A distribution network equipment state real-time monitoring and early warning method and system

Abstract

The present application relates to the technical field of power distribution equipment state monitoring, and discloses a power distribution equipment state real-time monitoring and early warning method and system, comprising: establishing a source-load balance analysis model for describing the energy supply and energy consumption balance relationship of power distribution equipment, outputting the power supply frequency fluctuation value of the power distribution equipment when it is in supply-demand imbalance; according to the power supply frequency fluctuation value, judging the duration and fault frequency of the power distribution equipment when it is in supply-demand imbalance state, and outputting the performance evaluation value of the power distribution equipment; combining the performance evaluation value with the load safety early warning level, and outputting the power distribution equipment early warning scheme; outputting the real-time running data of the power distribution equipment, the load safety result of the power distribution equipment in the real-time state and the future period, and based on the load safety result, matching and executing the power distribution equipment monitoring and early warning scheme. The present application can identify the load safety risk that may occur in the future period in advance before the current state deteriorates, and improve the safety and reliability of the power distribution equipment operation.

Description

Technical Field

[0001] This invention relates to the field of power distribution network equipment status monitoring technology, and more specifically, to a method and system for real-time monitoring and early warning of power distribution network equipment status. Background Technology

[0002] Distribution network equipment refers to various devices used for power transmission, distribution and dispatch in the distribution network. It mainly includes distribution transformers, switchgear, circuit breakers, load switches, distribution cables, line equipment, grounding devices, etc. The main function of distribution equipment is to transmit high-voltage power from the transmission network to the user end and ensure that power can be safely and stably distributed to each user.

[0003] Real-time monitoring and early warning of distribution network equipment status refers to the timely detection of potential fault risks or abnormal states of equipment by real-time monitoring of various equipment in the distribution network, combined with data analysis and prediction models, and issuing early warning signals in advance, thereby realizing early identification and handling of equipment faults and avoiding power outages or other safety hazards caused by equipment faults.

[0004] However, existing real-time monitoring and early warning schemes for distribution network equipment status are usually based on fixed thresholds and single operating condition assumptions. They mainly rely on whether voltage, current or load exceeds the limit to trigger alarms. They lack a systematic characterization of uncertain operating scenarios and long-term operating pressure on equipment, making it difficult to achieve refined and forward-looking risk management.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] In response to the problems in related technologies, this invention proposes a method and system for real-time monitoring and early warning of the status of distribution network equipment, so as to overcome the above-mentioned technical problems existing in the existing related technologies.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows:

[0008] In a first aspect, the present invention provides a method for real-time monitoring and early warning of the status of distribution network equipment, comprising:

[0009] Analyze the load-bearing capacity of distribution network equipment according to monitoring standards, and establish load safety early warning levels under uncertain scenarios based on load-bearing capacity;

[0010] A source-load balance analysis model is established based on historical operating data to describe the balance between energy supply and energy consumption of power distribution equipment. The source-load balance analysis model is randomly sampled to output the power supply frequency fluctuation value of the power distribution equipment when it is in a state of supply-demand imbalance.

[0011] Based on the power supply frequency fluctuation value, determine the duration and fault frequency of the power distribution equipment in the state of supply and demand imbalance, iteratively train the backpropagation neural network model, and output the performance evaluation value of the power distribution equipment.

[0012] By combining performance evaluation values ​​with load safety warning levels, a warning scheme for power distribution equipment is output.

[0013] By combining real-time operating data of power distribution equipment with frequency calculation technology, the future operating status of power distribution equipment is simulated. Based on the simulation results, the load safety results of power distribution equipment in real-time state and in future time period are output, and a power distribution equipment monitoring and early warning scheme is matched and executed based on the load safety results.

[0014] Preferably, a source-load balance analysis model is established based on historical operating data to describe the balance between energy supply and energy consumption of power distribution equipment. Random sampling is performed on the source-load balance analysis model, and the output power supply frequency fluctuation values ​​of the power distribution network equipment when it is in a state of supply-demand imbalance include:

[0015] Obtain historical operational data containing the operating load and power supply capacity of power distribution equipment, and generate an operational topology graph by using the power supply line connections between power distribution equipment as edges and the historical operational data as nodes.

[0016] Analyze the operational topology to obtain the balance relationship between energy supply and energy consumption during power transmission, and use elastic network fitting technology to establish a source-load balance analysis model based on the balance relationship;

[0017] Based on variational inference technology, different energy supply and energy consumption data are extracted from the source-load balance analysis model as sampling points to simulate the power supply frequency fluctuation state of power distribution equipment under power supply fluctuation conditions.

[0018] Preferably, the analysis of the operating topology diagram is used to obtain the balance relationship between energy supply and energy consumption during power transmission, and a source-load balance analysis model based on the balance relationship is established using elastic network fitting technology, including:

[0019] The running topology graph is transformed into a multi-relationship graph based on meta-paths. Power quality relationship selection graph convolutional layers are used to analyze the multi-relationship graph to obtain aggregated neighbor information in order to model the global correlation between meta-paths.

[0020] The aggregated neighbor information is filtered to distinguish the importance of each power quality relationship, and the power quality relationships that meet the importance threshold are retained as the energy supply and demand balance changes of the power distribution equipment under different operating conditions;

[0021] By using the operating time and maintenance information of power distribution equipment as independent variables and the changes in energy supply and demand balance as dependent variables, a pool of characteristic indicators is obtained for constructing the source-load balance analysis model.

[0022] Based on the feature index pool, the influencing factor feature indexes are screened to determine the index combination used to construct the source-load balance analysis model. The source-load balance analysis model is constructed by using the index combination and elastic network fitting technology.

[0023] Preferably, the running topology graph is transformed into a multi-relationship graph based on meta-paths. Power quality relationship selection graph convolutional layers are used to analyze the multi-relationship graph to obtain aggregated neighbor information, in order to model the global correlation between meta-paths, including:

[0024] Based on the power transmission relationship between power distribution equipment and the power supply node, the operating topology graph is defined as a heterogeneous graph, and the meta-path set is defined according to the business characteristics of the power distribution equipment. The meta-path set is used to obtain the neighbor set of the power supply node under each meta-path, so as to transform the heterogeneous graph into a multi-relationship graph based on meta-paths.

[0025] The power supply nodes are projected into the feature subspace of the graph convolutional network to obtain the influencing factor index, and the influencing factor index is assigned to each meta-path. The neighbor information of the power supply node under each meta-path is calculated by the mean aggregation function, and the global power supply quality information of each meta-path is extracted by the readout function.

[0026] The global neighbor information is obtained by calculating the correlation between neighbor information and global power quality information, and removing neighbor information that is less than the threshold in relation to global power quality information based on the correlation.

[0027] The global neighbor information is projected onto the same feature space to calculate the weight coefficients of the meta-path, and combined with the power supply node information to generate a weighted fusion node representation through an activation function, thus obtaining aggregated neighbor information.

[0028] Preferably, the method of selecting influencing factor characteristic indicators based on the characteristic indicator pool to determine the indicator combination used to construct the source-load balance analysis model, and constructing the source-load balance analysis model using the indicator combination and elastic network fitting technology includes:

[0029] Pearson linear correlation coefficient, Kendall rank correlation coefficient and Spearman rank correlation coefficient were used to quantify the degree of correlation between the independent variable indicators and the dependent variable indicators in the feature indicator pool, respectively.

