Intelligent commutation type three-phase imbalance dynamic treatment method for power distribution network

Abstract

The invention discloses an intelligent commutation type three-phase imbalance dynamic treatment method for a power distribution network, relates to the technical field of power distribution network treatment, and aims to solve the problem of poor treatment effect of three-phase imbalance. By using the data mining and analysis technology, not only can the cost minimization path be calculated according to the current state, but also the future change can be pre-judged in combination with the load dynamic trend, so that the formulated switching strategy is more prospective and accurate, the generated high-dimensional mapping vector fuses the node characteristics and the neighbor node characteristics, and the switching efficiency is improved. By combining the use of a nonlinear activation function, the model can adaptively learn a load change rule, sensor equipment integrates various sensors and chips, has the functions of environment monitoring, power management and the like, and is matched with structural components to ensure the reliable operation of the device, reduce the operation and maintenance cost and difficulty and prolong the service life of the equipment.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power distribution network management, in particular to a power distribution network intelligent commutation type three-phase imbalance dynamic management method. BACKGROUND

[0002] Power distribution network management refers to the systematic management, optimization and improvement of the power distribution network to ensure its safe, stable and efficient operation and meet user demands.

[0003] A Chinese patent with the publication number CN113300358A discloses an active power distribution network electric energy sensor device and a use method thereof, which mainly detects the real-time voltage and current signals of the active power distribution network, analyzes the quality of electric energy, and when the power grid has obvious voltage drop, low power factor, three-phase imbalance and large harmonic content due to power generation fluctuation or load mutation, the power control unit inputs the electric energy of the energy storage device into the power grid after inversion, the isolation transformer performs excitation compensation, the voltage signal after voltage reduction is obtained through the transformer, followed by voltage following of the third operational amplifier, one way is input into the MCU after inverse amplification and amplitude limiting through the fourth operational amplifier, and the other way is converted into a pulse signal through zero comparison of the fifth operational amplifier and then 100 times frequency division, so that the MCU can better obtain the voltage frequency, phase and change trend. Although the above patent solves the problem of electric energy management, there are still the following problems in actual operation:

[0004] 1. The sensor device is not effectively deployed, so that the sensor device cannot be applied to more management scenarios.

[0005] 2. The electrical quantity data is not subjected to more targeted data mining and analysis, so that the effective optimization management path generation cannot be performed according to the actual situation.

[0006] 3. The final management scheme is not subjected to targeted safety analysis, and the execution debugging is not performed according to the management scheme, so that the final management effect is poor. SUMMARY

[0007] The application aims to provide a power distribution network intelligent commutation type three-phase imbalance dynamic management method, which can not only calculate the cost minimization path according to the current state by using data mining and analysis technology, but also can predict future changes in combination with load dynamic trends, so that the switching strategy is more forward-looking and accurate, the generated high-dimensional mapping vector combines the node itself features and neighbor node features, and the use of the nonlinear activation function enables the model to adaptively learn the load change law, the sensor device integrates various sensors and chips, has functions such as environment monitoring and power management, and cooperates with the structural components to ensure the reliable operation of the device, reduces the operation and maintenance cost and difficulty, and prolongs the service life of the device, which can solve the problems in the prior art.

[0008] To achieve the above object, the present application provides the following technical solutions:

[0009] A power distribution network intelligent commutation type three-phase unbalanced dynamic treatment method, comprising:

[0010] First, the electrical quantity data of the busbar and the commutation switch branch in the distribution area are collected in real time; the collected electrical quantity data is analyzed for load characteristics; a graph theory topological model of the power distribution network is constructed according to the analysis result; the change rule and correlation of the load are mined according to the graph theory topological model; the load state and dynamic trend of each node in the mining result are mapped into a high-dimensional space by using a graph neural network; after the mapping is completed, an optimal path optimization model is established, and a commutation switch switching strategy is formulated according to the optimal path optimization model; the formulated optimal commutation switch switching strategy is analyzed for safety; a safety analysis report after the safety analysis is completed is communicated and transmitted; and the transmitted safety analysis report is embedded into a control terminal of a sensor device for execution and debugging.

[0011] Preferably, the electrical quantity data of the busbar and the commutation switch branch in the distribution area are collected in real time, including:

[0012] Before the electrical quantity data is acquired, the sensor device is deployed first;

[0013] The sensor device comprises a processor, a display screen, a solid-state relay, a commutation switch matrix, a surge suppressor, a voltage and current transformer, a temperature and humidity sensor, a safety protector, a power management chip and a structural assembly;

[0014] Among them, the voltage transformer is installed on the three-phase busbar in the distribution area, and the A, B and C three-phase voltages and the zero sequence voltage are measured respectively, and the current transformer is deployed at the outgoing end of the busbar to monitor the three-phase total current and the neutral line current, the micro CT is installed on the input and output branch of each commutation switch to measure the three-phase current value of each branch in real time, and the branch voltage sensor is integrated in the commutation switch cabinet to acquire the branch end voltage signal;

[0015] The high-voltage and large-current signal is converted into a low-voltage and small-current signal by the voltage and current transformer, and the surge suppressor is used for overvoltage and overcurrent protection of the signal;

[0016] After the deployment of the sensor device is completed, the electrical quantity data sampling frequency is set, and after the setting is completed, the electrical quantity data is collected.

[0017] Preferably, the collected electrical quantity data is analyzed for load characteristics, including:

[0018] The collected electrical quantity data is preprocessed, and the data preprocessing includes data cleaning and data alignment;

[0019] After data preprocessing, basic electrical quantity calculation is performed, including key parameter extraction and harmonic analysis;

[0020] Among them, the key parameter extraction is to extract the voltage and current effective value, power parameter and unbalance degree index in the electrical quantity data, the voltage and current effective value is to calculate the RMS value of each phase voltage and current; the power parameter is active power, reactive power, apparent power, power factor and phase angle difference; the unbalance degree index is voltage unbalance degree and current unbalance degree; the harmonic analysis is to decompose the current waveform by FFT transform, and calculate the total harmonic distortion rate;

[0021] According to the basic electrical quantity calculation data, load classification is performed, including residential load, commercial load and industrial load;

[0022] After the load classification is completed, the basic electrical quantity calculation data is used to cluster the branches with similar load characteristics into a class by using unsupervised learning algorithm;

[0023] Finally, the load characteristic analysis of electrical quantity data is completed.

[0024] Preferably, according to the characteristic analysis result, a graph theory topological model of the power distribution network is constructed, including:

[0025] According to the load characteristic analysis result, the nodes of the graph theory topological model are confirmed, including bus nodes, phase change switch nodes, load nodes and transformer nodes;

[0026] Then the edges of the graph theory topological model are confirmed, including electrical connection edges, power flow edges and logical association edges;

[0027] According to the nodes and edges, the framework of the graph theory topological model is constructed;

[0028] The bus node is taken as the core node, and the phase change switch node is connected outwardly, and each phase change switch node is connected to the corresponding load node;

[0029] Finally, the construction of the graph theory topological model is completed.

[0030] Preferably, according to the graph theory topological model, the change rule and association of the load are mined, including:

[0031] The time dimension rule of the graph theory topological model is mined, wherein, first, the time period in the graph theory topological model is identified, including daily rule, weekly rule and monthly rule; according to the identified time period, the fluctuation characteristic is quantified, including load fluctuation rate quantification, mutation frequency quantification and continuous high load duration quantification;

[0032] After the time dimension regularity mining is completed, the spatial dimension correlation mining is performed, wherein, the electrical correlation in the graph theory topological model is analyzed first, including the coupling degree analysis of the same phase branch and the cross-phase influence analysis, the coupling degree analysis of the same phase branch is to confirm the synchronous index mechanical energy of the load of each branch under the same phase bus;

[0033] According to the time dimension regularity and the spatial dimension correlation mining results, the correlated events are confirmed, wherein, the time dimension regularity and the spatial dimension correlation mining results are introduced into the event chain rule library to locate the correlated events;

[0034] According to the located correlated events, the dynamic behavior is predicted, and after the dynamic behavior is predicted, the unbalanced reasons and the unbalanced positions of different branches in the graph theory topological model are obtained.

