Power grid alarm perception method and system based on multi-source data

Abstract

The present invention relates to the field of data processing and prediction technology, and in particular to a power grid alarm perception method and system based on multi-source data. The method senses current imbalance by collecting the instantaneous values ​​of three-phase currents in the three-phase lines of the power grid, combines multi-dimensional operating data, uses a state prediction model to predict the operating status of nodes, and uses a graph theory algorithm to perform topological analysis on abnormal nodes to accurately identify power outage areas. The system integrates multiple decision-making factors such as communication status, historical power outage patterns, and meteorological data to dynamically assess power outage risks and realize intelligent alarm decision-making. Discrete Fourier transform is used to efficiently process data, significantly improving alarm accuracy, avoiding false alarms caused by parameter fluctuations in traditional threshold methods, effectively reducing missed alarms and false alarms, enhancing the reliability, safety and stability of power supply of the power grid, and providing an efficient and accurate alarm solution for complex power grid operating conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and prediction, and in particular to a power grid alarm perception method and system based on multi-source data. Background Art

[0002] With the continuous expansion of the scale of modern power systems and the increasing complexity of grid operation, power outage alarm perception is a key link in ensuring power supply reliability and rapid response to faults.

[0003] Currently, power grid outage alarm detection is primarily based on traditional threshold-based methods. These methods typically set thresholds for electrical parameters such as voltage and current at key monitoring nodes in the power grid. When the actual monitored parameter values ​​exceed the preset normal threshold range, a power outage is detected and an alarm is issued. However, threshold setting often relies on empirical experience and rough statistical analysis of historical data, making it difficult to adapt to complex and changing grid operating conditions. For example, during peak hours or when special distributed power sources are connected, normal fluctuations in electrical parameters can falsely trigger alarms. Summary of the Invention

[0004] In view of the above problems in the prior art, the present invention is proposed.

[0005] Therefore, the technical problem to be solved by the present invention is: to solve the problem that the traditional power grid power outage alarm method relies on fixed thresholds, resulting in a high false alarm rate, and the problem that under complex power grid operating conditions (such as peak power consumption and distributed power supply access), the alarm is inaccurate, real-time, and has a high risk of missed reports due to the lack of multi-source data fusion and intelligent analysis capabilities, thereby improving the accuracy, reliability and power supply security of the power grid power outage alarm.

[0006] To solve the above technical problems, the present invention provides the following technical solutions, a power grid alarm perception method based on multi-source data, including: collecting the instantaneous values ​​of the three-phase current in the three-phase lines of the power grid based on current transformers, and performing current imbalance perception based on the instantaneous values ​​of the three-phase current to obtain power grid current perception results; collecting multi-dimensional operation data of multiple nodes in the power grid based on the power grid current perception results; inputting the multi-dimensional operation data into a state prediction model to obtain the node operation status results output by the state prediction model; for target nodes with abnormal operation status results, performing topological analysis on the target nodes based on a graph theory algorithm combined with a power grid graph structure to obtain an identification result of a power outage area associated with the target node; performing power outage alarm perception based on the identification result combined with the current multi-decision fusion factors of the power grid to obtain a power outage alarm perception result of the power grid.

[0007] As a preferred embodiment of the multi-source data-based power grid alarm perception method of the present invention, obtaining the power grid current perception result includes performing a discrete Fourier transform on the instantaneous values ​​of the three-phase currents to extract the first current amplitude and first phase under the fundamental wave, and the second current amplitude and second phase under each harmonic;

[0008] Based on the first current amplitude, the first phase, the second current amplitude and the second phase, the amplitude changes and the phase changes of the positive sequence current, the negative sequence current and the zero sequence current are calculated; based on the amplitude changes and the phase changes, a current imbalance evaluation index is determined; based on the current imbalance evaluation index, the current imbalance of the power grid is sensed to obtain the current sensing result of the power grid.

[0009] As a preferred solution of the power grid alarm perception method based on multi-source data described in the present invention, the multi-dimensional operating data includes voltage data, current data, power data, phase data and equipment temperature data.

[0010] As a preferred solution of the multi-source data-based power grid alarm perception method of the present invention, the state prediction model is trained based on sample operation data and its corresponding state result labels, and includes a data processing layer, a deep learning neural network layer, and a Bayesian network layer;

[0011] The node operation status result includes inputting multi-dimensional operation data into the state prediction model, performing data processing on the multi-dimensional operation data based on the data processing layer, and outputting a multi-dimensional data feature vector;

[0012] Perform linear prediction on the multi-dimensional data feature vector based on the deep learning neural network layer and output the first node abnormality probability of each node;

[0013] Based on the Bayesian network layer, the prior knowledge and the first node abnormality probability are combined to perform probabilistic reasoning on the multi-dimensional data feature vector, and the second node abnormality probability of the output node is output;

[0014] Performing a state prediction based on the first node abnormality probability and the second node abnormality probability of the node, and outputting an operating state result of the node, including determining a final abnormality probability value of the node based on the first node abnormality probability and the second node abnormality probability of the node; if the final abnormality probability value is greater than or equal to a preset probability threshold, determining that the operating state result of the node is an abnormal state; if the final abnormality probability value is less than the preset probability threshold, determining that the operating state result of the node is a normal state;

[0015] Among them, the calculation of the final abnormal probability value is expressed as:

[0016] ,

[0017] ,

[0018] ,

[0019] ,

[0020] in, Indicates the The final abnormal probability value of the node, Indicates the The information entropy correlation degree of the nodes, Indicates the The first node abnormality probability of a node, Indicates the The second node abnormality probability of a node, Indicates the The information entropy of the first node abnormal probability of the node, Indicates the The information entropy of the second node abnormal probability of a node.

