Power distribution network evaluation method and system

Abstract

The invention provides a power distribution network evaluation method and system, and relates to the technical field of power system operation control, and the method comprises the steps: obtaining an extreme event index of a power distribution network in a current time period according to a measurement data flow, and judging whether an extreme event exists in the power distribution network or not according to the extreme event index; the measurement data streams are classified through a data quality grading mechanism, and then an observation weight matrix of the extreme events is constructed; supplementing missing data for the missing data segment according to the observation weight matrix, dynamically updating topological information of the power distribution network in combination with the supplemented missing data, and obtaining a real-time network model corresponding to the extreme event according to the topological information; and according to the real-time network model and the extreme event index, obtaining a state evaluation result of the power distribution network in the current time period. According to the invention, the evaluation precision of the power distribution network under extreme conditions is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system operation control, in particular to a power distribution network evaluation method and system. BACKGROUND

[0002] The evaluation of the state of the power distribution network is a key link to ensure the safe and stable operation of the power grid. The existing technology mainly relies on the steady-state measurement data provided by monitoring devices such as smart meters to realize real-time sensing and evaluation of the operating state of the node voltage and branch power of the power distribution network through weighted least squares, robust estimation or Kalman filtering algorithms, so as to better meet the demand for power distribution network evaluation under normal working conditions and provide an important decision basis for the operation and maintenance of the power grid.

[0003] In related technologies, when facing extreme situations such as typhoons, heavy rain floods or extremely low temperature weather, monitoring devices may produce a large amount of missing or abnormal data due to failure or interruption of data transmission. Moreover, when the topology of the power distribution network changes dramatically, such as large-scale line switching or rapid access and exit of distributed energy, data acquisition and transmission will be affected, resulting in a large amount of missing data, which seriously affects the accuracy and reliability of the power distribution network evaluation. SUMMARY

[0004] The problem solved by the present application is how to improve the evaluation accuracy of the power distribution network under extreme conditions.

[0005] To solve the above problems, the present application provides a power distribution network evaluation method and system.

[0006] In a first aspect, a power distribution network evaluation method of the present application comprises: obtaining a measurement data stream of the operating state of the power distribution network in a current time period; According to the measurement data stream, an extreme event index of the power distribution network in the current time period is obtained, and according to the extreme event index, it is judged whether there is an extreme event in the power distribution network; If yes, the measurement data stream is classified by a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream; According to the data in the complete data segment, an observation weight matrix of the extreme event is constructed; According to the observation weight matrix, a weighted tensor completion operation is performed on the missing data segment to obtain the missing data of the missing data segment; According to the complete data segment and the missing data, the topology information of the power distribution network is dynamically updated, and according to the topology information, a real-time network model corresponding to the extreme event is obtained; According to the real-time network model and the extreme event index, estimated values of node voltages and branch powers of the power distribution network are obtained through variable structure factor graph reasoning; According to the estimated values of the node voltages and the branch powers, a state evaluation result of the power distribution network in a current time period is obtained.

[0007] Optionally, the obtaining of the extreme event index of the power distribution network in the current time period according to the measurement data stream comprises: feature extraction is performed on the measurement data stream to obtain meteorological disturbance feature quantities and topology change feature quantities; comparison of the meteorological disturbance feature quantities with preset meteorological thresholds is performed to obtain a meteorological disturbance intensity index; comparison of the topology change feature quantities with preset topology thresholds is performed to obtain a topology change rate index; the extreme event index is generated by weighted fusion of the meteorological disturbance intensity index and the topology change rate index.

[0008] Optionally, the judging of whether the power distribution network has an extreme event according to the extreme event index comprises: comparison of the extreme event index with threshold ranges of preset extreme events is performed; if the extreme event index is in any of the threshold ranges, the preset extreme event corresponding to the threshold range is taken as the extreme event of the power distribution network; if the extreme event index is not in any of the threshold ranges of the preset extreme events, it is determined that the power distribution network is in a normal operating state.