[0035] Preferably, the graph neural network is used to map the load state and the dynamic trend of each node in the mining results to a high-dimensional space, including:

[0036] Each node in the graph theory topological model is mapped to a feature vector, and each edge in the graph theory topological model is mapped to a physical correlation feature;

[0037] The mapped nodes and edges are converted into input tensors of the graph neural network, and are subjected to normalization processing;

[0038] After the input tensor conversion of the nodes and the edges is completed, according to the edge feature weight, the features of adjacent nodes are aggregated, the attention weight of the current node and the neighbor nodes is calculated, after the calculation is completed, the key neighbors are highlighted, and then the self feature and the aggregated neighbor feature are fused through a nonlinear activation function to generate a new node hidden state;

[0039] According to the new node hidden state, a high-dimensional mapping vector of each node is generated, the high-dimensional mapping vector includes a load state and a dynamic trend, wherein, the load state is a real-time electrical parameter, a load type and an unbalanced degree of the current node; the dynamic trend is a fluctuation mode of the load with time and an association trend with adjacent nodes;

[0040] Finally, the mapping of the high-dimensional space is completed.

[0041] Preferably, after the mapping is completed, an optimal path optimization model is established, and a phase change switch switching strategy is formulated according to the optimal path optimization model, including:

[0042] Before the optimal path optimization model is established, the optimization target and the constraint condition are confirmed first, wherein, the optimization target is to minimize the three-phase unbalance degree, to minimize the phase change action times and to minimize the voltage fluctuation; the constraint condition is an electrical safety constraint, a device capacity constraint, a topological constraint and a user influence constraint;

[0043] The node and its connection relationship in the graph theory topology model after the high-dimensional space mapping is completed are converted into state nodes of the optimal path optimization model, wherein each state node contains real-time electrical parameters, load type and dynamic trend;

[0044] After the state node conversion is completed, a graph search algorithm is used to calculate the cost minimization of the path in the optimal path optimization model, and the node dynamic trend output by the graph neural network is input to output the path that can cope with the load growth and decay trend as the switching path;

[0045] After the cost minimization calculation of the path in the optimal path optimization model is completed, the optimal path in the optimal path optimization model is obtained, and the key switching nodes and sequence of the optimal path are extracted;

[0046] Finally, the optimal phase-change switch switching strategy is generated according to the optimal path and the key switching nodes and sequence, including target switch number, current phase, target phase, expected unbalance degree improvement value, power transfer amount and operation time window;

[0047] Finally, the optimal phase-change switch switching strategy is formulated.

[0048] Preferably, the formulated optimal phase-change switch switching strategy is subjected to safety analysis, including:

[0049] The safety analysis includes electrical parameter safety verification, device load capacity verification, topology result rationality check, operation risk analysis and dynamic risk prediction of the optimal phase-change switch switching strategy;

[0050] The electrical parameter safety verification is voltage and current out-of-limit check, unbalance degree compliance check and harmonic influence evaluation; the device load capacity verification is phase-change switch and solid-state relay load verification, sensor and protection device compatibility check and environmental parameter monitoring; the topology result rationality check is electrical connection logic verification, power flow direction check and loop and island risk elimination; the operation risk analysis is power outage influence evaluation, voltage fluctuation suppression verification and communication and control reliability check; the dynamic risk prediction is cross-phase coupling influence analysis, load dynamic trend matching check and protection device action logic verification;

[0051] The safety analysis results are comprehensively judged, and the risk level of the optimal phase-change switch switching strategy is determined according to the judgment results, and the risk level is divided into high risk, medium risk and low risk;

[0052] The optimal phase-change switch switching strategy divided into high risk is directly determined as a rejected scheme, and the corresponding risk points in the optimal phase-change switch switching strategies of medium risk and low risk are adjusted,

[0053] Finally, the optimal switching strategy of the medium-risk and low-risk phase-change switch is used to generate a security analysis report.

[0054] Preferably, the security analysis report is communicated after the security analysis is completed, including:

[0055] The generated security analysis report is transmitted to the processor port of the sensor device;

[0056] Before the security analysis report is transmitted, the transmission amount of the report is confirmed, and the transmission amount of the report is retrieved from the database;

[0057] Then, the remaining capacity of each transmission channel at the time of transmission is confirmed;

[0058] The transmission channel with a report transmission amount less than the remaining capacity of the channel is selected as the final transmission channel for transmitting the security analysis report to the processor port of the sensor device;

[0059] Finally, the data transmission of the security analysis report is completed.

[0060] Preferably, the security analysis report is embedded in the control terminal of the sensor device for execution debugging, including:

[0061] The port of the processor in the sensor device receives the security analysis report, and after receiving, the report content is parsed to obtain the key information in the report, including the target switch number, the current phase, the target phase, the expected unbalance improvement value, the power transfer amount, the operation time window and the safety constraint condition;

[0062] The parameters in the key information are verified for rationality, wherein the electrical parameters in the strategy are compared with the safety threshold value built-in the sensor device, and at the same time, the real-time data of the temperature and humidity sensor are combined to confirm the current environmental parameters;

[0063] The parameters that pass the rationality verification are sent by the processor to the solid-state relay and the phase-change switch matrix to execute the instruction and pre-configure the switch state;

[0064] Finally, the sensor device performs governance debugging according to the execution instruction.

[0065] Preferably, after the security analysis report is embedded in the control terminal of the sensor device for execution debugging, it further includes:

[0066] The execution debugging result of the sensor device is described to obtain a feature description vector; wherein the execution debugging result of the sensor device at least includes: switch state confirmation, unbalance improvement, power transfer amount verification, safety constraint condition satisfaction, time window execution, environment monitoring feedback and abnormal situation handling;

[0067] matching the feature description vector with a plurality of standard feature description vectors in the standard feature description vector library respectively to obtain a plurality of first matching degrees;

[0068] when the maximum first matching degree exceeds the first high threshold, performing on the sensor device a first depth governance debugging scheme corresponding to the standard feature description vector generating the maximum first matching degree;

[0069] Otherwise, calculating a fusion verification value of the second depth governance debugging scheme corresponding to each of the standard feature description vectors generating the top N first matching degrees by the following formula:

[0070]

[0071] wherein Y is the fusion verification value, K1 and K2 are preset weight values, S i is the i-th second matching degree among a plurality of second matching degrees between the governance debugging value distributions of the second depth governance debugging schemes, M is the total number of the second matching degrees, P i is the i-th first matching degree among the top N first matching degrees, and N is a preset positive integer;

[0072] when the fusion verification value exceeds the second high threshold, fusing the second depth governance debugging schemes corresponding to each of the standard feature description vectors generating the top N first matching degrees to obtain a fused depth governance debugging scheme;

[0073] performing on the sensor device the fused depth governance debugging scheme.

[0074] Preferably, the fusing the second depth governance debugging schemes corresponding to each of the standard feature description vectors generating the top N first matching degrees to obtain a fused depth governance debugging scheme comprises:

[0075] planning a scheme fusion rule optimal for the second depth governance debugging schemes corresponding to each of the standard feature description vectors generating the top N first matching degrees;

[0076] fusing the second depth governance debugging schemes corresponding to each of the standard feature description vectors generating the top N first matching degrees based on the planned optimal scheme fusion rule to obtain a fused depth governance debugging scheme;

[0077] wherein when planning the optimal scheme fusion rule, the following steps are specifically performed:

[0078] Based on the scheme fusion rule planning model, the second deep governance debugging scheme corresponding to each standard feature description vector with the first matching degree of the N largest standard feature description vector is planned as a candidate scheme fusion rule;

[0079] The planned candidate scheme fusion rule closest to the fusion value target is selected as the optimal scheme fusion rule.