[0021] As a preferred solution of the power grid alarm perception method based on multi-source data described in the present invention, wherein: the topological analysis of the target node based on the graph theory algorithm combined with the power grid graph structure includes searching in the power grid graph structure with the target node as the starting node to obtain the traversed current search node and the node depth of the current search node; when the node depth is less than or equal to the preset depth threshold, a depth-first search is performed on the current search node based on the depth-first search function to obtain the first power outage edge; until the node depth is greater than the preset depth threshold, a breadth-first search is performed on the current search node based on the breadth-first search function until all nodes in the power grid graph structure are traversed to obtain the second power outage edge;

[0022] Connecting corresponding nodes based on the first power outage edge and the second power outage edge, and determining the connected components obtained by the connection as the identification result of the power outage area associated with the target node;

[0023] The power grid graph structure includes a graph structure in which the devices and nodes of the power grid are vertices and line connection relationships are edges.

[0024] As a preferred solution of the multi-source data-based power grid alarm perception method of the present invention, the depth-first search function is expressed as , Indicates the current search node, Indicates the node depth of the current search node;

[0025] The execution process of the depth-first search function is expressed as:

[0026] Mark the current search node Has been visited; traverse the current search node Adjacency list Adjacent nodes in ;

[0027] If the node of , and the node Not the target node , then recursively call the depth-first search function ; Representation node The operating status of Representation node The running status is abnormal. Representation node The running status is normal;

[0028] For nodes that are in normal operation , calculate the current search node With node The first correlation between , expressed as:

[0029] ,

[0030] in, Representation node The power value, Representation node The power value; Representation node The operating status of

[0031] If the first correlation If it is greater than the first preset threshold, the edge Mark it as the first power outage edge and recursively call ;

[0032] The execution process of the breadth-first search function is expressed as:

[0033] Initialize the queue , and the current search node Join the queue ;

[0034] For the queue The head node in , traverse the head node Adjacency list Adjacent nodes in ;

[0035] If the node of , and the node Not the target node , then recursively call the breadth-first search function;

[0036] If it is the target node, calculate the head node With node The second correlation between , expressed as:

[0037] ,

[0038] in, Representation node The temperature value, Representation node The temperature value, Representation node The phase value of Representation node Phase value; if the second correlation If it is greater than the second preset threshold, the edge Marked as the second power outage side.

[0039] As a preferred solution of the power grid alarm perception method based on multi-source data described in the present invention, wherein: the multi-decision fusion factors include the communication link status, the historical power outage pattern of the adjacent area and the current meteorological data; based on the identification result and the current multi-decision fusion factors of the power grid, power outage alarm perception is performed to obtain the power grid power outage alarm perception result, including quantizing the current communication link status of the power grid, the historical power outage pattern of the adjacent area and the current meteorological data to obtain the communication status quantization value, the power outage time concentration and the meteorological quantization value; based on the communication status quantization value, the power outage time concentration and the meteorological quantization value and the number of nodes involved in the power outage area in the identification result, the probability of power outage alarm occurrence is determined; if the power outage alarm probability is greater than the preset alarm threshold, the power outage alarm perception result of the power grid is determined to be the issuance of a power outage alarm; wherein the alarm information in the issued power outage alarm includes the detailed range of the area, the power outage duration, the cause of the power outage and the alarm level.

[0040] Another object of the present invention is to provide a power grid alarm perception system based on multi-source data. This system uses the instantaneous values ​​of the three-phase current in the three-phase line to perceive current imbalance in the power grid. If the current perception result is current imbalance, the system then perceives a power outage alarm for the power grid, thereby avoiding false triggering of power outage alarms due to fluctuations in electrical parameters in complex and changeable power grid operating conditions. Furthermore, during the power outage alarm perception process, the operating status results of each node are predicted in conjunction with a state prediction model. Thus, abnormal nodes can be accurately identified in complex and changeable power grid operating conditions. Then, a graph theory algorithm and the power grid graph structure are combined to perform topological analysis on the abnormal target nodes, accurately obtaining the identification results of the power outage area, thereby improving the accuracy of power outage alarm perception and ensuring the reliability of the power supply to the power grid.

[0041] To solve the above technical problems, the present invention provides the following technical solutions: a power grid alarm perception system based on multi-source data, comprising: a power grid alarm perception center, a current imbalance perception unit, a data acquisition unit, a state prediction unit, a topology analysis unit, and a power outage alarm perception unit;

[0042] The power grid alarm sensing center is connected to the current imbalance sensing unit, data acquisition unit, state prediction unit, topology analysis unit and power outage alarm sensing unit to manage each unit;

[0043] The current imbalance sensing unit is used to collect the instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid based on the current transformer, and perform current imbalance sensing based on the instantaneous values ​​of the three-phase currents to obtain the current sensing results of the power grid;

[0044] The data acquisition unit collects multi-dimensional operating data of multiple nodes in the power grid based on the current sensing results, including voltage data, current data, power data, phase data and device temperature data;

[0045] The state prediction unit inputs the multi-dimensional operation data into the state prediction model to obtain the operation state result of each node output by the state prediction model; the state prediction model is trained based on the sample operation data and the corresponding state result label;

[0046] The topology analysis unit is used to perform a topological analysis on a target node with an abnormal operating status based on a graph theory algorithm combined with a power grid graph structure to obtain an identification result of a power outage area associated with the target node; the power grid graph structure is a graph structure constructed with each device and node of the power grid as vertices and line connection relationships as edges;

[0047] The power outage alarm perception unit is used to perform power outage alarm perception based on the recognition result in combination with the current multiple decision-making fusion factors of the power grid to obtain a power outage alarm perception result of the power grid.