[0030] By comparing the correlation degree with the selection criteria, key characteristic indicators affecting the source-load balance are screened, effective indicator combinations for constructing the source-load balance analysis model are determined, and the differences in the dimensions and dispersion of the effective indicator combinations are judged.

[0031] The effective index combination is preprocessed based on the differences in dimensionality and dispersion, and a source-load balance analysis model is constructed based on the preprocessed effective index combination using ridge regression and lasso regression algorithms.

[0032] By solving the optimization problem with quadratic programming, the elastic network estimate of the correlation coefficients between the independent variable index and the dependent variable index in the source-load balance analysis model is calculated, and the source-load balance analysis results are obtained.

[0033] Preferably, based on the power supply frequency fluctuation value, the duration and fault frequency of the power distribution equipment being in a state of supply-demand imbalance are determined, and the backpropagation neural network model is iteratively trained to output the performance evaluation values ​​of the power distribution network equipment, including:

[0034] The offset amplitude of power supply frequency fluctuations is extracted from historical operating data, and a judgment standard for supply and demand imbalance is defined based on the offset amplitude. The amplitude of power supply frequency fluctuations is analyzed through the judgment standard.

[0035] The duration and fault frequency of power supply frequency fluctuations of power distribution equipment are analyzed based on the trend of amplitude changes, and the power distribution equipment is determined to be in a state of supply and demand imbalance when the duration and fault frequency exceed the threshold.

[0036] A sliding window structure is generated using window partitioning technology to perform dependency analysis on duration and failure frequency, and a model training dataset is generated based on the dependency analysis results using incremental learning.

[0037] The model training dataset is used as input to train the backpropagation neural network model. The backpropagation neural network model is then used to output the estimated duration and fault frequency of the power distribution equipment in the state of supply and demand imbalance, and output the performance evaluation value of the power distribution network equipment.

[0038] Preferably, the real-time operating data of the power distribution equipment is combined with frequency calculation technology to simulate the future operating state of the power distribution equipment. Based on the simulation results, the load safety results of the power distribution equipment in the real-time state and in the future time period are output. Based on the load safety results, a power distribution equipment monitoring and early warning scheme is matched and executed, including:

[0039] The system acquires real-time operating data of power distribution equipment and uses frequency dynamic calculation technology to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data, thereby generating random scenarios for the future operation of the power distribution equipment.

[0040] Based on Monte Carlo simulation technology, the random scenarios of power distribution equipment are simulated to simulate the state performance of power distribution equipment under different loads and environmental fluctuation factors, and to evaluate the performance evaluation value of power distribution equipment.

[0041] Based on real-time operating data and performance evaluation values, the system outputs the load safety results of power distribution equipment in real-time and in future time periods, matches the power distribution equipment monitoring and early warning scheme, and executes early warning processing of power distribution equipment based on the matching results.

[0042] Preferably, real-time operating data of the power distribution equipment is acquired, and frequency dynamic calculation technology is used to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data to generate random scenarios for the future operation of the power distribution equipment, including:

[0043] The real-time operating data of the power distribution equipment is collected by sensors within the target time period. The moving average filtering technique is used to denoise and normalize the real-time operating data to complete the data preprocessing.

[0044] Based on real-time operational data and the specifications and structural parameters of power distribution equipment, a digital twin containing the structure, operating parameters, environmental variables, and operating logic of the power distribution equipment is constructed using dynamic modeling.

[0045] The mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. The operating characteristics of the power distribution equipment are combined to divide the state set and simulate the operating state probability of the power distribution equipment.

[0046] Based on the operating state probability, a Markov chain is used to construct the transition probability matrix for each operating state, and the transition probability matrix is ​​used to generate random operating scenarios for power distribution equipment in the future prediction period.

[0047] Preferably, the mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. Then, based on the operating characteristics of the power distribution equipment, a state set is defined to simulate the operating state probabilities of the power distribution equipment, including:

[0048] By performing state mapping and behavior simulation mapping, the real-time synchronization of the physical entity of the power distribution equipment to the digital twin is completed, and the ability to replicate the real-time operating status of the digital twin is added, resulting in a mature digital twin.

[0049] Based on the operating characteristics of power distribution equipment, a set of states is divided, and time windows are divided. The frequency of real-time operating data corresponding to each operating state in the mature digital twin within each time window is counted, and the probability of the operating state of the power distribution equipment is determined by the frequency.

[0050] Secondly, the present invention also provides a real-time monitoring and early warning system for the status of distribution network equipment, the system comprising:

[0051] The standard setting module is used to analyze the load-bearing capacity of distribution network equipment according to monitoring standards, and to establish load safety early warning levels under uncertain scenarios based on the load-bearing capacity.

[0052] The frequency fluctuation analysis module is used to establish a source-load balance analysis model based on historical operating data to describe the balance between energy supply and energy consumption of power distribution equipment, and to randomly sample the source-load balance analysis model to output the power supply frequency fluctuation value of the power distribution equipment when it is in a state of supply-demand imbalance.

[0053] The performance evaluation module is used to determine the duration and fault frequency of power distribution equipment in a state of supply and demand imbalance based on the power supply frequency fluctuation value, iteratively trains the backpropagation neural network model, and outputs the performance evaluation value of the power distribution equipment.

[0054] The early warning scheme output module is used to combine performance evaluation values ​​with load safety early warning levels to output early warning schemes for power distribution equipment.

[0055] The real-time monitoring output module is used to combine the real-time operating data of the power distribution equipment with frequency calculation technology to simulate the future operating status of the power distribution equipment. Based on the simulation results, it outputs the load safety results of the power distribution equipment in the real-time state and in the future period, and matches and executes the power distribution equipment monitoring and early warning scheme based on the load safety results.

[0056] The beneficial effects of this invention are as follows:

[0057] 1. This invention analyzes the load-bearing capacity of distribution network equipment based on monitoring standards, establishes load safety early warning levels under uncertain scenarios, and transforms traditional static threshold judgment into a risk-based hierarchical management and control method. Based on historical operating data, a source-load balance analysis model is constructed to fully reflect the real operating characteristics of power distribution equipment under different load disturbances and power supply fluctuations. At the same time, the duration of supply-demand imbalance and the frequency of faults are quantified by the power supply frequency fluctuation value, and the backpropagation neural network model is iteratively trained. This makes the performance evaluation results not only reflect the current state, but also implicitly include the equipment's sensitivity to long-term operating pressure and potential fault risks, significantly improving the precision and predictive ability of monitoring and early warning. In this way, it can identify potential load safety risks in the future before the current state deteriorates, thereby improving the safety and reliability of distribution network equipment operation.

[0058] 2. This invention utilizes historical operational data to generate an operational topology diagram and abstracts equipment and power supply lines as nodes and edges of the graph. This allows for a clear display of the interaction between power flow and energy transmission issues in the distribution network. Furthermore, it employs elastic network fitting technology to establish a source-load balance analysis model, which can better describe subtle changes in energy supply and consumption in the power grid. This approach changes the reliance on fixed parameters and static assumptions in traditional methods, thereby helping to identify potential supply-demand imbalances in advance, reducing the probability of power grid failures, and improving the stability and reliability of the power grid.

[0059] 3. This invention uses frequency dynamic calculation technology to transform real-time operating data into a probability distribution of equipment operating status. Combined with Monte Carlo simulation technology, multiple possible operating scenarios are generated through random sampling, thereby simulating the performance of power distribution equipment under different load fluctuations and environmental changes. This allows for the evaluation of its performance under various operating conditions, reflecting the operating status of the equipment under extreme or uncertain conditions, and predicting load safety in future periods, thus identifying potential supply and demand imbalances or equipment failure risks in advance. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of a method for real-time monitoring and early warning of the status of distribution network equipment according to an embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of a real-time monitoring and early warning system for the status of distribution network equipment according to an embodiment of the present invention.