[0080] Compared with the prior art, the present application has the following advantages:

[0081] 1. The power distribution network intelligent commutation type three-phase imbalance dynamic governance method provided by the present application integrates different devices and components in the sensor device, can collect multi-dimensional data such as bus and branch voltage, current, load, environmental parameters, etc. in real time, can find harmonic components in the current waveform through FFT transformation for harmonic analysis, can calculate the total harmonic distortion rate, and can help determine whether the device is abnormal.

[0082] 2. The power distribution network intelligent commutation type three-phase imbalance dynamic governance method provided by the present application fully utilizes data mining and analysis technology, can not only calculate the cost minimization path according to the current state, but also can predict future changes in combination with the dynamic trend of the load, so that the switching strategy formulated is more forward-looking and precise, and the high-dimensional mapping vector generated combines the node itself features and neighbor node features, and the use of a nonlinear activation function enables the model to adaptively learn the load change law.

[0083] 3. The power distribution network intelligent commutation type three-phase imbalance dynamic governance method provided by the present application ensures that the multi-dimensional analysis has no dead angle for safety evaluation, and realizes the reasonable allocation of transmission resources through the double confirmation of the transmission amount and the channel residual capacity before transmission, and through the deep analysis of the safety analysis report by the processor, the accurate extraction of key information such as target switch number and power transfer amount, the avoidance of redundant data interference, and the ensuring of the accuracy of subsequent execution instructions.

[0084] 4. The execution debugging result of the sensor device is used to perform deep governance debugging on the sensor device, which greatly improves the governance level of the system. Specifically, a standard feature description vector and its respective deep governance debugging scheme are set in advance, it is judged whether there is a deep governance debugging scheme that can be directly used based on the first matching degree, if yes, the sensor device is executed, otherwise, it is determined whether the second deep governance debugging scheme corresponding to each standard feature description vector with the first matching degree of the N largest standard feature description vector is suitable for fusion by calculating the fusion verification value, if yes, it is fused, and the sensor device is executed after the fusion of the fusion deep governance debugging scheme, which greatly improves the ability of the system to perform deep governance debugging on the sensor device, and improves the applicability and intelligent level of the system.

[0085] 5. Introduce a scheme fusion rule planning model and a fusion value target, based on the scheme fusion rule planning model, plan a to-be-selected scheme fusion rule for the second deep governance and debugging scheme corresponding to each of the standard feature description vectors that produce the first matching degrees of the top N standard feature description vectors, select the planned to-be-selected scheme fusion rule closest to the fusion value target as the optimal scheme fusion rule, and based on the planned optimal scheme fusion rule, fuse the second deep governance and debugging schemes corresponding to each of the standard feature description vectors that produce the first matching degrees of the top N standard feature description vectors, to obtain a fused deep governance and debugging scheme, greatly improving the fusion accuracy, comprehensiveness and efficiency of the fused deep governance and debugging scheme, and improving the effect of deep governance and debugging of the sensor equipment. BRIEF DESCRIPTION OF DRAWINGS

[0086] Figure 1 The figure is a schematic diagram of the commutation type three-phase imbalance governance step of the present application. DETAILED DESCRIPTION

[0087] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0088] Embodiment one

[0089] Electric quantity data acquisition and load characteristic analysis

[0090] In order to solve the problem that in the prior art, the sensor equipment is not effectively deployed, so that the sensor equipment cannot be applied to more governance scenarios, please refer to Figure 1 The technical solutions of the present embodiment are as follows:

[0091] A power distribution network intelligent commutation type three-phase imbalance dynamic governance method, comprising:

[0092] The electrical quantity data of the busbar and the commutation switch branch of the transformer area is collected in real time; the electrical quantity data collected is analyzed for load characteristics; a graph theory topological model of the distribution network is constructed according to the analysis result; the change rule and correlation of the load are mined according to the graph theory topological model; the load state and dynamic trend of each node in the mining result are mapped into a high-dimensional space by using a graph neural network; after the mapping is completed, an optimal path optimization model is established, and the commutation switch switching strategy is formulated according to the optimal path optimization model; the optimal commutation switch switching strategy formulated is analyzed for safety; the safety analysis report after the safety analysis is completed is communicated and transmitted; and the safety analysis report after the transmission is completed is embedded into the control terminal of the sensor device for execution and debugging.

[0093] Specifically, by collecting the electrical quantity data of the busbar and the commutation switch branch of the transformer area in real time, combining the load characteristic analysis and the graph theory topological model construction, the change rule and correlation of the load can be deeply mined. This process enables the system to accurately master the operation state of the distribution network, greatly improves the comprehensiveness and accuracy of data processing compared with the traditional way, provides solid data support for subsequent decision-making, maps the load state and dynamic trend into a high-dimensional space by using a graph neural network, establishes an optimal path optimization model to formulate a commutation switch switching strategy, realizes the intelligentization and scientization of strategy formulation, analyzes the formulated strategy for safety, and embeds the analysis report into the sensor device for execution and debugging, thereby avoiding safety hazards caused by improper strategies from the source.

[0094] The electrical quantity data of the busbar and the commutation switch branch of the transformer area is collected in real time, including:

[0095] Before the electrical quantity data is acquired, the sensor device is deployed first;

[0096] The sensor device includes a processor, a display screen, a solid-state relay, a commutation switch matrix, a surge suppressor, a voltage and current transformer, a temperature and humidity sensor, a safety protector, a power management chip and a structural assembly;

[0097] Among them, the voltage transformer is installed on the three-phase busbar of the distribution transformer area, and the A, B and C three-phase voltages and the zero sequence voltage are measured, respectively. The current transformer is deployed at the outgoing end of the busbar to monitor the three-phase total current and the neutral line current. The micro CT is installed on the input and output branch of each commutation switch to measure the three-phase current value of each branch in real time. The branch voltage sensor is integrated in the commutation switch cabinet to obtain the branch end voltage signal.

[0098] The high-voltage and large-current signal is converted into a low-voltage and small-current signal by the voltage and current transformer, and then the surge suppressor is used for overvoltage and overcurrent protection of the signal;

[0099] After the sensor deployment is completed, the electrical quantity data sampling frequency is set, and the electrical quantity data is collected after the setting is completed.

[0100] Specifically, the sensor device integrates different devices and components, which can collect multi-dimensional data such as voltage, current, load, and environmental parameters of the bus and branch in real time, providing full-domain data support for precise management. It integrates multiple sensors and chips, has environmental monitoring and power management functions, and cooperates with structural components to ensure reliable operation of the device, reduce operation and maintenance costs and difficulty, improve the service life of the device, and overall improve the operation efficiency and reliability of the distribution network. Install voltage transformers on the three-phase bus of the distribution area to measure three-phase and zero sequence voltage, install current transformers at the bus outlet to monitor three-phase total current and neutral line current, install micro CTs at the input and output branches of the commutating switch to measure three-phase current of each branch, and simultaneously obtain branch end voltage signals. This comprehensive sensor deployment achieves comprehensive coverage of electrical quantity data of the bus and commutating switch branch of the distribution area, accurately captures electrical parameters of each key node, and provides complete and accurate data basis for subsequent data analysis and decision-making. Through voltage and current transformers, high-voltage and large-current signals are converted into low-voltage and small-current signals, reducing safety risks in signal transmission and processing. At the same time, surge suppressors are used for overvoltage and overcurrent protection of signals, effectively avoiding damage to equipment caused by abnormal voltage and current fluctuations, improving the stability and reliability of the entire acquisition system, and ensuring the continuous and stable operation of data acquisition. The electrical quantity data sampling frequency can be set according to actual needs to meet the real-time and accuracy requirements of different application scenarios. Real-time acquisition of electrical quantity data can timely reflect the operation state of the distribution area, facilitating operators to quickly discover abnormal conditions and take appropriate measures.