[0048] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the power grid alarm perception method based on multi-source data are implemented as described above.

[0049] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the power grid alarm perception method based on multi-source data as described above are implemented.

[0050] Beneficial effects of the present invention: By integrating multi-source data with intelligent models and combining graph theory algorithms to perform precise topological analysis of abnormal power grid nodes, the present invention significantly improves the accuracy of power outage alarms and avoids false alarms caused by parameter fluctuations in traditional threshold methods. The system uses discrete Fourier transforms to efficiently process data, integrating multi-dimensional decision-making information such as communication status, historical power outage patterns, and meteorological factors in real time, dynamically assessing power outage risks and supporting intelligent alarm decision-making. Through multi-source collaborative analysis and dynamic adaptability, it effectively reduces missed reports and false alarms, enhances the reliability, safety, and stability of power grid power supply, and provides an efficient and accurate alarm solution for complex power grid conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A flowchart of a power grid alarm perception method based on multi-source data is provided in accordance with an embodiment of the present invention.

[0053] Figure 2 A system structure diagram of a power grid alarm perception method based on multi-source data provided by one embodiment of the present invention.

[0054] Figure 3 This is a diagram of an electronic device implementing a power grid alarm perception method based on multi-source data provided by one embodiment of the present invention.

[0055] Figure 4 A diagram of a computer-readable storage medium implementing a method for sensing power grid alarms based on multi-source data provided by one embodiment of the present invention.

[0056] In the figure: 300, electronic device; 310, memory; 311, computer-executable instructions; 320, processor; 400, computer-readable storage medium; 411, computer program. DETAILED DESCRIPTION

[0057] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0058] Example 1, with reference to Figure 1 , is an embodiment of the present invention, which provides a power grid alarm perception method based on multi-source data, including:

[0059] S1: The instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid are collected using current transformers, and current imbalance sensing is performed based on the instantaneous values ​​of the three-phase currents to obtain the power grid current sensing results.

[0060] It should be noted that if Figure 1 As shown in S1, the instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid are collected based on the current transformer, and current imbalance perception is performed based on the instantaneous values ​​of the three-phase currents to obtain the current perception results of the power grid; the instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid are collected through the current transformer, and the discrete Fourier transform is performed on the instantaneous values ​​of the three-phase currents, and current imbalance perception is performed based on the discrete Fourier transform results to obtain the current perception results of the power grid.

[0061] Furthermore, obtaining the grid current sensing result includes performing discrete Fourier transform on the instantaneous values ​​of the three-phase currents, extracting the first current amplitude and first phase under the fundamental wave, and the second current amplitude and second phase under each harmonic;

[0062] The collected three-phase current instantaneous value sequences are: , , represents the number of sampling points, and the discrete Fourier transform (DFT) is expressed as:

[0063] ,

[0064] For three-phase current Perform DFT and obtain discrete Fourier transform results respectively ; Fundamental frequency The corresponding value is ( , is the sampling frequency);

[0065] The first current amplitude extracted and the first phase are the amplitudes of the discrete Fourier transform results at the fundamental frequency and phase ,Right now:

[0066] ,

[0067] ,

[0068] in, express The real part of express The imaginary part of ; for each harmonic, where the harmonic order is , the corresponding value is ;

[0069] The second current amplitude extracted and the second phase Expressed as:

[0070] ,

[0071] ,

[0072] Calculating amplitude changes and phase changes of the positive-sequence current, the negative-sequence current, and the zero-sequence current based on the first current amplitude, the first phase, the second current amplitude, and the second phase;

[0073] In the positive sequence component, the phasor relationship of the three phases is Leading Mutually , Leading Mutually Therefore, according to the first current amplitude, the first phase, the second current amplitude and the second phase, the positive sequence current is calculated Expressed as:

[0074] ,

[0075] in, Indicates plural, ;

[0076] In the negative sequence component, the phasor relationship of the three phases is Falling behind Mutually , Falling behind Mutually Therefore, the negative sequence current is calculated based on the first current amplitude, the first phase, the second current amplitude and the second phase. Expressed as:

[0077] ,

[0078] For zero-sequence current, the magnitude and direction are the same in the three phases;

[0079] Calculate the zero-sequence current based on the first current amplitude, the first phase, the second current amplitude and the second phase Expressed as:

[0080] ,

[0081] For positive sequence current The amplitude change of , positive sequence current The phase change is ; For negative sequence current The amplitude change of , negative sequence current The phase change is ; For zero sequence current The amplitude change of , zero sequence current The phase change is ;

[0082] Determine the current imbalance evaluation index based on the amplitude change and phase change; 、 and , and phase changes 、 and , determine the current imbalance evaluation index , expressed as:

[0083] ,

[0084] Based on the current imbalance evaluation index, the current imbalance of the power grid is sensed to obtain the current perception result of the power grid; the current imbalance evaluation index is compared with the preset threshold value. Compare the values ​​and get the comparison result; if the current imbalance evaluation index is greater than the preset threshold , then the current sensing result of the power grid is determined to be in a current imbalance state; if the current imbalance evaluation index is less than or equal to the preset threshold , it is determined that the current sensing result of the power grid is in a current balance state.