[0063] In the picture:

[0064] 1. Standard setting module; 2. Frequency fluctuation analysis module; 3. Performance evaluation module; 4. Early warning scheme output module; 5. Real-time monitoring output module. Detailed Implementation

[0065] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention.

[0066] According to an embodiment of the present invention, a method and system for real-time monitoring and early warning of the status of distribution network equipment are provided.

[0067] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, the real-time monitoring and early warning method for the status of distribution network equipment according to an embodiment of the present invention includes:

[0068] Step S1: Analyze the load-bearing capacity of the distribution network equipment according to the monitoring standards, and establish a load safety early warning level under uncertain scenarios based on the load-bearing capacity.

[0069] In one embodiment, establishing load safety warning levels under uncertain scenarios requires a comprehensive review of relevant industry standards, factory technical parameters, historical operation and maintenance specifications, and customized requirements for regional distribution network operation. This includes clarifying the benchmark requirements for different types of distribution network equipment, such as transformers, switchgear, and cables, in key indicators like rated voltage, rated current, temperature rise threshold, and insulation strength. Simultaneously, it involves integrating operational data, fault records, and environmental impact data throughout the entire lifecycle of the distribution network equipment. Furthermore, based on the reviewed monitoring standards, a multi-dimensional analysis method is employed to accurately assess load capacity. This includes real-time acquisition of data such as current, voltage, and power factor through electrical parameter detection; analysis of the matching degree between the current load and rated load of the distribution network equipment; evaluation of the equipment's thermal tolerance limits under different loads through thermal imaging detection and temperature rise model calculations; and assessment of the equipment's insulation withstand capability during load changes through insulation performance testing.

[0070] Simultaneously considering the impact of factors such as equipment aging and operation and maintenance quality on load-bearing capacity attenuation, a baseline value and attenuation curve for the load-bearing capacity of the equipment under normal operating conditions are ultimately formed. Furthermore, various uncertain scenarios in the operation of distribution network equipment are defined, mainly covering sudden changes in environmental parameters caused by extreme weather, sudden load increases caused by regional load peaks, load transfers caused by grid fault cascading reactions, and load fluctuations brought about by new energy grid connection. The variation patterns, fluctuation amplitudes, and durations of distribution network equipment load under different uncertain scenarios are analyzed, clarifying the impact weight and mechanism of each scenario on the equipment load-bearing capacity. Based on the baseline value of equipment load-bearing capacity, Analysis of attenuation curves and the impact of different uncertain scenarios establishes a multi-level load safety early warning system, clarifying the triggering conditions for each early warning level, such as normal load range, early warning load threshold, dangerous load threshold, judgment indicators, and quantitative standards. For example, setting the equipment load below 80% of its capacity as a safe level, 80% to 90% as a yellow warning level, 90% to 100% as an orange warning level, and exceeding 100% or experiencing a sudden increase in load that lasts for a certain period as a red warning level. By defining the impact of uncertain scenarios in advance, the early warning system becomes more forward-looking and adaptable, capable of coping with complex and ever-changing distribution network operating environments.

[0071] Step S2: Based on historical operating data, establish a source-load balance analysis model to describe the balance between energy supply and energy consumption of power distribution equipment, and randomly sample the source-load balance analysis model to output the power supply frequency fluctuation value of the power distribution equipment when it is in a state of supply-demand imbalance.

[0072] In one embodiment, a source-load balance analysis model is established based on historical operating data to describe the balance between energy supply and energy consumption of power distribution equipment. Random sampling is performed on the source-load balance analysis model, and the output power supply frequency fluctuation values ​​of the power distribution network equipment when it is in a state of supply-demand imbalance include:

[0073] Obtain historical operational data containing the operating load and power supply capacity of power distribution equipment, and generate an operational topology graph by using the power supply line connections between power distribution equipment as edges and the historical operational data as nodes.

[0074] Analyze the operational topology to obtain the balance relationship between energy supply and energy consumption during power transmission, and use elastic network fitting technology to establish a source-load balance analysis model based on the balance relationship;

[0075] Based on variational inference technology, different energy supply and energy consumption data are extracted from the source-load balance analysis model as sampling points to simulate the power supply frequency fluctuation state of power distribution equipment under power supply fluctuation conditions.

[0076] In one embodiment, analyzing the running topology to obtain the balance relationship between energy supply and energy consumption during power transmission, and establishing a source-load balance analysis model based on the balance relationship using elastic network fitting technology includes:

[0077] The running topology graph is transformed into a multi-relationship graph based on meta-paths. Power quality relationship selection graph convolutional layers are used to analyze the multi-relationship graph to obtain aggregated neighbor information in order to model the global correlation between meta-paths.

[0078] The aggregated neighbor information is filtered to distinguish the importance of each power quality relationship, and the power quality relationships that meet the importance threshold are retained as the energy supply and demand balance changes of the power distribution equipment under different operating conditions;

[0079] By using the operating time and maintenance information of power distribution equipment as independent variables and the changes in energy supply and demand balance as dependent variables, a pool of characteristic indicators is obtained for constructing the source-load balance analysis model.

[0080] Based on the feature index pool, the influencing factor feature indexes are screened to determine the index combination used to construct the source-load balance analysis model. The source-load balance analysis model is constructed by using the index combination and elastic network fitting technology.

[0081] In one embodiment, the running topology graph is transformed into a multi-relationship graph based on meta-paths. Power quality relationship selection graph convolutional layers are used to analyze the multi-relationship graph to obtain aggregated neighbor information, in order to model the global correlation between meta-paths, including:

[0082] Based on the power transmission relationship between power distribution equipment and the power supply node, the operating topology graph is defined as a heterogeneous graph, and the meta-path set is defined according to the business characteristics of the power distribution equipment. The meta-path set is used to obtain the neighbor set of the power supply node under each meta-path, so as to transform the heterogeneous graph into a multi-relationship graph based on meta-paths.

[0083] The power supply nodes are projected into the feature subspace of the graph convolutional network to obtain the influencing factor index, and the influencing factor index is assigned to each meta-path. The neighbor information of the power supply node under each meta-path is calculated by the mean aggregation function, and the global power supply quality information of each meta-path is extracted by the readout function.

[0084] The global neighbor information is obtained by calculating the correlation between neighbor information and global power quality information, and removing neighbor information that is less than the threshold in relation to global power quality information based on the correlation.

[0085] The global neighbor information is projected onto the same feature space to calculate the weight coefficients of the meta-path, and combined with the power supply node information to generate a weighted fusion node representation through an activation function, thus obtaining aggregated neighbor information.

[0086] In one embodiment, the method of determining the combination of influencing factor characteristic indicators based on the characteristic indicator pool to construct the source-load balance analysis model, and constructing the source-load balance analysis model using the indicator combination and elastic network fitting technology includes:

[0087] Pearson linear correlation coefficient, Kendall rank correlation coefficient and Spearman rank correlation coefficient were used to quantify the degree of correlation between the independent variable indicators and the dependent variable indicators in the feature indicator pool, respectively.

[0088] By comparing the correlation degree with the selection criteria, key characteristic indicators affecting the source-load balance are screened, effective indicator combinations for constructing the source-load balance analysis model are determined, and the differences in the dimensions and dispersion of the effective indicator combinations are judged.