[0009] Optionally, the classification of the measurement data stream through a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream comprises: data quality levels of each measurement point in the measurement data stream are obtained according to data integrity, timestamp continuity and numerical rationality of the measurement point; whether there is data in the measurement point is determined according to the data quality level; when there is data in the measurement point, the measurement point is taken as a complete data point; when there is no data in the measurement point, the measurement point is taken as a complete data point; the complete data segments and the missing data segments in the measurement data stream are determined according to the complete data points and the missing data points.

[0010] Optionally, the construction of the observation weight matrix of the extreme event according to data in the complete data segments comprises: determine a first weight component negatively correlated with the extreme event indicator according to the numerical credibility, time credibility and spatial correlation of each of the measurement points in the complete data segment; determine a second weight component positively correlated with the data quality level according to the data quality level of each of the measurement points in the complete data segment; fuse the first weight component and the second weight component to obtain a comprehensive weight of each of the measurement points; arrange the comprehensive weight according to the time dimension, node dimension and data type dimension to obtain the observation weight matrix.

[0011] Optionally, the weighting tensor completion operation on the missing data segment according to the observation weight matrix to obtain the missing data of the missing data segment comprises: take time, node and data type as three-dimensional coordinates, and map the complete data segment and the missing data segment into an initial tensor, wherein the initial tensor is zero at the missing position of the missing data; apply a weighted mask to the initial tensor according to the observation weight matrix to obtain a weighted tensor completion target function; solve the global optimal solution of the tensor completion target function by a convex optimization iterative algorithm to obtain a completion tensor; take the numerical value corresponding to the missing position in the completion tensor as the missing data of the missing data segment.

[0012] Optionally, the dynamic updating of the topology information of the power distribution network according to the complete data segment and the missing data, and the obtaining of the real-time network model corresponding to the extreme event according to the topology information comprise: fuse the complete data segment and the missing data to obtain fused measurement data flow; extract data from the fused measurement data flow to generate a real-time topology of the current power grid connection relationship of the power distribution network; differentially compare the real-time topology with a pre-stored reference topology, and correct the line parameters, distributed power access points and load models of the power distribution network according to the comparison result to obtain the real-time network model corresponding to the extreme event.

[0013] Optionally, the obtaining of the estimated value of the node voltage and branch power of the power distribution network by the variable structure factor graph inference according to the real-time network model and the extreme event indicator comprises: generate an initial factor graph containing node variables, measurement factors, pseudo-measurement factors and event-topology coupling factors according to the node-branch relationship of the real-time network model; determine an adaptive weight of the event-topology coupling factor according to the extreme event indicator; adjust node variables and factor nodes in the initial factor graph according to the adaptive weight to obtain a variable structure factor graph corresponding to the real-time topology; perform belief propagation inference on the variable structure factor graph to obtain estimated values of the node voltage and the branch power of the power distribution network.

[0014] Optionally, the obtaining of the state evaluation result of the power distribution network in the current time period according to the estimated values of the node voltage and the branch power includes: performing out-of-limit scanning on the estimated values according to a preset voltage safety interval and a branch power current carrying limit to generate a node voltage deviation indicator and a branch power out-of-limit indicator; mapping the node voltage deviation indicator and the branch power out-of-limit indicator to a risk quantification model to obtain a comprehensive risk score of the power distribution network in the current time period; obtaining the state evaluation result according to the comprehensive risk score.

[0015] In a second aspect, a power distribution network evaluation system is provided, which includes: a data monitoring unit configured to obtain a measurement data stream of an operating state of a power distribution network in a current time period; an extreme event judgment unit configured to obtain an extreme event indicator of the power distribution network in the current time period according to the measurement data stream, and determine whether an extreme event exists in the power distribution network according to the extreme event indicator; a data classification unit configured to, if so, classify the measurement data stream through a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream; a matrix construction unit configured to construct an observation weight matrix of the extreme event according to data in the complete data segments; a completion operation unit configured to perform a weighted tensor completion operation on the missing data segments according to the observation weight matrix to obtain missing data of the missing data segments; a model establishment unit configured to dynamically update topology information of the power distribution network according to the complete data segments and the missing data, and obtain a real-time network model corresponding to the extreme event according to the topology information; an estimation unit configured to obtain estimated values of node voltage and branch power of the power distribution network through variable structure factor graph inference according to the real-time network model and the extreme event indicator; an evaluation unit configured to obtain a state evaluation result of the power distribution network in the current time period according to the estimated values of the node voltage and the branch power.

