( c ) is the direct power supply node of the communication node c; R[ p , c ] is the power node p The proportion of the amount of data transmitted by the communication node c to the total amount of power data transmitted; p I P [ p ]、 I C [ c ] respectively are the power node topology importance score, the communication node topology importance score; K is a normalization constant;​ the dynamic updating formula: Where: σ is the normalized activation function. Q c ( t ) is a communication node c No. t Round of scoring, Q ref This serves as a reference value for the health status of communication nodes. S p ( t ) is a power node p No. t Round of scoring, S ref This serves as a reference value for the health status of power nodes.

5. A method of node state assessment for a power-communication network coupling according to claim 4, characterized in that, Communication node c State score Q c Is: wherein the weights α 1+ α 2 + α 3+ α 4=1, I C [ c ] are the communication node topology importance scores; Self performance score Q self,c : wherein, D c is a node c service latency, D th is a service latency threshold, L c is a packet loss rate, T c is a communication device measured temperature, T nom , T max are a communication device rated temperature and a maximum temperature, respectively; In-network topology impact score Q com,c : Cross-network power feedback score Q elec,c : Ψ c is a set of power nodes c transmitting data through the communication node is a power node p through c a measurement distortion of transmitting data, L p,c is c a packet loss rate of transmitting p data, D p,c is c a time delay of transmitting p data, D req,p is p a service time delay requirement.

6. A method of node state assessment for a power-communication network coupling according to claim 5, characterized in that, The outputting of the state score of the power node by unsupervised iterative optimization of a graph convolutional neural network comprises: Constructing input feature vector x p : wherein, , is a normalized electrical quantity, U 0 is a power node p rated voltage, U p 0 is a power node p measured voltage, P max 0 is a power node p rated maximum power, P p 0 is a power node p measured power; is a correlated communication node quality weighting value, H p 0 is a power node p set of communication nodes transmitting data; calculating the state of the power node by using a graph convolutional neural network, including first-layer aggregation of the same-network characteristics: where ReLU is a rectified linear unit, W (1) is a first layer weight matrix, b (1) is a first layer bias; second-layer output of the score: wherein the sigmoid function is an activation function, W (2) is a second layer weight matrix, b (2) is a second layer bias; unsupervised iterative optimization of the objective function: wherein parameters λ1+λ2+λ3=1; Topology smoothing constraints L topo : wherein, , are power nodes p, q state scores; Physical property constraints L phys : in, P p,q For directional power flow. For nodes p The load power, For inflow node p Total power; Cross-network coupling consistency constraints L cross : R θ ) is a regularization term, η is a regularization coefficient, θ is a complexity of the model parameters;​ Iteration process: update parameters with Adam optimizer until loss converges, output final power node state score S p .

7. A method of node state assessment for a power-communication network coupling according to claim 1, characterized in that, After obtaining the state score of the node, it further comprises screening weak nodes based on the state scores of the communication nodes and the power nodes and the topological importance, locating weak links of cross-network association in combination with the cross-network coupling matrix, and outputting the causes and association relationships of the weak points.

8. A method of node state assessment for a power-communication network coupling according to claim 7, characterized in that, The locating of the weak links of cross-network association in combination with the cross-network coupling matrix comprises: determining the nodes whose state scores of the communication nodes and the power nodes are lower than a healthy state threshold and whose topological importance is higher than an importance threshold as key weak nodes, and determining the nodes lower than the healthy state threshold as general weak nodes, and arranging them in descending order of the scores; based on the cross-network coupling matrix, if the coupling strength is higher than a close coupling threshold and the corresponding power nodes and communication nodes are both key weak nodes, it is determined as a cross-network core weak link; if the coupling strength is higher than the close coupling threshold and at least one of the corresponding power nodes and communication nodes is a general weak node, it is determined as a cross-network ordinary weak link.

9. A system for assessing the state of nodes of a power-communication network coupling, characterized in that, The method for implementing the power grid-communication network coupling node state evaluation method according to any one of claims 1-8 comprises: a network structure modeling module for constructing a power grid internal weighted association matrix and a communication network internal weighted association matrix to quantify the topological association strength between nodes in the same network; A topology importance evaluation module calculates the topology importance scores of power nodes and communication nodes respectively according to the weighted correlation matrix and real-time states of the nodes; A cross-network coupling matrix construction module constructs a cross-network coupling matrix based on the business characteristics and the topology importance, and dynamically updates the elements of the cross-network coupling matrix in combination with the states of the power nodes and the communication nodes; the cross-network coupling matrix and the states of the power nodes and the communication nodes form an iterative updating relationship driven in both directions, so that the cross-network coupling matrix and the states of the power nodes and the communication nodes remain dynamically adaptive; A communication node state quantitative evaluation module calculates a communication node state score through weighted summation according to the communication node's own performance score, the in-network topology influence score, the cross-network power feedback score and the topology importance score; A power node state deep evaluation module takes the internal weighted correlation matrix of the power grid as a graph structure, fuses the communication node state, the topology importance score and the power node characteristics to construct an input vector, and outputs a power node state score through unsupervised iterative optimization of a graph convolutional neural network.

10. A computer storage medium having stored thereon a computer program, characterized in that The computer executes the computer program to implement the power grid-communication network coupled node state evaluation method of any one of claims 1-8.

11. An electronic device comprising a memory and a processor, characterized in that, The memory stores the computer program, and the processor runs the computer program stored on the memory, and the computer program is executed to implement the power grid-communication network coupled node state evaluation method of any one of claims 1-8.

Images