[0089] The effective index combination is preprocessed based on the differences in dimensionality and dispersion, and a source-load balance analysis model is constructed based on the preprocessed effective index combination using ridge regression and lasso regression algorithms.

[0090] By solving the optimization problem with quadratic programming, the elastic network estimate of the correlation coefficients between the independent variable index and the dependent variable index in the source-load balance analysis model is calculated, and the source-load balance analysis results are obtained.

[0091] It needs to be explained that in the process of outputting the power supply frequency fluctuation value of the distribution network equipment when it is in a state of supply and demand imbalance, it is necessary to perform historical operation data collection and preprocessing. The data scope is clearly defined as the operating load data of core power distribution equipment such as 10kV transformers and 10kV switchgear in the distribution network system for the past 3 years. For example, active power, reactive power and power supply capacity data are collected every 15 minutes, such as generator output power, grid input power and line transmission capacity. At the same time, the connection relationship data of power supply lines between power distribution equipment are collected, including line type, length, impedance, etc. The collected historical data is cleaned, and an operation topology is constructed based on the processed data: each power distribution equipment and power supply node is regarded as a node, the power supply line connection between equipment is regarded as an edge, the edge weight is assigned to the line transmission capacity, and the node attributes are assigned to the preprocessed historical operation data. For example, the attributes of node 1 10kV transformer include the load power every 15 minutes from 2022 to 2024, the power supply capacity of the line it is on, etc.

[0092] The operational topology graph is transformed into a multi-relationship graph based on meta-paths. Since the operational topology graph contains various types of nodes such as equipment, lines, and power supply nodes, as well as various relationships such as power transmission and power supply support, it is a heterogeneous graph. Therefore, a meta-path set needs to be defined according to the business characteristics of the power distribution equipment. Examples include power supply node, line, power distribution equipment, line, power consumption node; power distribution equipment, maintenance node, power distribution equipment, etc. The neighbor set of each power supply node under different meta-paths is obtained through the meta-path set, completing the transformation from a heterogeneous graph to a multi-relationship graph. Simultaneously, utilizing power supply quality relationships, including voltage deviation, frequency stability, and power supply reliability, an appropriate graph convolutional layer (GCN layer) is selected to project the power supply nodes into the feature subspace of the graph convolutional network, obtaining influencing factor indicators such as voltage, current, and power. Each metapath is assigned a corresponding influencing factor index. The neighbor information of power supply nodes under each metapath is calculated using a mean aggregation function. For example, the mean load information of neighboring equipment under the power supply node, line, and distribution equipment metapaths for a given power supply node is calculated. Simultaneously, readout functions, such as summation functions, are used to extract global power quality information for each metapath, such as the average power supply frequency of the area covered by that metapath. The Pearson correlation between neighbor information and global power quality information is calculated, and neighbor information with a correlation below 0.3 (the threshold is set according to power quality requirements) is removed to obtain global neighbor information. This global neighbor information is projected onto the same feature space, and the weight coefficients of each metapath are calculated using a softmax function. Combined with the power supply node's own information, a ReLU activation function is applied. The system generates weighted fusion node representations to obtain aggregated neighbor information, enabling the modeling of global correlations between metapaths. The aggregated neighbor information is then filtered, and a random forest algorithm is used to calculate the importance of each power quality relationship to the energy supply and demand balance. An importance threshold of 0.2 is set, retaining power quality relationships with importance higher than this threshold, such as load power fluctuations and power supply stability, to characterize the changes in energy supply and demand balance of power distribution equipment under different operating conditions. The system then uses the operating time and maintenance information of the power distribution equipment as independent variables, and the changes in energy supply and demand balance, such as the supply-demand difference and balance duration, as dependent variables to construct a feature index pool. Finally, mutual information entropy is used to filter out influencing factors with a correlation higher than 0.4 with the dependent variable to determine the influencing factors. The system combines indicators and utilizes elastic network fitting technology to construct a source-load balance analysis model with L1 regularization coefficient α=0.1 and L2 regularization coefficient β=0.01. Simultaneously, it employs variational inference technology to randomly sample the source-load balance analysis model, setting the sampling number to 1000 times. Different energy supply and energy consumption data are extracted from the model's characteristic indicator distribution as sampling points, such as extreme scenario data like a 20% drop in power supply or a 15% increase in load power. These data are then substituted into the model to simulate the power supply frequency fluctuation state of the distribution equipment under power supply fluctuation conditions. The output value for each sampling is the power supply frequency fluctuation value, such as ±0.5Hz or ±0.8Hz, thus accurately capturing global factors affecting power supply quality and improving the comprehensiveness of the balance relationship analysis.

[0093] Assuming a 10kV distribution network system in a certain urban area is taken as the research object, it covers 10kV transformers numbered T1-T5, power supply nodes numbered P1-P3, power supply lines numbered L1-L8, and 20 power consumption nodes. The monitoring period is from January 1, 2022 to December 31, 2024. The core objective is to construct a source-load balance analysis model and accurately output the power supply frequency fluctuation value when supply and demand are unbalanced.

[0094] The historical operational data comes from the data acquisition and monitoring system of the distribution network dispatch center and the power production management system. Specifically, it includes: operational data collected by the data acquisition and monitoring system every 15 minutes from 2022 to 2024, totaling 35,040 time slices. Each slice contains 12 indicators, including the active power, reactive power, current, and voltage of 5 transformers, the output power of 3 power supply nodes, the transmission power of 8 lines, and voltage loss, as well as the operation and maintenance data collected by the power production management system, including maintenance records (maintenance time, maintenance content) and fault records (fault time, fault type, and repair time) for each piece of equipment.

[0095] A hybrid algorithm architecture combining Graph Convolutional Networks (GCN) and Elastic Networks is adopted. The overall structure consists of three core layers: feature extraction layer, feature fusion layer, and model fitting layer. The specific details are as follows:

[0096] The core components of the feature extraction layer include two GCN layers, namely GCN1 and GCN2, with an input dimension of 12 and output dimensions of 8 and 6 respectively. In the inter-layer connection method, the input of GCN1 is the node feature vector (12-dimensional), while the output of GCN1 is directly used as the input of GCN2. Thus, local and global features of nodes can be extracted from the multi-relationship graph to generate a 6-dimensional feature vector.

[0097] The core components of the feature fusion layer include one fully connected layer with 6 input dimensions and 4 output dimensions. The input of the fully connected layer is the output feature vector of GCN2. It is connected to the bias term through the weight matrix, which can reduce the dimensionality of the extracted feature vector and fuse it to select core features.

[0098] The core component of the model fitting layer is the elastic network layer, which has an input dimension of 4 and an output dimension of 1. The input of the elastic network layer is the output feature vector of the fully connected layer, and it directly outputs the energy supply and demand balance value. In this way, the quantitative relationship between the core features and the energy supply and demand balance can be established, and the model fitting can be completed.

[0099] Pearson linear correlation coefficient, Kendall rank correlation coefficient, Spearman rank correlation coefficient, and monotonic correlation coefficient were used to quantify the correlation between the eight independent variables in the feature index pool: summer peak period, winter peak period, load power, power supply (kW), maintenance status in the past 3 months, number of faults in the past year, line transmission loss (kW), and ambient temperature and the dependent variable index.