[0101] The collected electrical quantity data is analyzed for load characteristics, including:

[0102] The collected electrical quantity data is pre-processed, including data cleaning and data alignment.

[0103] After data preprocessing, basic electrical quantity calculation is performed, including key parameter extraction and harmonic analysis.

[0104] Among them, the key parameter extraction is to extract the voltage and current effective value, power parameter and unbalance degree index in the electrical quantity data. The voltage and current effective value is the RMS value of each phase voltage and current. The power parameter is active power, reactive power, apparent power, power factor and phase angle difference. The unbalance degree index is voltage unbalance degree and current unbalance degree. Harmonic analysis is to decompose the current waveform by FFT transform and calculate the total harmonic distortion rate.

[0105] According to the basic electrical quantity calculation data, the load is classified, including residential load, commercial load and industrial load.

[0106] The basic electrical quantity calculation data of completed load classification is classified into one category by using an unsupervised learning algorithm.

[0107] Finally, the load characteristics of the electrical quantity data are analyzed.

[0108] Specifically, data cleaning can eliminate abnormal data generated by noise, equipment failure, etc. during the collection process, ensuring the reliability of the analysis data; data alignment solves the time difference problem of data collected by different devices, laying a foundation for subsequent accurate analysis; the extracted voltage, current effective value, power parameter and unbalance degree index can accurately present the running efficiency, power quality and other information of the system; harmonic analysis through FFT transformation can find the harmonic components in the current waveform, calculate the total harmonic distortion rate, help to judge whether the equipment is abnormal, and provide data support for equipment maintenance and system optimization; combined with the unsupervised learning algorithm clustering, the branches with similar load characteristics can be mined, which helps the power enterprise to understand the power utilization rules of different user groups, reasonably plan the power grid resources, and improve the power supply reliability and economy. At the same time, the scheme process is complete and the logic is rigorous, from data collection to the final analysis result output, each link is closely connected, forming a closed loop, which provides a scientific and effective technical means for power system operation analysis and optimal scheduling.

[0109] Embodiment two

[0110] Constructing a distribution network topology model, mining load rules, and generating an optimal commutation switch switching strategy

[0111] In order to solve the problem that in the prior art, the electrical quantity data is not subjected to more targeted data mining and analysis, so that the effective optimization treatment path cannot be generated according to the actual situation, please refer to Figure 1 The embodiment provides the following technical scheme:

[0112] According to the characteristic analysis result, a graph theory topology model of the distribution network is constructed, including:

[0113] According to the load characteristic analysis result, the nodes of the graph theory topology model are confirmed, including bus nodes, commutation switch nodes, load nodes and transformer nodes;

[0114] Then, the edges of the graph theory topology model are confirmed, including electrical connection edges, power flow edges and logical association edges;

[0115] According to the nodes and edges, the framework of the graph theory topology model is constructed;

[0116] The bus node is taken as a core node, and the commutation switch nodes are connected outwardly and radiately, and each commutation switch node is connected to a corresponding load node;

[0117] Finally, the construction of the graph theory topology model is completed.

[0118] Specifically, the nodes of the power distribution network are explicitly divided into bus nodes, commutation switch nodes, load nodes and transformer nodes, and the edges are divided into electrical connection edges, power flow edges and logical association edges. This classification makes the topology of the power distribution network clear at a glance. Taking the bus node as the core, it is connected to the commutation switch node, and then to the load node. The hierarchical structure facilitates the planning, design and analysis of the power system, and can quickly identify the functions and connection relationships of each part. Based on the analysis results of the load characteristics, the nodes are confirmed, so that the model can accurately reflect the operation of the power distribution network under different load conditions. The setting of various edges, especially the power flow edges, intuitively shows the transmission path and direction of power in the network, which helps to analyze power loss, power flow distribution and other issues, providing strong support for the optimal operation of the power system. The clear topology structure and the definition of nodes and edges can quickly locate the fault node and the related impact range when a fault occurs in the power distribution network. By analyzing the connection relationship between the fault node and other nodes, a scheme for fault isolation and power restoration can be quickly developed, the power outage time is shortened, and the power supply reliability is improved. The modularized node connection method makes it easy to expand the existing model when new loads or equipment are added to the power distribution network. The setting of the commutation switch node also provides the possibility for flexible regulation of the system, which facilitates the response to various changes in the operation of the power system and enhances the universality and adaptability of the model.

[0119] According to the variation law and association of the graph theory topology model, including:

[0120] The time dimension law of the graph theory topology model is mined, wherein the time period in the graph theory topology model is identified, including daily law, weekly law and monthly law; the fluctuation characteristics are quantified according to the identified time period, including load fluctuation rate quantification, mutation frequency quantification and continuous high load duration quantification;

[0121] After the time dimension law mining is completed, the spatial dimension association mining is performed, wherein the electrical association in the graph theory topology model is analyzed, including the analysis of the coupling degree of the same phase branch and the analysis of the cross-phase influence, and the coupling degree analysis of the same phase branch is the load synchronization index mechanical energy confirmation of each branch under the same phase bus;

[0122] According to the time dimension law and the spatial dimension association mining results, the associated events are confirmed, wherein the time dimension law and the spatial dimension association mining results are imported into the event chain rule library for associated event positioning;

[0123] According to the positioned associated events, the dynamic behavior is predicted, and the unbalance reason and unbalance position of different branches in the graph theory topology model are obtained after the dynamic behavior prediction.

[0124] Specifically, the time and space dimensions are deeply mined, respectively. On the time dimension, the day, week, and month rules are accurately identified, and the load change trend and fluctuation characteristics are carefully described through quantitative indicators such as load fluctuation rate, mutation frequency, and duration of continuous high load. On the space dimension, the electrical correlation is focused on, and the in-phase branch coupling degree and cross-phase influence analysis are performed. The time dynamics and spatial correlation of load changes are covered in all directions to avoid information omission. The abstract load change rule is converted into quantifiable indicators, making the research on load change rule and correlation characteristics more scientific and objective. These quantitative data not only intuitively reflect the load change situation, but also provide a solid data foundation for subsequent analysis and prediction, enhancing the credibility and persuasiveness of the research results. By importing the mining results of time and space dimensions into the event chain rule library, the associated events are accurately located, and then based on this, the dynamic behavior prediction is performed, which can predict the unbalance reasons and positions of different branches in advance. This provides forward-looking decision-making basis for the operation and dispatching, fault prevention, and optimization management of the power system, and helps to improve the stability and reliability of the power system operation. From rule mining to correlation analysis, to event confirmation and behavior prediction, the entire scheme forms a complete closed loop, with each link closely connected and progressive, having strong systematicness and logicality, and can effectively guide actual operation.

[0125] The graph neural network is used to map the load state and dynamic trend of each node in the mining results to a high-dimensional space, including:

[0126] Each node in the graph theory topology model is mapped to a feature vector, and each edge in the graph theory topology model is mapped to a physical correlation feature;

[0127] The mapped nodes and edges are converted into input tensors of the graph neural network and normalized;

[0128] After the conversion of the input tensors of the nodes and edges is completed, the features of adjacent nodes are aggregated according to the edge feature weight, the attention weight of the current node and the neighbor nodes is calculated, the key neighbors are highlighted after the calculation, and then the self-feature and the aggregated neighbor features are fused through a nonlinear activation function to generate a new node hidden state;

[0129] A high-dimensional mapping vector of each node is generated according to the new node hidden state, and the high-dimensional mapping vector includes a load state and a dynamic trend, wherein the load state is the real-time electrical parameter, load type, and unbalance degree of the current node, and the dynamic trend is the fluctuation mode of the load over time and the correlation trend with adjacent nodes;

[0130] Finally, the mapping of the high-dimensional space is completed.