[0085] S2: Collect multi-dimensional operating data of multiple nodes in the power grid based on the grid current sensing results.

[0086] It should be noted that if Figure 1 As shown in S2, multi-dimensional operation data of multiple nodes in the power grid are collected based on the current sensing results.

[0087] Furthermore, the multi-dimensional operating data includes voltage data, current data, power data, phase data and equipment temperature data, and the equipment includes transformers, switch cabinets, etc.

[0088] S3: Input the multi-dimensional operation data into the state prediction model to obtain the node operation status result output by the state prediction model.

[0089] It should be noted that if Figure 1 As shown in S3, the multi-dimensional operating data is input into the state prediction model to obtain the operating status results of each node output by the state prediction model; the state prediction model is trained based on the sample operating data and its corresponding state result labels; the multi-dimensional operating data including voltage data, current data, power data, phase data and equipment temperature data is input into the state prediction model to obtain the operating status results of each node output by the state prediction model.

[0090] Furthermore, the state prediction model is trained based on the sample operation data and its corresponding state result labels, and includes a data processing layer, a deep learning neural network layer, and a Bayesian network layer;

[0091] Furthermore, obtaining the node operation status result includes inputting the multi-dimensional operation data into the state prediction model, performing data processing on the multi-dimensional operation data based on the data processing layer, and outputting a multi-dimensional data feature vector;

[0092] Standardize multi-dimensional operation data. The voltage data of each node is , the current data is , the power data is , the phase data is , the device temperature data is ; For voltage data , standardized formula Expressed as:

[0093] ,

[0094] in, represents the mean value of all node voltage data, Indicates the standard deviation of voltage data;

[0095] Current data , power data , phase data , device temperature data Standardize and combine the standardized data into a multi-dimensional data feature vector ;

[0096] Perform linear prediction on the multi-dimensional data feature vector based on the deep learning neural network layer and output the first node abnormality probability of each node;

[0097] Among them, the deep learning neural network layer is a multi-layer perceptron (MLP) structure, which includes the input layer, hidden layers and output layers; the number of neurons in the input layer is the same as the dimension of the multidimensional data feature vector, that is, the number of neurons in the input layer , corresponding to 5 types of data; the first hidden layer to the The number of neurons in the hidden layers is , the number of neurons in the output layer , used to output the probability value of whether the node is abnormal;

[0098] For the The output of neurons in the hidden layer With the The output of neurons in the hidden layer The relationship is expressed as:

[0099] ,

[0100] in, Indicates the The hidden layer neurons to the The hidden layer The weight of a neuron, Indicates the The hidden layer The bias of a neuron, represents the activation function;

[0101] Activation Function Expressed as:

[0102] ,

[0103] For the output layer, the output layer The output of each hidden layer is linearly processed to obtain the first node abnormality probability of each output node;

[0104] The calculation of the abnormal probability of the first node is expressed as:

[0105] ,

[0106] in, Indicates the The first node abnormality probability of a node, Indicates the The output of the hidden layer;

[0107] Loss functions when training deep learning neural network layers Expressed as:

[0108] ,

[0109] in, represents the number of training samples, Indicates the The true label of the sample node (0 means normal, 1 means abnormal), represents the regularization coefficient;

[0110] Based on the Bayesian network layer, the prior knowledge and the first node abnormality probability are combined to perform probabilistic reasoning on the multi-dimensional data feature vector, and the second node abnormality probability of the output node is output;

[0111] Among them, the Bayesian network layer includes nodes, corresponding to different features in the multi-dimensional data feature vector; each node represents a random variable, such as Nodes represent voltage characteristics, Nodes represent current characteristics, etc.; edges between nodes represent dependencies between variables;

[0112] For each node, determine the prior probability distribution; for example, for the node (voltage characteristics), based on historical data and experience, determine the prior probability under normal and abnormal conditions and ;

[0113] For nodes with parent nodes, determine the conditional probability distribution; node Has a parent node and parent node , then the conditional probability is ;node ,node and nodes is a continuous variable, the conditional probability Calculated by kernel density estimation combined with neural network prediction; 、 and Node ,node and nodes The probability density function of Node and nodes For input prediction node The mean , through neural networks Prediction Node Variance , expressed as:

[0114] ,

[0115] According to the abnormal probability of the first node, the Bayesian network is used for probabilistic reasoning to calculate the node Abnormal probability of the second node in abnormal state ;

[0116] According to the joint probability distribution formula of Bayesian network ,in, Representation node The parent node set is inferred by variable elimination method or joint tree algorithm, based on the hierarchical decomposition and parallel calculation of the graph; the Bayesian network is decomposed according to the hierarchical structure of the nodes, and the conditional probability is calculated in parallel at different levels. The node is obtained through information transmission and fusion. Abnormal probability of the second node in abnormal state ;

[0117] The state prediction is performed based on the first node abnormality probability and the second node abnormality probability of the node, and the operation state result of the node is output; specifically, the corresponding information entropy is calculated according to the first node abnormality probability and the second node abnormality probability of each node, which is expressed as:

[0118] ,

[0119] ,

[0120] in, Indicates the The information entropy of the first node abnormal probability of the node, Indicates the The information entropy of the second node abnormal probability of a node;

[0121] According to the first node abnormal probability and the second node abnormal probability of each node, as well as the information entropy of the first node abnormal probability and the information entropy of the second node abnormal probability, the information entropy correlation degree between the information entropies of each node is calculated, which is expressed as:

[0122] ,

[0123] in, Indicates the The degree of correlation of information entropy of nodes;

[0124] The final abnormal probability value of each node is calculated based on the first node abnormal probability and the second node abnormal probability of each node, as well as the information entropy correlation degree, which is expressed as:

[0125] ,

[0126] Among them, among them, Indicates the The final abnormal probability value of the node, Indicates the The first node abnormality probability of a node, Indicates the The probability of a node's second node being abnormal;

[0127] If it is determined that the final abnormal probability value is greater than or equal to the preset probability threshold, the operating status result is determined to be that each node is in an abnormal state, where the preset probability threshold is set according to actual conditions; if it is determined that the final abnormal probability value is less than the preset probability threshold, the operating status result is determined to be that each node is in a normal state.

[0128] S4: For target nodes with abnormal operating status, a topological analysis is performed on the target nodes based on a graph theory algorithm combined with the power grid graph structure to obtain an identification result of a power outage area associated with the target node.

[0129] It should be noted that if Figure 1 As shown in S4, for the target node whose operating status result is abnormal, a topological analysis is performed on the target node based on a graph theory algorithm combined with a power grid graph structure to obtain an identification result of the power outage area associated with the target node; a power grid graph structure is pre-constructed, and the power grid graph structure is a graph structure constructed with each device and node of the power grid as vertices and line connection relationships as edges; for the target node whose operating status result is in an abnormal state, a topological analysis is performed on the target node through a graph theory algorithm combined with a power grid graph structure, wherein the topological analysis includes a depth-first search and a breadth-first search; a depth-first search and a breadth-first search are performed on the target node through a graph theory algorithm combined with a power grid graph structure to determine the power outage impact range associated with the target node, and mark the power outage area to obtain an identification result of the power outage area associated with the target node.

[0130] Furthermore, the target node is used as the starting node to search in the power grid graph structure, and the traversed current search node and the node depth of the current search node are obtained;

[0131] Representing the power grid graph structure with adjacency lists ,in, Represents a vertex set (i.e., various devices and nodes in the power grid), Represents an edge set (i.e., line connection relationship);

[0132] For each vertex , with a linked list Storage and Vertices Adjacent vertices; the running status of each node is expressed in the state array Indicates that, Representation node It runs normally. Representation node Abnormal operation;

[0133] For the power outage area, use a two-dimensional array Indicates that, Representation node and nodes Whether the edges between them are in the power outage area, the initial values ​​are , that is, node and nodes There is no connection between them;

[0134] The depth-first search function can be expressed as ,in, Indicates the current search node, Indicates the node depth of the current search node;

[0135] From the abnormal target node Start calling the depth-first search function Perform depth-first search to obtain the target node The current search section during the search and the current search section Node depth ;

[0136] When the node depth is less than or equal to the preset depth threshold, a depth-first search is performed on the current search node based on the depth-first search function to obtain the first power outage edge. When the node depth is greater than the preset depth threshold, a breadth-first search is performed on the current search node based on the breadth-first search function until all nodes in the power grid graph structure are traversed to obtain the second power outage edge. When the node depth is greater than the preset depth threshold, a breadth-first search is performed on the current search node based on the breadth-first search function to obtain the second power outage edge. Less than or equal to the preset depth threshold When the depth-first search function is used to perform a depth-first search on the current search node, the first power outage edge is obtained, where the preset depth threshold , depth-first search function The execution process is expressed as:

[0137] Mark the current search node Has been visited, that is ;

[0138] Traverse the current search node Adjacency list Each adjacent node in ;

[0139] Check Node Running status , if the node of , and, node Not the target node To avoid repeated searches for abnormal nodes, recursively call .in, Representation node The operating status of Representation node The running status is abnormal. Representation node The running status is normal;

[0140] For nodes that are in normal operation , then calculate the current search node With node The first correlation between , expressed as:

[0141] ,

[0142] in, and Represents nodes respectively and nodes The power value; Representation node The operating status of

[0143] If the first correlation is greater than the first preset threshold, wherein the first preset threshold is set according to the actual situation. For example, if the first preset threshold is 0.6, the edge Marked as the first power outage edge, , , and recursively call ;

[0144] When the node depth Reaching the preset depth threshold When , stop the depth-first search, need to search the surrounding nodes more comprehensively, and more accurately determine the scope of the power outage, then switch to breadth-first search; breadth-first search function The execution process is expressed as:

[0145] With a queue Store the nodes to be accessed and initialize the queue , and the current search node Join the queue ; When the queue When it is not empty, remove the queue The head node in ; For queues The head node in , traverse the head node Adjacency list Each adjacent node in ;

[0146] If the node of , and the node Not the target node , then continue to recursively call the breadth-first search function, that is, the node Join the queue Perform a breadth-first search ;

[0147] For nodes that are in normal operation , then calculate the head node With node The second correlation between , expressed as:

[0148] ,

[0149] in, and Node and nodes The temperature value, and Node and nodes The phase value of

[0150] If the second correlation is greater than the second preset threshold, wherein the second preset threshold is set according to the actual situation. For example, if the second preset threshold is set to 0.7, the edge Marked as the second power outage side, , , and the node Join the queue ;

[0151] Connect the corresponding nodes based on the first power outage edge and the second power outage edge, and determine the connected components obtained by the connection as the power outage area associated with the target node; traverse Array, for all first power outage edges and second power outage edges, , The edge of the two nodes and nodes Divide into different connected components, each connected component is a power outage area that may be associated with the target node; calculate the node The impact of power outage , expressed as:

[0152] ,

[0153] For all nodes, Sort the values ​​by Starting from the node with the larger value, the nodes connected to the node through the power outage edge are clustered into a power outage area; if there is a power outage edge between the nodes of two different clusters, and the average If the difference between the two values ​​is less than a preset threshold, such as 0.5, the two clusters are merged; the final power outage area clustering result is the identification result of the power outage area associated with the target node, and the set of nodes in each cluster and the power outage edges between them constitute a power outage area.

[0154] S5: Based on the recognition result and the current multi-decision fusion factors of the power grid, power outage alarm perception is performed to obtain the power outage alarm perception result of the power grid.

[0155] It should be noted that if Figure 1 As shown in S5, based on the recognition result and the current multi-decision fusion factors of the power grid, power outage alarm perception is performed to obtain the power outage alarm perception result of the power grid; based on the recognition result and the current multi-decision fusion factors of the power grid, power outage alarm perception is performed, that is, power outage alarm perception is performed in combination with the current communication link status of the power grid, the historical power outage pattern of the adjacent area and the current meteorological data, and the power outage alarm perception result of the power grid is obtained.

[0156] Furthermore, the current communication link status of the power grid, the historical power outage patterns in adjacent areas, and the current meteorological data are quantified to obtain the communication status quantization value, the power outage time concentration, and the meteorological quantization value;

[0157] The current communication link status of the power grid is quantified based on the received signal strength indicator (RSSI). RSSI is a measure of received signal power, typically measured in decibel milliwatts (dBm). In power grid communication systems, such as wireless communications, RSSI values ​​can be directly read on communication devices (e.g., smart meters, communication modules between distributed energy resource controllers and the grid control center). Higher RSSI values ​​indicate stronger signal strength.

[0158] The communication link status is divided into three levels: good, general, and poor. When the communication link status is good, such as the RSSI value is greater than -70dBm, the communication status quantization value ; The communication link status is normal, such as when the RSSI value is between -70dBm and -90dBm, the communication status quantization value When the communication link status is poor, such as when the RSSI value is less than -90dBm, the communication status quantization value ;

[0159] For the quantitative processing of the historical power outage patterns of the adjacent areas of the current power grid, the power outage data of the adjacent areas in the past period of time (such as the past year) are collected. The power outage data includes information such as the power outage time, power outage duration and power outage range; the number of adjacent areas is , for the The historical power outage frequency of adjacent areas is expressed as:

[0160] ,

[0161] Considering the distribution of power outage time, calculate the concentration of power outage time Divide a day into 24 periods, and count the proportion of power outages in each period to the total number of power outages. The power outage time concentration is calculated as follows: Expressed as:

[0162] ,

[0163] in, Indicates the The region in The probability of power outage during a certain period;

[0164] Meteorological data includes temperature, humidity, wind speed and rainfall; the current meteorological data of the power grid is quantified to obtain the temperature impact factor , humidity influencing factors , wind speed influencing factors and rainfall influencing factors , expressed as:

[0165] ,

[0166] in, It represents the average temperature of the region during the same period in history. and Indicates the highest and lowest temperatures in the same period of history; similarly, the degree impact factor is obtained , wind speed influencing factors and rainfall influencing factors ;

[0167] According to the temperature influence factor , humidity influencing factors , wind speed influencing factors and rainfall influencing factors Determine meteorological quantitative values , expressed as:

[0168] ,

[0169] in, represents the weight coefficient, and ;

[0170] Based on the communication state quantization value, the power outage time concentration, the meteorological quantization value, and the number of nodes involved in the power outage area in the identification result, the probability of power outage alarm occurrence is determined; the number of nodes involved in the power outage area in the identification result is obtained, and the power outage alarm occurrence probability is determined based on the communication state quantization value, the power outage time concentration, the historical power outage frequency, the meteorological quantization value, and the number of nodes involved in the power outage area in the identification result, expressed as:

[0171] ,

[0172] in, Indicates the probability of power outage alarm; and Represents the weight coefficient related to the adjacent area, which is determined by the distance between the adjacent area and the current power outage area and the degree of power transmission correlation; Indicates the number of nodes involved in the power outage area in the identification result, Indicates the total number of nodes in the power grid;

[0173] If the probability of a power outage alarm is greater than a preset alarm threshold, the power outage alarm perception result of the power grid is determined to be the issuance of a power outage alarm; the power outage alarm probability is compared with the preset alarm threshold to obtain a comparison result, wherein the preset alarm threshold is set according to actual conditions; if it is determined that the probability of a power outage alarm is greater than the preset alarm threshold, the power outage alarm perception result of the power grid is determined to be the issuance of a power outage alarm;

[0174] The alarm information in the issued power outage alarm includes the detailed area range, power outage duration, power outage cause and alarm level. Among them, the power outage duration can be inferred based on the historical power outage pattern of the adjacent area; the power outage cause can be obtained based on meteorological data analysis, such as high temperature causing equipment failure, strong wind damaging the line, etc.; the alarm level can be refined according to different decision fusion factors. For example, if the communication link status is good and the historical power outage frequency in the adjacent area is low, but the meteorological impact index is high and the power outage area is large, the alarm level is orange warning; if the communication link status is poor and the historical power outage pattern in the adjacent area shows frequent power outages in the near future, and the meteorological conditions are severe, the alarm level is red warning.