[0100] In setting the selection criteria, the absolute values ​​of all three correlation coefficients were ≥0.3, and at least two coefficients had absolute values ​​≥0.4. The correlation results of each indicator were compared with the selection criteria to screen key characteristic indicators. At the same time, the coefficient of variation was used to judge the difference in dimensionality and dispersion of effective indicator combinations. A coefficient of variation ≥0.2 was considered to have significant dispersion differences. For the selected effective indicator combinations, if there were differences in dimensionality, Z-score standardization was used, and if there were significant dispersion differences, normalization was used. Based on the preprocessed effective indicator combinations, an elastic network source-load balance analysis model was constructed that integrates ridge regression (L2 regularization) and lasso regression (L1 regularization).

[0101] The parameter solving of the elastic network model is transformed into an optimization problem with quadratic programming. The optimization problem is solved by the interior point method to obtain the elastic network estimate of the correlation coefficients of each effective independent variable index with the dependent variable index, and then the source load balance analysis results are output.

[0102] A composite loss function combining mean squared error, L1 regularization, and L2 regularization is adopted. While ensuring the model fitting accuracy, L1 regularization enables feature selection, and L2 regularization suppresses excessive weights to avoid model overfitting. The Adam optimization algorithm is used to optimize the source-load balance analysis model. The initial learning rate is 0.001, the momentum parameter is 0.9, the second-order momentum parameter is 0.999, and the numerical stability parameter is 1e-8. During training, a learning rate decay strategy is adopted. When the validation set loss does not decrease for 5 consecutive epochs, the learning rate is decayed to 0.5. The source-load balance analysis model is output based on the training results.

[0103] Step S3: Determine the duration and fault frequency of the power distribution equipment when it is in a state of supply and demand imbalance based on the power supply frequency fluctuation value, iteratively train the backpropagation neural network model, and output the performance evaluation value of the power distribution equipment.

[0104] In one embodiment, the duration and fault frequency of power distribution equipment being in a state of supply-demand imbalance are determined based on power supply frequency fluctuation values. A backpropagation neural network model is iteratively trained, and the output performance evaluation values ​​of the power distribution equipment include:

[0105] The offset amplitude of power supply frequency fluctuations is extracted from historical operating data, and a judgment standard for supply and demand imbalance is defined based on the offset amplitude. The amplitude of power supply frequency fluctuations is analyzed through the judgment standard.

[0106] The duration and fault frequency of power supply frequency fluctuations of power distribution equipment are analyzed based on the trend of amplitude changes, and the power distribution equipment is determined to be in a state of supply and demand imbalance when the duration and fault frequency exceed the threshold.

[0107] A sliding window structure is generated using window partitioning technology to perform dependency analysis on duration and failure frequency, and a model training dataset is generated based on the dependency analysis results using incremental learning.

[0108] The model training dataset is used as input to train the backpropagation neural network model. The backpropagation neural network model is then used to output the estimated duration and fault frequency of the power distribution equipment in the state of supply and demand imbalance, and output the performance evaluation value of the power distribution network equipment.

[0109] It should be explained that this embodiment extracts the original power supply frequency data corresponding to the frequency fluctuation value from the historical operating data stored in the distribution network system's SCADA system and PMS system. It calculates the offset of the frequency fluctuation value relative to the rated frequency (50Hz) under each time slice. At the same time, it combines the distribution network equipment operation procedures and historical fault case data to statistically analyze the probability of equipment failure under different offset amplitudes. Based on this, it defines the supply and demand imbalance judgment standard. For example, it sets the offset amplitude ≥0.5% as the first-level imbalance warning threshold, ≥1.0% as the second-level imbalance warning threshold, and ≥2.0% as the severe imbalance threshold.

[0110] Based on this judgment standard, amplitude analysis is performed on the power supply frequency fluctuation value. The duration of imbalance and fault frequency are analyzed based on the amplitude change trend to determine the imbalance state. First, time-series trend fitting is performed on the dataset with frequency fluctuation offsets after screening according to the judgment standard. Linear interpolation is used to fill in missing values ​​in the time-series data. A sliding window is used to traverse and identify continuous time segments where the frequency fluctuation amplitude is consistently above the corresponding imbalance threshold. The duration of this continuous time segment is the duration of the supply-demand imbalance. Simultaneously, fault records of the power distribution equipment within the corresponding time segment are extracted from the PMS system, including fault occurrence time and fault type. The single fault value is then calculated. The number of faults occurring within a given time period is the fault frequency. Simultaneously, duration thresholds are set in conjunction with distribution network equipment operation and maintenance standards, such as a 30-minute threshold for Level 1 imbalance, 15 minutes for Level 2, and 5 minutes for severe imbalance. These are combined with fault frequency thresholds (e.g., a fault frequency ≥ 0.5 times / hour). When the duration exceeds the corresponding threshold or the fault frequency exceeds the threshold, the distribution equipment can be determined to be in a state of supply and demand imbalance. Thus, by combining the two core parameters of duration and fault frequency, the severity of the imbalance state of the distribution equipment can be comprehensively reflected, providing key label data for subsequent model training.

[0111] A windowing technique and incremental learning are used to generate the model training dataset. First, a sliding window structure is designed based on the continuous characteristics of time series. The window size is set to contain 20 time slices, and the window sliding step size is set to 5 time slices. This sliding window structure is used to slide and truncate the duration data and fault frequency data that are determined to be in an imbalanced state, and obtain the duration change sequence, fault frequency change sequence and the imbalance state label at the end of the corresponding window within each window. At the same time, dependency analysis is performed on the data within the window to calculate the Pearson correlation coefficient between the duration and fault frequency of adjacent windows and identify the temporal dependency relationship between the data. Then, an incremental learning approach is used to generate the model training dataset. That is, an initial training dataset is first built based on historical imbalance data, and then the duration and fault frequency data corresponding to the new running data are added in real time. The dataset is updated through incremental learning, and historical outdated data with large differences from the distribution of the new data are removed simultaneously.

[0112] The backpropagation neural network model is iteratively trained to output the performance evaluation value of the distribution network equipment. First, the backpropagation neural network model architecture is constructed. The number of input layer nodes corresponds to the feature dimensions within the sliding window, such as 20 dimensions for duration change sequence and 20 dimensions for fault frequency change sequence, for a total of 40 input nodes. Two hidden layers are set, with 32 nodes in the first layer and 16 nodes in the second layer. The ReLU function is used as the activation function to solve the gradient vanishing problem. The output layer is set with 3 nodes, corresponding to the intermediate parameters of the duration prediction value, fault frequency prediction value, and performance evaluation value, respectively. The model training dataset generated above is divided into training set and validation set in a 7:3 ratio. The mean squared error is used as the loss function, and the Adam optimization algorithm is used to iteratively train the model. During the training process, the validation set is used for validation every 50 epochs. When the validation set loss does not decrease for 10 consecutive epochs, the learning rate decay strategy is adopted until the model converges (the convergence threshold is set to loss value ≤ 1e-5).

[0113] After training, new duration and fault frequency related data are input into the model. The model first outputs the estimated values ​​of duration and fault frequency, and then calculates and outputs the performance evaluation value of the distribution network equipment based on the preset performance evaluation standard. Then, the backpropagation neural network model is used to achieve accurate prediction of imbalance characteristic parameters and quantitative evaluation of equipment performance.

[0114] Step S4: Combine the performance evaluation value with the load safety warning level to output a warning scheme for power distribution equipment.