[0131] Specifically, the feature vector mapping and the physical correlation feature mapping are respectively performed on the nodes and edges of the graph theory topology model, and the physical connection and electrical characteristics of the power distribution network are converted into computer processable information. In this way, the complex relationship between each node and line of the power distribution network is fully considered, and the network topology structure can be accurately reflected. Compared with the traditional method which only relies on node data, the network overall situation can be more comprehensively and accurately presented, laying a solid foundation for subsequent analysis. Through normalization processing of the input tensor and aggregation of node features based on edge feature weights, the attention weight is calculated to highlight the key neighbors. The scheme can efficiently extract the key features of the load state and dynamic trend. This mechanism not only captures the real-time electrical parameters, load types and other static information of the node itself, but also excavates the dynamic characteristics such as the fluctuation pattern of the load over time and the correlation trend with adjacent nodes, avoiding information redundancy and improving data utilization efficiency. The generated high-dimensional mapping vector integrates the node features and neighbor node features, and the use of a nonlinear activation function enables the model to adaptively learn the load change rule. Based on this, the model can better predict the future change trend of the load, provide a forward-looking basis for the development of three-phase imbalance control strategy, enhance the ability of the system to respond to dynamic changes of the load, and improve the timeliness and effectiveness of the control. After mapping the load state and dynamic trend to the high-dimensional space, the load distribution and change trend of each node in the power distribution network can be more clearly presented. Power management personnel can optimize resource allocation and develop more scientific phase-change switch switching strategies based on this information, reduce the loss caused by three-phase imbalance, and improve the efficiency and stability of power grid operation.

[0132] After the mapping is completed, an optimal path optimization model is established, and a phase-change switch switching strategy is developed according to the optimal path optimization model, including:

[0133] Before establishing the optimal path optimization model, the optimization target and the constraint condition are confirmed, wherein the optimization target is to minimize the three-phase imbalance degree, minimize the number of phase-change actions, and minimize the voltage fluctuation; the constraint condition is the electrical safety constraint, the device capacity constraint, the topology constraint, and the user influence constraint;

[0134] The nodes and their connection relationships in the graph theory topology model after the high-dimensional space mapping are converted into state nodes of the optimal path optimization model, wherein each state node contains real-time electrical parameters, load types, and dynamic trends;

[0135] After the state node conversion is completed, a graph search algorithm is used to calculate the minimum cost of the path in the optimal path optimization model, and the node dynamic trend output by the graph neural network is input to output the path that can cope with the load growth and decay trend as the switching path;

[0136] After the cost minimization calculation of the path in the optimal path optimization model is completed, the optimal path in the optimal path optimization model is obtained, and the key switching nodes and sequence of the optimal path are extracted;

[0137] Finally, the optimal commutation switch switching strategy is generated according to the optimal path and the key switching nodes and sequence, including target switch number, current phase, target phase, expected unbalance degree improvement value, power transfer amount and operation time window;

[0138] Finally, the optimal commutation switch switching strategy is formulated.

[0139] Specifically, the minimization of three-phase unbalance degree, commutation action number and voltage fluctuation is taken as the optimization target, and the core problems of power grid operation are comprehensively considered. Reducing three-phase unbalance degree can improve power quality and reduce line loss; controlling the number of commutation actions can prolong the service life of equipment and reduce operation and maintenance costs; and reducing voltage fluctuation ensures user power stability. This multi-objective optimization realizes the balance of power grid operation efficiency, economy and reliability, and fully guarantees the safety and feasibility of system operation through electrical safety constraints, equipment capacity constraints, topology constraints and user influence constraints. The electrical safety constraint prevents dangerous conditions such as overvoltage and overcurrent; the equipment capacity constraint ensures that the commutation operation is within the equipment bearing range; the topology constraint conforms to the characteristics of the power grid structure; and the user influence constraint reduces the interference to user power consumption, building a safety line for strategy implementation. The graph theory topology model after high-dimensional space mapping is converted into a state node containing real-time electrical parameters, load type and dynamic trend, combined with graph search algorithm and graph neural network, fully utilizing data mining and analysis technology. Not only can the cost minimization path be calculated according to the current state, but also the future changes can be predicted in combination with the dynamic trend of the load, making the formulated switching strategy more forward-looking and accurate, effectively dealing with complex and variable power grid operation scenarios. The generated optimal commutation switch switching strategy covers detailed information such as target switch number, current phase, target phase, etc., clearly defining the operation object, operation content, expected effect and operation time window, providing clear and executable operation guidelines for operation and maintenance personnel, reducing human decision-making errors, improving commutation operation efficiency, and assisting efficient and stable operation of the power grid.

[0140] Embodiment three

[0141] Safety analysis, communication transmission and execution debugging

[0142] In order to solve the problem that in the prior art, the final treatment scheme is not subjected to safety analysis, and execution debugging is not performed according to the treatment scheme, resulting in poor final treatment effect, please refer to Figure 1 The embodiment provides the following technical solutions:

[0143] The safety analysis of the formulated optimal commutation switch switching strategy includes:

[0144] The safety analysis includes electrical parameter safety verification of the optimal commutation switch switching strategy, device load capacity verification, topology result rationality check, operation risk analysis and dynamic risk prediction;

[0145] The electrical parameter safety verification is voltage and current out-of-limit check, unbalance degree compliance check and harmonic influence evaluation; the device load capacity verification is commutation switch and solid-state relay load verification, sensor and protection device compatibility check and environmental parameter monitoring; the topology result rationality check is electrical connection logic verification, power flow direction check and loop and island risk investigation; the operation risk analysis is power-off influence evaluation, voltage fluctuation suppression verification and communication and control reliability check; the dynamic risk prediction is cross-phase coupling influence analysis, load dynamic trend matching check and protection device action logic verification;

[0146] The safety analysis results are comprehensively judged, and the optimal commutation switch switching strategy is risk level judged according to the judgment results, and the risk level is divided into high risk, medium risk and low risk;

[0147] The optimal commutation switch switching strategy divided into high risk is directly judged as a rejected scheme, and the corresponding risk points in the optimal commutation switch switching strategy of medium risk and low risk are adjusted,

[0148] Finally, the optimal commutation switch switching strategy of medium risk and low risk generates a safety analysis report.

[0149] Specifically, the electrical parameter safety verification accurately controls key indicators such as voltage, current, unbalance degree and harmonic, and guarantees system operation safety from the core electrical quantity level; the device load capacity verification focuses on devices such as commutation switches and solid-state relays, and combines with environmental parameter monitoring to ensure stable operation of devices during strategy implementation; the topology result rationality check goes deep into bottom logic such as electrical connection logic and power flow direction, and investigates loop and island hazards to guarantee grid structure safety; the operation risk analysis and dynamic risk prediction respectively target operation process and future trend to mine risks, and this multi-dimensional analysis ensures that there is no dead angle in strategy safety evaluation, and through comprehensive judgment of safety analysis results, risk levels are divided, different treatments are taken for different risk levels, high-risk strategies are directly rejected to avoid major safety accidents; and risk points of medium and low risk strategies are adjusted to effectively reduce potential risks. This grading control mechanism can quickly exclude serious hazards and flexibly optimize strategies to improve resource utilization efficiency.

[0150] The safety analysis report completed by the safety analysis is communicated, including:

[0151] The generated safety analysis report is transmitted to the processor port of the sensor device;

[0152] The transmission amount of the safety analysis report is confirmed before the report is transmitted, and the transmission amount of the report is retrieved from the database;

[0153] The remaining capacity of each transmission channel at the time of transmission is also confirmed;

[0154] The transmission channel with a transmission amount less than the remaining capacity of the channel is selected as the final transmission channel for transmitting the safety analysis report to the processor port of the sensor device;

[0155] The data transmission of the safety analysis report is finally completed.