[0175] The above is a schematic scheme of a multi-source data-based power grid alarm perception method according to this embodiment. It should be noted that the technical scheme of the system for the multi-source data-based power grid alarm perception method and the technical scheme of the multi-source data-based power grid alarm perception method described above are based on the same concept. For details not described in detail in the technical scheme of the multi-source data-based power grid alarm perception system according to this embodiment, please refer to the description of the technical scheme of the multi-source data-based power grid alarm perception method described above.

[0176] Example 2, reference Figure 2-Figure 4 , as one embodiment of the present invention, provides a power grid alarm perception system based on multi-source data, comprising:

[0177] like Figure 2 As shown, the power grid alarm perception center, current imbalance perception unit, data acquisition unit, state prediction unit, topology analysis unit and power outage alarm perception unit;

[0178] The power grid alarm sensing center is connected to the current imbalance sensing unit, data acquisition unit, state prediction unit, topology analysis unit and power outage alarm sensing unit to manage each unit;

[0179] The current imbalance sensing unit is used to collect the instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid based on the current transformer, and perform current imbalance sensing based on the instantaneous values ​​of the three-phase currents to obtain the current sensing results of the power grid;

[0180] The data acquisition unit collects multi-dimensional operating data of multiple nodes in the power grid based on the current sensing results, including voltage data, current data, power data, phase data and device temperature data;

[0181] The state prediction unit inputs the multi-dimensional operation data into the state prediction model to obtain the operation state result of each node output by the state prediction model; the state prediction model is trained based on the sample operation data and the corresponding state result label;

[0182] The topology analysis unit is used to perform a topological analysis on a target node with an abnormal operating status based on a graph theory algorithm combined with a power grid graph structure to obtain an identification result of a power outage area associated with the target node; the power grid graph structure is a graph structure constructed with each device and node of the power grid as vertices and line connection relationships as edges;

[0183] The power outage alarm perception unit is used to perform power outage alarm perception based on the recognition result in combination with the current multiple decision-making fusion factors of the power grid to obtain a power outage alarm perception result of the power grid.

[0184] like Figure 3 As shown, this embodiment further provides an electronic device 300, which is applicable to the case of a power grid alarm perception method based on multi-source data, including:

[0185] Memory 310 and processor 320; the memory 310 is used to store computer-executable instructions 311, and the processor 320 is used to execute the computer-executable instructions 311 to implement the power grid alarm perception method based on multi-source data as proposed in the above embodiment.

[0186] like Figure 4As shown, this embodiment further provides a computer-readable storage medium 400 on which a computer program 411 is stored. When the computer program 411 is executed by a processor, the power grid alarm perception method based on multi-source data proposed in the above embodiment is implemented.

[0187] The storage medium proposed in this embodiment and the power grid alarm perception method based on multi-source data proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, please refer to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0188] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0189] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0190] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A power grid alarm perception method based on multi-source data, characterized by: include: The instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid are collected based on the current transformers, and current imbalance is sensed based on the instantaneous values ​​of the three-phase currents to obtain the power grid current sensing results; Collect multi-dimensional operating data of multiple nodes in the power grid based on the power grid current sensing results; Inputting multi-dimensional operation data into a state prediction model to obtain the node operation status result output by the state prediction model; wherein the state prediction model is trained based on sample operation data and its corresponding state result labels, and includes a data processing layer, a deep learning neural network layer, and a Bayesian network layer; The node operation status result includes inputting multi-dimensional operation data into the state prediction model, performing data processing on the multi-dimensional operation data based on the data processing layer, and outputting a multi-dimensional data feature vector; Perform linear prediction on the multi-dimensional data feature vector based on the deep learning neural network layer and output the first node abnormality probability of each node; Based on the Bayesian network layer, the prior knowledge and the first node abnormality probability are combined to perform probabilistic reasoning on the multi-dimensional data feature vector, and the second node abnormality probability of the output node is output; Performing a state prediction based on the first node abnormality probability and the second node abnormality probability of the node, and outputting an operating state result of the node, including determining a final abnormality probability value of the node based on the first node abnormality probability and the second node abnormality probability of the node; if the final abnormality probability value is greater than or equal to a preset probability threshold, determining that the operating state result of the node is an abnormal state; if the final abnormality probability value is less than the preset probability threshold, determining that the operating state result of the node is a normal state; For target nodes with abnormal operating status, a topological analysis of the target nodes is performed based on graph theory algorithms combined with the power grid graph structure to obtain the identification results of the power outage areas associated with the target nodes. Based on the recognition results and the current multi-decision fusion factors of the power grid, power outage alarm perception is performed to obtain the power outage alarm perception results of the power grid; The topological analysis of the target node based on the graph theory algorithm combined with the power grid graph structure includes searching the power grid graph structure with the target node as the starting node to obtain the traversed current search node and the node depth of the current search node; when the node depth is less than or equal to a preset depth threshold, performing a depth-first search on the current search node based on a depth-first search function to obtain a first power outage edge; until the node depth is greater than the preset depth threshold, performing a breadth-first search on the current search node based on a breadth-first search function until all nodes in the power grid graph structure are traversed to obtain a second power outage edge; Connecting corresponding nodes based on the first power outage edge and the second power outage edge, and determining the connected components obtained by the connection as the identification result of the power outage area associated with the target node; The power grid graph structure includes a graph structure in which the devices and nodes of the power grid are vertices and the line connection relationships are edges; Among them, the calculation of the final abnormal probability value is expressed as: in, represents the final abnormal probability value of the i-th node, Y(i) represents the information entropy correlation degree of the i-th node, represents the first node abnormality probability of the i-th node, represents the abnormal probability of the second node of the i-th node, H n (i) represents the information entropy of the first node abnormality probability of the i-th node, H b (i) represents the information entropy of the second node abnormality probability of the i-th node.