[0115] It should be explained that in the process of the power distribution equipment early warning scheme, the quantitative score of the current equipment load safety early warning level is recorded as A, and the quantitative score corresponding to the performance evaluation value is recorded as B. Substituting these into the fusion calculation formula: Comprehensive Early Warning Score = A × 0.6 + B × 0.4, the comprehensive early warning level is determined based on the comprehensive early warning score. It is assumed that a comprehensive score ≥ 3.6 is a safety early warning, 2.8 to 3.6 is a yellow early warning, 2.0 to 2.8 is an orange early warning, and < 2.0 is a red early warning. At the same time, the comprehensive early warning level is calibrated by combining auxiliary parameters such as power supply frequency fluctuation amplitude, imbalance duration, and fault frequency. For example, when the comprehensive score is 3.0 and it is in a yellow early warning, but the power supply frequency fluctuation amplitude is ≥ 2.0% and it is in a severe imbalance, the early warning level is upgraded to orange; when the comprehensive score is 2.2 and it is in an orange early warning, but the fault frequency is < 0.2 times / hour and it is below the threshold, the early warning level is downgraded to yellow. During the calibration process, the operation and maintenance procedures of the distribution network equipment and historical early warning handling cases are referenced to ensure the rationality of the calibration results.

[0116] Based on the calibrated comprehensive early warning level, targeted early warning plans are developed. For different comprehensive early warning levels, the format of early warning information output, the responsible parties for handling, the handling process, and the time limits are clearly defined. For example, a safety early warning only requires generating a routine operation record in the distribution network dispatch system, without on-site handling; a yellow early warning requires sending a reminder message to the maintenance team through the dispatch system, requiring equipment status inspection to be completed within 24 hours; an orange early warning requires immediately sending an early warning notification to the maintenance team and the dispatch center, requiring on-site special testing and load adjustment measures to be taken within 6 hours; a red early warning requires activating the emergency response mechanism, pushing early warning information to the emergency command center, maintenance team, and power company management at multiple levels, and finally outputting an early warning plan and synchronizing it to the distribution network dispatch system.

[0117] Step S5: Combine the real-time operating data of the power distribution equipment with frequency calculation technology to simulate the future operating state of the power distribution equipment. Based on the simulation results, output the load safety results of the power distribution equipment in the real-time state and in the future period. Based on the load safety results, match and execute the power distribution equipment monitoring and early warning scheme.

[0118] In one embodiment, the real-time operating data of the power distribution equipment is combined with frequency calculation technology to simulate the future operating state of the power distribution equipment. Based on the simulation results, the load safety results of the power distribution equipment in the real-time state and in the future time period are output. Based on the load safety results, a power distribution equipment monitoring and early warning scheme is matched and executed, including:

[0119] The system acquires real-time operating data of power distribution equipment and uses frequency dynamic calculation technology to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data, thereby generating random scenarios for the future operation of the power distribution equipment.

[0120] Based on Monte Carlo simulation technology, the random scenarios of power distribution equipment are simulated to simulate the state performance of power distribution equipment under different loads and environmental fluctuation factors, and to evaluate the performance evaluation value of power distribution equipment.

[0121] Based on real-time operating data and performance evaluation values, the system outputs the load safety results of power distribution equipment in real-time and in future time periods, matches the power distribution equipment monitoring and early warning scheme, and executes early warning processing of power distribution equipment based on the matching results.

[0122] In one embodiment, acquiring real-time operating data of the power distribution equipment and using frequency dynamic calculation technology to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data to generate random scenarios for the future operation of the power distribution equipment includes:

[0123] The real-time operating data of the power distribution equipment is collected by sensors within the target time period. The moving average filtering technique is used to denoise and normalize the real-time operating data to complete the data preprocessing.

[0124] Based on real-time operational data and the specifications and structural parameters of power distribution equipment, a digital twin containing the structure, operating parameters, environmental variables, and operating logic of the power distribution equipment is constructed using dynamic modeling.

[0125] The mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. The operating characteristics of the power distribution equipment are combined to divide the state set and simulate the operating state probability of the power distribution equipment.

[0126] Based on the operating state probability, a Markov chain is used to construct the transition probability matrix for each operating state, and the transition probability matrix is ​​used to generate random operating scenarios for power distribution equipment in the future prediction period.

[0127] In one embodiment, a mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. The operating characteristics of the power distribution equipment are then combined to divide the state set, simulating the probability of the power distribution equipment's operating state, including:

[0128] By performing state mapping and behavior simulation mapping, the real-time synchronization of the physical entity of the power distribution equipment to the digital twin is completed, and the ability to replicate the real-time operating status of the digital twin is added, resulting in a mature digital twin.

[0129] Based on the operating characteristics of power distribution equipment, a set of states is divided, and time windows are divided. The frequency of real-time operating data corresponding to each operating state in the mature digital twin within each time window is counted, and the probability of the operating state of the power distribution equipment is determined by the frequency.

[0130] It should be explained that this embodiment collects and preprocesses real-time operating data of the power distribution equipment. Sensors deployed at key parts of the equipment collect real-time operating data within a target time period and perform noise reduction processing. Based on the preprocessed real-time operating data, combined with the specifications and structural parameters of the power distribution equipment, a digital twin is constructed using a dynamic modeling approach that integrates SolidWorks 3D modeling and MATLAB / Simulink simulation modeling. The model includes equipment structure modules, operating parameter modules, environmental variable modules, and operating logic modules. Then, a mapping process between the digital twin and the physical entity is performed, including state mapping, synchronizing the real-time operating parameters of the physical entity to the corresponding modules in the digital twin. The frequency is mapped to behavioral simulation at 1 minute / time. Based on the operating rules of physical entities, the digital twin is endowed with the ability to replicate real-time operating states, such as simulating the temperature rise process when the transformer load changes. By repeatedly calibrating and adjusting the model parameters, the error between the digital twin output and the measured data of the physical entity is ≤3%, resulting in a mature digital twin. Then, combined with the operating characteristics of the power distribution equipment, the state set is divided, such as normal operating state, light load state, heavy load state, minor fault state, and severe fault state. Time windows are divided to count the frequency of real-time operating data corresponding to each operating state in the mature digital twin within each time window. The probability of an operating state is calculated using the frequency (probability of a state = frequency of data corresponding to that state / total amount of data in the window).

[0131] Based on the probability of the operating state, a transition probability matrix is ​​constructed between each operating state. The state space of the Markov chain is divided into 5 operating states. By statistically analyzing the number of transitions of each state within adjacent time windows in historical operating data, the transition probability is calculated to obtain the transition probability matrix. The future prediction period is set to 24 hours, and the prediction period is divided into 144 windows according to the time window. Based on the transition probability matrix, a random sampling method is used to generate the operating state within each prediction window, and then combine them to form multiple random operating scenarios of the power distribution equipment in the next 24 hours. Thus, the Markov chain can effectively capture the temporal transition characteristics of the operating state, and the generated random scenarios cover a variety of possible operating conditions.

[0132] Monte Carlo simulation technology is used to simulate and evaluate random scenarios. 1,000 generated future random operating scenarios are input into a mature digital twin. Based on Monte Carlo simulation technology, the equipment status performance under different loads and environmental fluctuations is simulated. During the simulation, key parameters of the equipment in each scenario are collected in real time. Combined with the backpropagation neural network model, the simulation parameters are input to calculate the equipment performance evaluation value in each scenario. At the same time, the load safety value distribution of each time period in the next 24 hours of the 1,000 simulation scenarios is statistically analyzed, and the load safety results in the future time period are output. Based on the real-time and future load safety results, a power distribution equipment monitoring and early warning scheme is matched to execute early warning processing.