[0156] Specifically, by confirming the report transmission amount and the remaining capacity of each transmission channel, and selecting the channel with a transmission amount less than the remaining capacity of the channel for transmission, the problem of data transmission congestion, packet loss and other problems caused by insufficient channel capacity is effectively avoided. This precise channel matching method ensures that the safety analysis report can be transmitted stably and efficiently to the processor port of the sensor device, providing a reliable basis for subsequent data processing and analysis. The double confirmation of transmission amount and channel remaining capacity before transmission realizes the reasonable allocation of transmission resources. Avoiding the transmission delay caused by blind selection of channels, the safety analysis report can reach the destination at the fastest speed, allowing the sensor device to obtain data in time and speeding up the response and processing speed of safety problems, improving the operation efficiency of the entire system. Strictly according to the relationship between the transmission amount and the remaining capacity of the channel to select the transmission channel, the possibility of data loss is reduced from the source. In the process of data transmission, data loss may cause important safety information to be missing, affecting the accurate judgment of the sensor device on the safety status of the system. This scheme enhances the integrity and reliability of data transmission through a scientific channel selection strategy, retrieves the report transmission amount from the database, and dynamically confirms the remaining capacity of the channel. This way fully utilizes the existing data resources and transmission resources of the system. According to the actual situation, the transmission channel is selected flexibly to avoid resource waste, improve the utilization rate of the transmission channel, realize the optimal allocation of data transmission resources, and reduce the system operation cost.

[0157] The safety analysis report is embedded in the control terminal of the sensor device for execution and debugging, including:

[0158] The processor port of the sensor device receives the safety analysis report, and after receiving, the report content is parsed to obtain the key information in the report, including the target switch number, the current phase, the target phase, the expected unbalance improvement value, the power transfer amount, the operation time window and the safety constraint condition;

[0159] The parameters in the key information are subjected to parameter rationality verification, wherein the electrical parameters in the strategy are compared with the safety threshold values built in the sensor device, and meanwhile, the current environmental parameters are confirmed in combination with the real-time data of the temperature and humidity sensor;

[0160] The parameters that pass the rationality verification are sent by the processor to the solid-state relay and the commutation switch matrix to send execution instructions, and the switch states are pre-configured;

[0161] Finally, the sensor device performs management debugging according to the execution instructions.

[0162] Specifically, the safety analysis report is deeply analyzed by the processor to accurately extract key information such as target switch number and power transfer amount, avoid redundant data interference, and ensure the accuracy of subsequent execution instructions. This accurate information processing mechanism provides a reliable basis for accurate debugging of the sensor device, so that the management operation can be targeted, and in the parameter rationality verification link, both the electrical parameters and the built-in safety threshold values are compared, and the real-time environmental data of the temperature and humidity sensor are comprehensively judged. This double verification mechanism fully considers the electrical safety and environmental adaptability of the device operation, effectively avoids safety hazards caused by unreasonable parameters or environmental factors, and greatly improves the safety and stability of the sensor device operation. The parameters that pass the verification are sent to the solid-state relay and the commutation switch matrix to pre-configure the switch state, so that the sensor device can enter the "ready" state in advance. Compared with the traditional debugging method, the waiting time and operation steps of on-site debugging are reduced, the efficiency of management debugging is effectively improved, and the unbalanced problem in the power system can be quickly responded to and optimized and adjusted in time.

[0163] Embodiment Four

[0164] After the safety analysis report is embedded in the control terminal of the sensor device for execution debugging, the following steps are further included:

[0165] The execution debugging result of the sensor device is subjected to feature description to obtain a feature description vector; wherein the execution debugging result of the sensor device at least includes: switch state confirmation, unbalance degree improvement, power transfer amount verification, safety constraint condition satisfaction, time window execution, environmental monitoring feedback and abnormal situation handling situation;

[0166] The feature description vector is matched with a plurality of standard feature description vectors in a standard feature description vector library respectively to obtain a plurality of first matching degrees;

[0167] When the maximum first matching degree exceeds a first high threshold value, a first deep management debugging scheme corresponding to the standard feature description vector generating the maximum first matching degree is executed on the sensor device;

[0168] Otherwise, the fusion verification value of the second deep governance debugging scheme corresponding to each of the standard feature description vectors with the top N first matching degrees is calculated by the following formula:

[0169]

[0170] wherein Y is the fusion verification value, K1 and K2 are preset weight values, S i is the i-th second matching degree among the plurality of second matching degrees between the governance debugging values of the second deep governance debugging schemes, M is the total number of the second matching degrees, P i is the i-th first matching degree among the top N first matching degrees, and N is a preset positive integer.

[0171] When the fusion verification value exceeds the second high threshold value, the second deep governance debugging schemes corresponding to each of the standard feature description vectors with the top N first matching degrees are fused to obtain a fusion deep governance debugging scheme.

[0172] The fusion deep governance debugging scheme is executed on the sensor device.

[0173] The working principle and beneficial effects of the above technical solution are as follows:

[0174] The execution debugging result of the sensor device is obtained, and the execution debugging result is described by features and converted into a feature description vector in vector form. A standard feature description vector library is set in advance, and the library has a plurality of standard feature description vectors. Each standard feature description vector corresponds to a first deep governance debugging scheme. The first deep governance debugging scheme is a scheme for deeply governing and debugging the sensor device, and is suitable for the situation represented by the execution debugging result of other sensor devices corresponding to the first standard feature description vector. Therefore, the feature description vector is matched with the plurality of standard feature description vectors in the standard feature description vector library, and the greater the first matching degree obtained is, the more suitable the first deep governance debugging scheme corresponding to the corresponding standard feature description vector is for the current deep governance debugging of the sensor device.

[0175] The first high threshold value is a threshold value representing that the first matching degree is extremely high, which is set in advance. When the maximum first matching degree exceeds the first high threshold value, it means that the first deep governance debugging scheme corresponding to the standard feature description vector with the maximum first matching degree can be used for deeply governing and debugging the sensor device, and the sensor device is executed accordingly.

[0176] Otherwise (the maximum first matching degree does not exceed the first high threshold value), it is indicated that there is no first deep governance debugging scheme that can be directly used for deep governance debugging of the sensor device. The fusion verification value of the second deep governance debugging scheme corresponding to each of the standard feature description vectors with the top N first matching degrees is calculated through the above formula, and the fusion verification value represents the appropriate degree of size of the fusion of the second deep governance debugging scheme corresponding to each of the standard feature description vectors with the top N first matching degrees. The second high threshold value is a threshold value representing a larger fusion verification value. When the fusion verification value exceeds the second high threshold value, the second deep governance debugging scheme corresponding to each of the standard feature description vectors with the top N first matching degrees is fused to obtain a fused deep governance debugging scheme, and finally the sensor device is executed.

[0177] When calculating the fusion verification value, the governance debugging value distribution of the second deep governance debugging scheme includes multiple values of the second deep governance debugging scheme for deep governance debugging of the sensor device, at least including: improving system stability, optimizing power management, improving electrical imbalance, and improving safety. If the second matching degree between the governance debugging value distributions of two second deep governance debugging schemes is larger, it represents that the degree of repetition of the multiple values of the two for deep governance debugging of the sensor device is higher, and then it is less appropriate, and the fusion verification value is negatively correlated. In addition, the larger the first matching degree, the more suitable the corresponding second deep governance debugging scheme for the current deep governance debugging of the sensor device, and the first matching degree is positively correlated with the fusion verification value. Based on this, the calculation formula of the fusion verification value is designed, and then the weight value set in advance according to the influence degree of the second matching degree and the first matching degree on the fusion verification value is weighted to obtain the fusion verification value, so that the calculated fusion verification value can represent the appropriate degree of size of the fusion of the second deep governance debugging scheme corresponding to each of the standard feature description vectors with the top N first matching degrees.