[0016] The power distribution network evaluation method and system of the present application can accurately monitor and quickly respond to the operation state of the power distribution network by obtaining the extreme event index of the power distribution network in the current time period according to the measurement data stream. When an extreme event such as extreme weather or a dramatic change in the topology structure occurs, the traditional method is often difficult to accurately evaluate the state of the power distribution network due to the large amount of missing or abnormal monitoring data. Therefore, it is necessary to analyze the measurement data stream to obtain the extreme event index and determine whether the power distribution network has an extreme event. The present application avoids the evaluation delay or failure caused by data problems under extreme conditions through the active monitoring and judgment mechanism. Secondly, the combination of the data quality grading mechanism and the weighted tensor completion operation effectively solves the problem of data missing and abnormality under extreme events, further improving the accuracy of the evaluation. When it is determined that an extreme event exists, the measurement data stream is classified by the data quality grading mechanism to distinguish between complete data segments and missing data segments. Then, based on the complete data segments, an observation weight matrix of the extreme event is constructed, and the weighted tensor completion operation is performed on the missing data segments using the matrix to obtain the estimated value of the missing data. The present application not only makes full use of the information of the existing complete data, but also considers the quality and correlation of the data through the weighted method, making the completed data more reliable. Compared with the traditional data interpolation or simple estimation method, the weighted tensor completion operation can more accurately restore the missing data, reduce the influence of data errors on the evaluation results, and thus improve the accuracy of the evaluation. By dynamically updating the topology information of the power distribution network and constructing a real-time network model, the evaluation efficiency and accuracy are improved. Specifically, when an extreme event occurs, the topology structure of the power distribution network may change dramatically, such as line switching, rapid access or exit of distributed energy, etc. The traditional method is often difficult to capture these changes in time, resulting in a mismatch between the evaluation model and the actual operation state, and distorted evaluation results. According to the complete data segments and the completed missing data, the present application dynamically updates the topology information of the power distribution network and obtains the real-time network model corresponding to the extreme event, ensuring that the evaluation model can accurately reflect the actual operation state of the power distribution network under extreme conditions, providing strong support for subsequent accurate evaluation. At the same time, the construction of the real-time network model also improves the evaluation efficiency, avoiding repeated correction and iteration due to model update lag. According to the real-time network model and the extreme event index, the present application obtains the estimated value of the node voltage and branch power of the power distribution network through variable structure factor graph reasoning. The variable structure factor graph can be flexibly adjusted according to the real-time topology structure and operation state of the power distribution network, so as to more accurately capture the change relationship of the node voltage and branch power. Compared with the traditional fixed model evaluation method, the variable structure factor graph reasoning can more quickly respond to the changes brought by the extreme event, reduce the calculation redundancy in the reasoning process, and improve the evaluation efficiency. At the same time, by comprehensively considering the real-time network model and the extreme event index, the estimated value is closer to the actual operation state, thereby improving the accuracy of the evaluation.In summary, the application significantly improves the evaluation accuracy of the power distribution network in extreme conditions by active monitoring of extreme event indicators, data quality grading and weighted tensor completion, dynamic topology updating and real-time network model construction, variable structure factor graph reasoning, and detailed and accurate state evaluation result generation. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flowchart of a power distribution network evaluation method in an embodiment of the application is shown in FIG. 1. Figure 2 A structural diagram of a power distribution network evaluation system in another embodiment of the application is shown in FIG. 2. DETAILED DESCRIPTION

[0018] In order to make the above objectives, features and advantages of the application more apparent, specific embodiments of the application will be described in detail below with reference to the accompanying drawings. Although some embodiments of the application are shown in the drawings, it should be understood that the application can be implemented in various forms, and should not be interpreted as being limited to the embodiments described herein, but rather, these embodiments are provided to make the application more thorough and complete. It should be understood that the drawings and embodiments of the application are only for illustrative purposes, and are not intended to limit the scope of protection of the application.