[0133] Assuming that the No. 1 10kV transformer (model S11-1000 / 10, rated capacity of 1000kVA, rated voltage of 10kV / 0.4kV, and rated current of 57.7A / 1443A) in a 10kV power distribution system of a certain urban area is taken as the research object, the monitoring target is the load safety status in real time and the next 24 hours (e.g., 00:00-24:00 on August 1, 2025), and the corresponding early warning scheme is matched and executed.

[0134] The training data includes: historical operating data of transformer No. 1 from May 1, 2025 to July 31, 2025, a total of 92 days, collected daily at a 1-minute granularity, including 15 indicators such as winding temperature, three-phase current, and three-phase voltage, totaling 32,480 data points; operating status records within the same time period, which were manually labeled and matched with equipment fault records to obtain the number of transitions for each state; and 20,000 data points containing operating parameters and corresponding performance evaluation values ​​were extracted from the historical data as a fine-tuning dataset.

[0135] The test data uses real-time operating data (60 records) of transformer No. 1 from 00:00 to 01:00 on August 1, 2025. Moving average filtering (window size 5) and Z-score standardization are applied to these 60 real-time data records. Based on the preprocessed data, the initial probabilities of five operating states are statistically obtained. Assuming normal operation (D1) is 65%, light load (D2) is 20%, heavy load (D3) is 10%, minor fault (D4) is 3%, and severe fault (D5) is 2%, a transition probability matrix is ​​constructed. For example, D1 to D5 is 80%, D1 to D2 is 15%, D1 to D3 is 4%, D1 to D4 is 1%, and D1 to D5 is 0%. 1000 random scenarios for the next 24 hours are generated, each covering different combinations of load fluctuations and environmental changes. The test scenarios include three typical future operating conditions, as shown in Table 1.

[0136] Table 1 Typical Future Operating Conditions

[0137]

[0138] Three typical test scenarios and 1000 random scenarios were input into the digital twin. The simulation output parameters such as winding temperature, current, and power every 10 minutes for each scenario. Taking the peak load impact period of Scenario 2 (10:00-10:10) as an example, the simulation results showed that the A-phase current was 52.3A, the winding temperature was 78℃, and the active power was 950kW. The load factor was calculated as simulated active power / rated capacity (1000kVA), and the performance evaluation deviation was calculated as |simulated performance evaluation value - actual value| / actual value. Taking the 10:00-10:10 period of Scenario 2 as an example, the load factor was... 950 / 1000 = 95%, performance evaluation deviation = 2%, then the load safety value = 0.7 × (1 - 0.95) + 0.3 × (1 - 0.02) = 0.329; The load safety value distribution for each time period in the next 24 hours is statistically analyzed in 1000 simulation scenarios. When ≥80% of the scenarios have a load safety value ≥ 0.8 in a certain time period, it is considered safe; 60%-80% is relatively safe; 40%-60% is risky; and <40% is high-risk. Simultaneously, the prediction deviation is calculated by comparing this with the actual operating results, i.e., the actual operating data on August 1st. The results are shown in Table 2 below.

[0139] Table 2 Prediction Deviation Table

[0140]

[0141] The results show that the deviation between the simulated load safety value and the actual operating value is less than 7% in all three scenarios, and the accuracy of the load safety result determination is 100%. This indicates that the simulation technology in this embodiment can accurately predict the load safety result of the future operating state.

[0142] The load safety result matching early warning scheme based on simulation prediction was compared with the failure rate of the traditional post-event handling mode and the pre-event prediction false handling mode in this embodiment. The results are shown in Table 3 below:

[0143] Table 3 Comparison Results of Early Warning Schemes

[0144]

[0145] The comparative results show that the failure rate of this embodiment is reduced by 78.6% compared with the traditional model, the average power outage time is shortened by 70.8%, and the operation and maintenance cost is reduced by 42.5%. This fully demonstrates that this embodiment can effectively improve the operation and maintenance efficiency of distribution network equipment, reduce failure losses, and has significant practical application effects.

[0146] like Figure 2 As shown, according to another embodiment of the present invention, a real-time monitoring and early warning system for the status of distribution network equipment is provided, the system comprising:

[0147] Standard setting module 1 is used to analyze the load-bearing capacity of distribution network equipment according to monitoring standards, and to establish load safety early warning levels under uncertain scenarios based on load-bearing capacity;

[0148] Frequency fluctuation analysis module 2 is used to establish a source-load balance analysis model based on historical operating data to describe the balance relationship between energy supply and energy consumption of power distribution equipment, and to randomly sample the source-load balance analysis model to output the power supply frequency fluctuation value of the power distribution equipment when it is in a state of supply-demand imbalance.

[0149] Performance evaluation module 3 is used to determine the duration and fault frequency of power distribution equipment in a state of supply and demand imbalance based on the power supply frequency fluctuation value, iteratively train the backpropagation neural network model, and output the performance evaluation value of the power distribution equipment.

[0150] The early warning scheme output module 4 is used to combine the performance evaluation value with the load safety early warning level and output the early warning scheme for the power distribution equipment.

[0151] The real-time monitoring output module 5 is used to combine the real-time operating data of the power distribution equipment with frequency calculation technology to simulate the future operating status of the power distribution equipment. Based on the simulation results, it outputs the load safety results of the power distribution equipment in the real-time state and in the future period, and matches and executes the power distribution equipment monitoring and early warning scheme based on the load safety results.

[0152] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time monitoring and early warning of the status of distribution network equipment, characterized in that, include: Analyze the load-bearing capacity of distribution network equipment according to monitoring standards, and establish load safety early warning levels under uncertain scenarios based on load-bearing capacity; Obtain historical operational data containing the operating load and power supply capacity of power distribution equipment, and generate an operational topology graph by using the power supply line connections between power distribution equipment as edges and the historical operational data as nodes. Analyze the operational topology to obtain the balance relationship between energy supply and energy consumption during power transmission, and use elastic network fitting technology to establish a source-load balance analysis model based on the balance relationship; Based on variational inference technology, different energy supply and energy consumption data are extracted from the source-load balance analysis model as sampling points to simulate the power supply frequency fluctuation state of power distribution equipment under power supply fluctuation conditions. The offset amplitude of power supply frequency fluctuations is extracted from historical operating data, and a judgment standard for supply and demand imbalance is defined based on the offset amplitude. The amplitude of power supply frequency fluctuations is analyzed through the judgment standard. The duration and fault frequency of power supply frequency fluctuations of power distribution equipment are analyzed based on the trend of amplitude changes, and the power distribution equipment is determined to be in a state of supply and demand imbalance when the duration and fault frequency exceed the threshold. A sliding window structure is generated using window partitioning technology to perform dependency analysis on duration and failure frequency, and an incremental learning method is used to generate a model training dataset based on the dependency analysis results. The model training dataset is used as input to train the backpropagation neural network model. The backpropagation neural network model is used to output the estimated duration and fault frequency of the power distribution equipment in the state of supply and demand imbalance, and output the performance evaluation value of the power distribution network equipment. By combining performance evaluation values ​​with load safety warning levels, a warning scheme for power distribution equipment is output. By combining real-time operating data of power distribution equipment with frequency calculation technology, the future operating status of power distribution equipment is simulated. Based on the simulation results, the load safety results of power distribution equipment in real-time state and in future time period are output, and a power distribution equipment monitoring and early warning scheme is matched and executed based on the load safety results.

2. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that, The analysis of the topology graph obtains the balance between energy supply and energy consumption during power transmission, and establishes a source-load balance analysis model based on the balance relationship using elastic network fitting technology, including: The running topology graph is transformed into a multi-relationship graph based on meta-paths. Power quality relationship selection graph convolutional layers are used to analyze the multi-relationship graph to obtain aggregated neighbor information in order to model the global correlation between meta-paths. The aggregated neighbor information is filtered to distinguish the importance of each power quality relationship, and the power quality relationships that meet the importance threshold are retained as the energy supply and demand balance changes of the power distribution equipment under different operating conditions; By using the operating time and maintenance information of power distribution equipment as independent variables and the changes in energy supply and demand balance as dependent variables, a pool of characteristic indicators is obtained for constructing the source-load balance analysis model. Based on the feature index pool, the influencing factor feature indexes are screened to determine the index combination used to construct the source-load balance analysis model. The source-load balance analysis model is constructed by using the index combination and elastic network fitting technology.

3. The method for real-time monitoring and early warning of distribution network equipment status according to claim 2, characterized in that, The process of transforming the running topology graph into a multi-relationship graph based on meta-paths, and using a power quality relationship selection graph convolutional layer to analyze the multi-relationship graph to obtain aggregated neighbor information, in order to model the global correlation between meta-paths, includes: Based on the power transmission relationship between power distribution equipment and the power supply node, the operating topology graph is defined as a heterogeneous graph, and the meta-path set is defined according to the business characteristics of the power distribution equipment. The meta-path set is used to obtain the neighbor set of the power supply node under each meta-path, so as to transform the heterogeneous graph into a multi-relationship graph based on meta-paths. The power supply nodes are projected into the feature subspace of the graph convolutional network to obtain the influencing factor index, and the influencing factor index is assigned to each meta-path. The neighbor information of the power supply node under each meta-path is calculated by the mean aggregation function, and the global power supply quality information of each meta-path is extracted by the readout function. The global neighbor information is obtained by calculating the correlation between neighbor information and global power quality information, and removing neighbor information that is less than the threshold in relation to global power quality information based on the correlation. The global neighbor information is projected onto the same feature space to calculate the weight coefficients of the meta-path, and combined with the power supply node information to generate a weighted fusion node representation through an activation function, thus obtaining aggregated neighbor information.

4. The method for real-time monitoring and early warning of distribution network equipment status according to claim 3, characterized in that, The step of determining the index combination for constructing the source-load balance analysis model by screening influencing factor characteristic indicators based on the characteristic index pool, and constructing the source-load balance analysis model using the index combination and elastic network fitting technology includes: Pearson linear correlation coefficient, Kendall rank correlation coefficient and Spearman rank correlation coefficient were used to quantify the degree of correlation between the independent variable indicators and the dependent variable indicators in the feature indicator pool, respectively. By comparing the correlation degree with the selection criteria, key characteristic indicators affecting the source-load balance are screened, effective indicator combinations for constructing the source-load balance analysis model are determined, and the differences in the dimensions and dispersion of the effective indicator combinations are judged. The effective index combination is preprocessed based on the differences in dimensionality and dispersion, and a source-load balance analysis model is constructed based on the preprocessed effective index combination using ridge regression and lasso regression algorithms. By solving the optimization problem with quadratic programming, the elastic network estimate of the correlation coefficients between the independent variable index and the dependent variable index in the source-load balance analysis model is calculated, and the source-load balance analysis results are obtained.

5. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that, The process of combining real-time operating data of power distribution equipment with frequency calculation technology to simulate the future operating state of the power distribution equipment, outputting load safety results of the power distribution equipment in the real-time state and in the future time period based on the simulation results, and matching and executing a power distribution equipment monitoring and early warning scheme based on the load safety results includes: The system acquires real-time operating data of power distribution equipment and uses frequency dynamic calculation technology to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data, thereby generating random scenarios for the future operation of the power distribution equipment. Based on Monte Carlo simulation technology, the random scenarios of power distribution equipment are simulated to simulate the state performance of power distribution equipment under different loads and environmental fluctuation factors, and to evaluate the performance evaluation value of power distribution equipment. Based on real-time operating data and performance evaluation values, the system outputs the load safety results of power distribution equipment in real-time and in future time periods, matches the power distribution equipment monitoring and early warning scheme, and executes early warning processing of power distribution equipment based on the matching results.

6. The method for real-time monitoring and early warning of distribution network equipment status according to claim 5, characterized in that, The process of acquiring real-time operating data of power distribution equipment and using frequency dynamic calculation technology to simulate the probability of the operating state of the power distribution equipment based on the real-time operating data to generate random scenarios for the future operation of the power distribution equipment includes: The real-time operating data of the power distribution equipment is collected by sensors within the target time period. The moving average filtering technique is used to denoise and normalize the real-time operating data to complete the data preprocessing. Based on real-time operational data and the specifications and structural parameters of power distribution equipment, a digital twin containing the structure, operating parameters, environmental variables, and operating logic of the power distribution equipment is constructed using dynamic modeling. The mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. The operating characteristics of the power distribution equipment are combined to divide the state set and simulate the operating state probability of the power distribution equipment. Based on the operating state probability, a Markov chain is used to construct the transition probability matrix for each operating state, and the transition probability matrix is ​​used to generate random operating scenarios for power distribution equipment in the future prediction period.

7. The method for real-time monitoring and early warning of distribution network equipment status according to claim 6, characterized in that, The mapping process between the digital twin and the physical entity of the power distribution equipment is performed to obtain a mature digital twin. The operating characteristics of the power distribution equipment are then combined to divide the state set and simulate the operating state probabilities of the power distribution equipment, including: By performing state mapping and behavior simulation mapping, the real-time synchronization of the physical entity of the power distribution equipment to the digital twin is completed, and the ability to replicate the real-time operating status of the digital twin is added, resulting in a mature digital twin. Based on the operating characteristics of power distribution equipment, a set of states is divided, and time windows are divided. The frequency of real-time operating data corresponding to each operating state in the mature digital twin within each time window is counted, and the probability of the operating state of the power distribution equipment is determined by the frequency.

8. A real-time monitoring and early warning system for the status of distribution network equipment, used to implement the real-time monitoring and early warning method for the status of distribution network equipment as described in any one of claims 1-7, characterized in that, The system includes: The standard setting module is used to analyze the load-bearing capacity of distribution network equipment according to monitoring standards, and to establish load safety early warning levels under uncertain scenarios based on the load-bearing capacity. The frequency fluctuation analysis module is used to establish a source-load balance analysis model based on historical operating data to describe the balance between energy supply and energy consumption of power distribution equipment, and to randomly sample the source-load balance analysis model to output the power supply frequency fluctuation value of the power distribution equipment when it is in a state of supply-demand imbalance. The performance evaluation module is used to determine the duration and fault frequency of power distribution equipment in a state of supply and demand imbalance based on the power supply frequency fluctuation value, iteratively trains the backpropagation neural network model, and outputs the performance evaluation value of the power distribution equipment. The early warning scheme output module is used to combine performance evaluation values ​​with load safety early warning levels to output early warning schemes for power distribution equipment. The real-time monitoring output module is used to combine the real-time operating data of the power distribution equipment with frequency calculation technology to simulate the future operating status of the power distribution equipment. Based on the simulation results, it outputs the load safety results of the power distribution equipment in the real-time state and in the future period, and matches and executes the power distribution equipment monitoring and early warning scheme based on the load safety results.

Images