[0178] The above technical solution is aimed at the execution debugging result of the sensor device, and the sensor device is deeply managed and debugged, which greatly improves the management level of the system. Specifically, the standard feature description vector and its respective deep management and debugging scheme are set in advance, it is judged whether there is a deep management and debugging scheme that can be directly used based on the first matching degree, if yes, the sensor device is executed, otherwise, it is determined whether the second deep management and debugging scheme corresponding to the standard feature description vector with the first matching degree of the N largest standard feature description vectors generated before the standard feature description vector is suitable for fusion, if yes, the fusion is performed, and the sensor device is executed after the fusion of the fusion deep management and debugging scheme, which greatly improves the ability of the system to deeply manage and debug the sensor device, and improves the applicability and intelligent level of the system.

[0179] The second deep management and debugging scheme corresponding to the standard feature description vector with the first matching degree of the N largest standard feature description vectors generated before the standard feature description vector is fused to obtain a fusion deep management and debugging scheme, including:

[0180] Planning a scheme fusion rule that is optimal for the second deep management and debugging scheme corresponding to the standard feature description vector with the first matching degree of the N largest standard feature description vectors generated before the standard feature description vector;

[0181] Fusing the second deep management and debugging scheme corresponding to the standard feature description vector with the first matching degree of the N largest standard feature description vectors generated before the standard feature description vector based on the planned optimal scheme fusion rule to obtain a fusion deep management and debugging scheme;

[0182] Wherein, when planning the optimal scheme fusion rule, the following steps are specifically executed:

[0183] Planning a candidate scheme fusion rule for the second deep management and debugging scheme corresponding to the standard feature description vector with the first matching degree of the N largest standard feature description vectors generated before the standard feature description vector based on a scheme fusion rule planning model;

[0184] Selecting the planned candidate scheme fusion rule closest to the fusion value target as the optimal scheme fusion rule.

[0185] The working principle and beneficial effects of the above technical solution are:

[0186] The scheme fusion rule planning model is obtained by machine learning training based on a large number of deep governance debugging schemes marked with scheme fusion rules, and can plan a candidate scheme fusion rule for each second deep governance debugging scheme corresponding to a standard feature description vector with a first matching degree to the standard feature description vector, the candidate scheme fusion rule being a scheme fusion rule output by the scheme fusion rule planning model, a fusion value target being set in advance, the fusion value target being an optimization target of the scheme fusion rule (such as a scheme cooperation capability target), and the planned candidate scheme fusion rule closest to the fusion value target being selected as the optimal scheme fusion rule. Based on the planned optimal scheme fusion rule, the second deep governance debugging schemes corresponding to the standard feature description vectors with the first matching degrees to the standard feature description vectors are fused to obtain a fused deep governance debugging scheme.

[0187] The above technical scheme introduces a scheme fusion rule planning model and a fusion value target, plans a candidate scheme fusion rule for each second deep governance debugging scheme corresponding to a standard feature description vector with a first matching degree to the standard feature description vector based on the scheme fusion rule planning model, selects the planned candidate scheme fusion rule closest to the fusion value target as the optimal scheme fusion rule, and fuses the second deep governance debugging schemes corresponding to the standard feature description vectors with the first matching degrees to the standard feature description vectors based on the planned optimal scheme fusion rule to obtain a fused deep governance debugging scheme, greatly improving the fusion accuracy, comprehensiveness and efficiency of the fused deep governance debugging scheme and improving the effect of deep governance debugging of the sensor equipment.

[0188] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or device.

[0189] Although the embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application.

Claims

1. A method for dynamic management of three-phase imbalance in a distribution network using intelligent commutation, characterized in that, include: First, real-time electrical quantity data of the busbar and phase-switching switch branches in the distribution area are collected; load characteristic analysis is performed on the collected electrical quantity data; a graph theory topology model of the distribution network is constructed based on the characteristic analysis results; the load variation patterns and correlations are mined based on the graph theory topology model; and the load status and dynamic trends of each node in the mined results are mapped to a high-dimensional space using a graph neural network. After mapping is completed, an optimal path optimization model is established, and a commutator switching strategy is formulated based on the optimal path optimization model; a safety analysis is performed on the formulated optimal commutator switching strategy. The security analysis report, once completed, is transmitted via communication. The completed security analysis report is embedded into the control terminal of the sensor device for execution and debugging. After embedding the completed security analysis report into the control terminal of the sensor device for execution and debugging, the process further includes: The execution and debugging results of the sensor device are characterized to obtain a feature description vector; wherein, the execution and debugging results of the sensor device include at least: switch status confirmation, imbalance improvement, power transfer verification, safety constraint satisfaction, time window execution, environmental monitoring feedback, and abnormal situation handling. The feature description vector is matched with multiple standard feature description vectors in the standard feature description vector library to obtain multiple first matching degrees; When the maximum first matching degree exceeds the first high threshold, the first deep governance and debugging scheme corresponding to the standard feature description vector that produces the maximum first matching degree with the standard feature description vector is executed on the sensor device. When the fusion verification value exceeds the second highest threshold, the second deep governance and debugging schemes corresponding to the standard feature description vectors that generate the top N first matching degrees with the standard feature description vectors will be fused to obtain the fusion deep governance and debugging scheme. Implement a deep fusion governance and debugging scheme for sensor devices; This includes the optimal scheme fusion rule for the second deep governance and debugging scheme corresponding to the top N standard feature description vectors with the first matching degree generated by the planning and standard feature description vectors; Based on the optimal scheme fusion rule of the plan, the second deep governance and debugging schemes corresponding to the standard feature description vectors that generate the top N first matching degrees with the standard feature description vectors are fused to obtain the fused deep governance and debugging scheme.

2. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 1, characterized in that, Real-time acquisition of electrical quantity data for the transformer substation busbars and phase-change switch branches, including: Before acquiring electrical quantity data, the sensor equipment must be deployed. Sensor devices include processors, displays, solid-state relays, commutation switch matrices, surge suppressors, voltage and current transformers, temperature and humidity sensors, safety protectors, power management chips, and structural components. Among them, voltage transformers are installed on the three-phase busbars of the distribution area to measure the three-phase voltages and zero-sequence voltage of A, B, and C respectively. In addition, current transformers are deployed at the outgoing terminals of the busbars to monitor the total three-phase current and neutral current. Miniature current transformers are installed on the input and output branches of each phase-switching switch to measure the three-phase current values ​​of each branch in real time. Furthermore, branch voltage sensors are integrated in the phase-switching switch cabinet to acquire the branch terminal voltage signal. High-voltage, high-current signals are converted into low-voltage, low-current signals using voltage and current transformers, and then surge suppressors are used to protect the signals from overvoltage and overcurrent. After the sensor equipment is deployed, the sampling frequency of electrical quantity data is set, and then electrical quantity data is collected.

3. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 2, characterized in that, The collected electrical quantity data is subjected to load characteristic analysis, including: The collected electrical quantity data undergoes data preprocessing, which includes data cleaning and data alignment. After data preprocessing, basic electrical quantity calculations are performed, including key parameter extraction and harmonic analysis. The key parameter extraction involves extracting the effective values ​​of voltage and current, power parameters, and unbalance indices from electrical quantity data. The effective values ​​of voltage and current are the RMS values ​​of each phase voltage and current. The power parameters include active power, reactive power, apparent power, power factor, and phase angle difference. The unbalance indices include voltage unbalance and current unbalance. The harmonic analysis involves decomposing the current waveform through FFT transformation and calculating the total harmonic distortion rate. Loads are classified based on basic electrical quantity calculation data, including residential loads, commercial loads, and industrial loads; The basic electrical quantity calculation data after load classification is used to cluster branches with similar load characteristics into one class using an unsupervised learning algorithm; Finally, the load characteristic analysis of the electrical quantity data was completed.