2. The power grid alarm perception method based on multi-source data according to claim 1, characterized in that: The depth-first search function is expressed as DFS(v, depth), where v represents the current search node and depth represents the node depth of the current search node; The execution process of the depth-first search function is expressed as: Mark the current search node v as visited; traverse the adjacent node u1 in the adjacency list adj(v) of the current search node v; If Status[u1] of node u1 == 1, and node u1 is not the target node target i , then recursively call the depth-first search function DFS(u1,depth+1); Status[u1] represents the running status of node u1, Status[u1]==1 means that the running status of node u1 is abnormal, Status[u1]==0 means that the running status of node u1 is normal; For the node u1 in normal operation, calculate the first association degree association(v,u1) between the current search node v and node u1, which is expressed as: Where power(v) represents the power value of node v, power(u1) represents the power value of node u1; Status[v] represents the operating status of node v; If the first association (v, u1) is greater than the first preset threshold, the edge (v, u1) is marked as the first power outage edge, and DFS (u, depth + 1) is recursively called; The execution process of the breadth-first search function is expressed as: Initialize the queue Q and add the current search node v to the queue Q; For the head node u2 in queue Q, traverse the adjacent node w in the adjacency list adj(u2) of the head node u2; If Status[w] of node w == 1, and node w is not the target node target i , then recursively call the breadth-first search function; If it is the target node, the second association degree association(u2,w) between the head node u2 and the node w is calculated, which is expressed as: Among them, temperature(u2) represents the temperature value of node u2, temperature(w) represents the temperature value of node w, represents the phase value of node u2, represents the phase value of node w; if the second association (u2, w) is greater than the second preset threshold, the edge (u2, w) is marked as the second power outage edge; The multi-decision fusion factors include the communication link status, historical power outage patterns in adjacent areas, and current meteorological data; based on the identification results and combined with the current multi-decision fusion factors of the power grid, power outage alarm perception is performed to obtain the power outage alarm perception result of the power grid, including quantifying the current communication link status of the power grid, historical power outage patterns in adjacent areas, and current meteorological data to obtain a communication status quantization value, a power outage time concentration, and a meteorological quantization value; based on the communication status quantization value, the power outage time concentration, the meteorological quantization value, and the number of nodes involved in the power outage area in the identification results, the probability of power outage alarm occurrence is determined; if the power outage alarm probability is greater than the preset alarm threshold, the power outage alarm perception result of the power grid is determined to be the issuance of a power outage alarm; wherein the alarm information in the issued power outage alarm includes the detailed range of the area, the power outage duration, the cause of the power outage, and the alarm level.

3. A power grid alarm perception system based on multi-source data, applying the power grid alarm perception method based on multi-source data according to any one of claims 1 to 2, characterized in that: include: Grid alarm perception center, current imbalance perception unit, data acquisition unit, state prediction unit, topology analysis unit, and power outage alarm perception unit; The power grid alarm sensing center is connected to the current imbalance sensing unit, data acquisition unit, state prediction unit, topology analysis unit and power outage alarm sensing unit to manage each unit; The current imbalance sensing unit is used to collect the instantaneous values ​​of the three-phase currents in the three-phase lines of the power grid based on the current transformer, and perform current imbalance sensing based on the instantaneous values ​​of the three-phase currents to obtain the current sensing results of the power grid; The data acquisition unit collects multi-dimensional operating data of multiple nodes in the power grid based on the current sensing results, including voltage data, current data, power data, phase data and device temperature data; The state prediction unit inputs the multi-dimensional operation data into the state prediction model to obtain the operation state result of each node output by the state prediction model; the state prediction model is trained based on the sample operation data and the corresponding state result label; The topology analysis unit is configured to perform a topology analysis on a target node whose operating status result is abnormal based on a graph theory algorithm combined with a power grid graph structure, and obtain an identification result of a power outage area associated with the target node; The power grid graph structure is a graph structure constructed with the power grid devices and nodes as vertices and line connection relationships as edges; The power outage alarm perception unit is used to perform power outage alarm perception based on the recognition result in combination with the current multiple decision-making fusion factors of the power grid to obtain a power outage alarm perception result of the power grid.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the power grid alarm perception method based on multi-source data according to any one of claims 1 to 2 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the power grid alarm perception method based on multi-source data according to any one of claims 1 to 2 are implemented.

Images