[0019] It should be understood that the various steps described in the method embodiments of the application can be performed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the application is not limited in this respect.

[0020] The term "comprising" and variations thereof as used herein are open-ended, that is, "comprising but not limited to"; the term "based on" is "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". Related definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc. mentioned in the application are only used to distinguish different devices, modules or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modification of "one" or "multiple" mentioned in the application is illustrative and not limiting, and those skilled in the art should understand that unless otherwise explicitly stated in the context, it should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0023] In response to the problems existing in the above-mentioned related technologies, this embodiment provides a distribution network evaluation method and system.

[0024] Combine Figure 1 As shown, an embodiment of the present invention provides a distribution network evaluation method, including: Get the measurement data stream of the distribution network's operating status in the current time period.

[0025] Specifically, acquiring a stream of measurement data on the distribution network's operating status during the current time period is fundamental to the entire assessment method. Specifically, this step uses multiple monitoring devices, such as smart meters, μPMUs, and SCADA systems, to collect real-time operating status data on the distribution network. This operating status data includes, but is not limited to, node voltages, branch currents, power flows, and switch states. By centralizing this data, a continuous measurement data stream is formed, providing the raw input for subsequent analysis and processing. This process ensures the real-time and integrity of the data, providing a solid data foundation for subsequent extreme event detection and status assessment.

[0026] An extreme event index of the distribution network in the current time period is obtained according to the measurement data stream, and whether an extreme event exists in the distribution network is determined according to the extreme event index.

[0027] Specifically, it is first necessary to conduct in-depth analysis of the measured data to identify whether the distribution network is in an extreme operating state. In embodiments of the present invention, statistical analysis, signal processing, and other techniques are used to extract features from the data and compare them with features from normal operating conditions. By setting appropriate thresholds or pattern matching rules, it is possible to determine whether the current state meets the definition of an extreme event.

[0028] If so, the measurement data stream is classified through a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream.

[0029] Specifically, once an extreme event is detected, the reliability of the data becomes particularly important. Therefore, it is necessary to identify which data is reliable and which data may be missing or abnormal data due to device failure or transmission problems through a data quality grading mechanism. Based on this, each data point is carefully checked, such as determining the integrity of the data point, the continuity of the timestamp, and the reasonableness of the value, to ensure that subsequent data processing is based only on reliable data segments, and unreasonable data is also judged as data missing, so it is also divided into missing data segments, and also provides a basis for the completion of missing data. In addition, if not, it is determined that there is no extreme event in the power distribution network.

[0030] According to the data in the complete data segment, an observation weight matrix of the extreme event is constructed.

[0031] Specifically, in order to effectively complete the missing data, it is necessary to further analyze the existing complete data and construct a weight matrix that can reflect the importance and correlation of the data. Among them, the spatial and temporal characteristics of the data are modeled to determine the relative importance of each data point in describing the state of the power distribution network. The construction of the weight matrix needs to consider the position, time and relationship with other data points of the data, so as to provide a scientific basis for subsequent tensor completion.

[0032] According to the observation weight matrix, a weighted tensor completion operation is performed on the missing data segment to obtain the missing data of the missing data segment.

[0033] Specifically, the constructed observation weight matrix is used to complete the missing data. In an embodiment of the present application, a weighted tensor completion operation can be used, and through a mathematical model and algorithm, the missing data value is estimated according to the known data and its weight, so that the missing information can be restored as accurately as possible on the basis of considering data quality and correlation, making the entire data set more complete and reliable. In a preferred embodiment of the present application, a weighted low-rank approximation model can be used, an initial tensor is constructed according to the complete data, the missing data is disposed of zero and is given a lower weight. Then, through iterative solution by the alternating least squares method, the completed tensor is gradually obtained, so that the missing node voltage and branch power data are estimated, and complete data support is provided for the state evaluation of the power distribution network.

[0034] According to the complete data segment and the missing data, the topology information of the power distribution network is dynamically updated, and according to the topology information, a real-time network model corresponding to the extreme event is obtained.