4. The method for dynamic management of three-phase imbalance in a distribution network using intelligent commutation as described in claim 3, characterized in that, Based on the characteristic analysis results, a graph theory topology model of the distribution network is constructed, including: Based on the load characteristic analysis results, the nodes of the graph theory topology model are confirmed, including bus nodes, commutation switch nodes, load nodes, and transformer nodes. Next, the edges of the graph theory topology model are confirmed, including electrical connection edges, power flow edges, and logical association edges; The framework of the graph theory topology model is constructed based on nodes and edges; The busbar node is used as the core node, and the phase switch nodes are connected outward. Each phase switch node is connected to the corresponding load node. Finally, the graph theory topology model was constructed.

5. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 4, characterized in that, Based on graph theory topology models, the variation patterns and correlations of loads are analyzed, including: The temporal dimension patterns of graph theory topology models are mined. First, the time periods in the graph theory topology models are identified, including daily, weekly, and monthly patterns. Then, the fluctuation characteristics are quantified based on the identified time periods, including load volatility quantification, mutation frequency quantification, and duration of sustained high load quantification. After the temporal pattern mining is completed, spatial dimension correlation mining is carried out. First, the electrical correlation in the graph theory topology model is analyzed, including the coupling degree analysis of in-phase branches and the cross-phase influence analysis. The coupling degree analysis of in-phase branches is to confirm the load synchronization index mechanical energy of each branch under the same phase bus. The correlation results are used to confirm related events based on the temporal and spatial correlation mining results. Specifically, the temporal and spatial correlation mining results are imported into the event chain rule base for related event location. Dynamic behavior prediction is performed based on the associated events of the location. After dynamic behavior prediction, the causes and locations of imbalances in different branches in the graph theory topology model are obtained.

6. The method for dynamic management of three-phase imbalance in a distribution network using intelligent commutation as described in claim 5, characterized in that, Graph neural networks are used to map the load status and dynamic trends of each node in the mining results to a high-dimensional space, including: Each node in the graph theory topology model is mapped with a feature vector, and each edge in the graph theory topology model is mapped with a physical association feature. The mapped nodes and edges are converted into input tensors for the graph neural network and then normalized. After the input tensors of nodes and edges are transformed, the features of neighboring nodes are aggregated according to the edge feature weights. The attention weights of the current node and its neighboring nodes are calculated. After the calculation is completed, key neighbors are highlighted. Then, the node's own features and the aggregated neighbor features are fused through a non-linear activation function to generate a new node hidden state. A high-dimensional mapping vector is generated for each node based on the new hidden state of the nodes. The high-dimensional mapping vector includes the load status and dynamic trend. The load status is the real-time electrical parameters, load type and imbalance degree of the current node; the dynamic trend is the load fluctuation pattern over time and the correlation trend with neighboring nodes. Finally, the mapping of high-dimensional space is completed.

7. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 6, characterized in that, After mapping is completed, an optimal path optimization model is established, and a commutator switching strategy is formulated based on the optimal path optimization model, including: Before establishing the optimal path optimization model, the optimization objectives and constraints are first confirmed. The optimization objectives are to minimize the three-phase imbalance, minimize the number of commutation operations, and minimize voltage fluctuations. The constraints are electrical safety constraints, equipment capacity constraints, topology constraints, and user impact constraints. The nodes and their connections in the graph theory topology model mapped in high-dimensional space are transformed into state nodes of the optimal path optimization model. Each state node contains real-time electrical parameters, load type, and dynamic trends. After the state node transition is completed, a graph search algorithm is used to calculate the cost minimization of the path in the optimal path optimization model. At the same time, the node dynamic trend output by the graph neural network is used to select the path that can cope with the load growth and decay trends as the switching path. After the cost minimization calculation of the path in the optimal path optimization model is completed, the optimal path in the optimal path optimization model is obtained, and the key switching nodes and their order of the optimal path are extracted. Finally, the optimal commutation switch switching strategy is generated based on the optimal path and key switching nodes and sequence, including the target switch number, current phase, target phase, expected imbalance improvement value, power transfer amount, and operating time window; Finally, the optimal commutator switching strategy was determined.

8. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 7, characterized in that, A security analysis is performed on the established optimal commutator switching strategy, including: Safety analysis includes verifying the electrical parameters of the optimal commutator switching strategy, checking the equipment load capacity, checking the rationality of the topology results, analyzing operational risks, and predicting dynamic risks. Among them, electrical parameter safety verification includes voltage and current over-limit checks, unbalance compliance verification, and harmonic impact assessment; equipment load capacity verification includes commutation switch and solid-state relay load verification, sensor and protection device compatibility checks, and environmental parameter monitoring; topology result rationality checks include electrical connection logic verification, power flow direction verification, and loop and islanding risk investigation; operational risk analysis includes power outage impact assessment, voltage fluctuation suppression verification, and communication and control reliability checks; dynamic risk prediction includes cross-phase coupling impact analysis, load dynamic trend matching verification, and protection device action logic verification. The results of the safety analysis are comprehensively judged, and the risk level of the optimal commutation switch switching strategy is determined based on the judgment results. The risk level is divided into high risk, medium risk and low risk. The optimal commutator switching strategy classified as high-risk is directly rejected, while the risk points in the medium-risk and low-risk optimal commutator switching strategies are adjusted. Finally, a safety analysis report is generated from the optimal commutator switching strategies for medium-risk and low-risk situations. The security analysis report, once completed, is transmitted via communication, including: The generated security analysis report is transmitted to the processor port of the sensor device; Before transmitting the security analysis report, the transmission volume of the report is confirmed, and the transmission volume of the report is retrieved from the database. Next, confirm the remaining capacity of each transmission channel during transmission; Select a transmission channel with a report transmission volume less than the remaining channel capacity as the final transmission channel for transmitting the security analysis report to the processor port of the sensor device; Finally, the data transmission of the security analysis report is completed; The completed security analysis report is embedded into the control terminal of the sensor device for execution debugging, including: The processor port in the sensor device receives the security analysis report, parses the report content, and obtains the key information in the report, including the target switch number, current phase, target phase, expected imbalance improvement value, power transfer amount, operating time window, and security constraints. The parameters in the key information are verified for reasonableness. Specifically, the electrical parameters in the strategy are compared with the safety thresholds built into the sensor devices. At the same time, the current environmental parameters are confirmed by combining real-time data from temperature and humidity sensors. The parameters that have passed the rationality verification are sent by the processor to the solid-state relays and commutation switch matrix to execute instructions and pre-configure the switch states; Finally, the sensor equipment is managed and debugged according to the executed instructions.

9. The method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 1, characterized in that, After embedding the completed security analysis report into the control terminal of the sensor device for debugging, the process also includes: The fusion verification value of the second deep governance and debugging scheme corresponding to each of the standard feature description vectors that generate the top N first-degree matches with the standard feature description vectors is calculated using the following formula: Where Y is the fusion verification value, K1 and K2 are preset weight values, and S i Let P be the i-th second matching degree among multiple second matching degrees between the governance and debugging value distributions of pairwise second-depth governance and debugging schemes, where M is the total number of second matching degrees, and P is the second matching degree among multiple second matching degrees. i Let be the i-th first matching degree among the top N largest first matching degrees, where N is a preset positive integer.

10. A method for dynamic management of three-phase imbalance in a distribution network using intelligent phase commutation as described in claim 9, characterized in that, The second deep governance and debugging schemes corresponding to the standard feature description vectors that generate the top N first matching degrees with the standard feature description vectors are fused together to obtain a fused deep governance and debugging scheme. include: When planning the optimal scheme fusion rules, the following steps are specifically executed: Based on the scheme fusion rule planning model, the second deep governance and debugging schemes corresponding to the standard feature description vectors that generate the top N first matching degrees with the standard feature description vectors are planned with the scheme fusion rules of the candidate schemes. The fusion rule of the candidate scheme that is closest to the fusion value goal is selected as the optimal fusion rule.

Images