[0035] Specifically, when an extreme event occurs, the structure of the power distribution network may change, such as switching of lines or failure of equipment. Therefore, it is necessary to update the topology information of the power distribution network according to the latest data. At this time, it is necessary to monitor the changes in the network structure in real time and adjust the network model accordingly. By dynamically updating the topology information, it can ensure that the network model can truly reflect the current physical structure and operating state of the power distribution network, and provide an accurate basis for subsequent analysis.

[0036] According to the real-time network model and the extreme event index, the estimated values of the node voltage and branch power of the power distribution network are obtained by variable structure factor graph reasoning.

[0037] Specifically, in combination with the real-time network model and the extreme event index, variable structure factor graph reasoning is used to estimate the key parameters of the power distribution network. Specifically, the structure and parameters of the reasoning model need to be adjusted flexibly according to the current network structure and operating state. Through such dynamic reasoning, even in complex and uncertain conditions caused by extreme events, the key parameters such as node voltage and branch power can be more accurately estimated.

[0038] According to the estimated values of the node voltage and the branch power, the state evaluation result of the power distribution network in the current time period is obtained.

[0039] Specifically, based on the estimated parameters such as node voltage and branch power, a comprehensive analysis is performed to evaluate the current state of the power distribution network. First, each parameter needs to be compared with the preset safety standard to identify potential risk points and problem areas. Through comprehensive evaluation of these parameters, a comprehensive state report is generated, thereby providing decision support in extreme cases, and facilitating staff to receive accurate judgments of the power distribution network, thereby ensuring the safe and stable operation of the power distribution network.

[0040] The power distribution network evaluation method of the present application, by obtaining the extreme event index of the power distribution network in the current time period according to the measurement data stream, thereby realizing the accurate monitoring and rapid response of the operation state of the power distribution network. When extreme events such as extreme weather or dramatic changes in the topology structure occur, the traditional method is often difficult to accurately evaluate the state of the power distribution network due to the large amount of missing or abnormal monitoring data, therefore, the present application analyzes the measurement data stream to obtain the extreme event index, and judges whether there is an extreme event in the power distribution network according to the extreme event index. Through the active monitoring and judgment mechanism, the evaluation delay or failure caused by data problems under extreme conditions is avoided. Secondly, the combination of the data quality grading mechanism and the weighted tensor completion operation effectively solves the problem of data missing and abnormality under extreme events, and further improves the accuracy of the evaluation. When it is judged that the extreme event exists, the measurement data stream is classified by the data quality grading mechanism to distinguish the complete data segment and the missing data segment. Then, the observation weight matrix of the extreme event is constructed based on the complete data segment, and the weighted tensor completion operation is performed on the missing data segment using the matrix to obtain the estimated value of the missing data. The present application not only makes full use of the information of the existing complete data, but also considers the quality and correlation of the data through the weighted method, so that the completed data is more reliable. Compared with the traditional data interpolation or simple estimation method, the weighted tensor completion operation can more accurately restore the missing data and reduce the influence of data errors on the evaluation result, thereby improving the accuracy of the evaluation. The topology information of the power distribution network is dynamically updated and the real-time network model is constructed to improve the evaluation efficiency and accuracy. Specifically, when the extreme event occurs, the topology structure of the power distribution network may change dramatically, such as line switching, rapid access or exit of distributed energy, etc. The traditional method is often difficult to capture these changes in time, resulting in that the evaluation model does not match the actual operation state, and the evaluation result is distorted. According to the complete data segment and the completed missing data, the topology information of the power distribution network is dynamically updated, and the real-time network model corresponding to the extreme event is obtained, which ensures that the evaluation model can accurately reflect the actual operation state of the power distribution network under extreme conditions, and provides strong support for subsequent accurate evaluation. At the same time, the construction of the real-time network model also improves the evaluation efficiency and avoids repeated correction and iteration caused by model update lag. According to the real-time network model and the extreme event index, the estimated value of the node voltage and branch power of the power distribution network is obtained through the variable structure factor graph reasoning. Among them, the variable structure factor graph can be flexibly adjusted according to the real-time topology structure and operation state of the power distribution network, so as to more accurately capture the change relationship of the node voltage and branch power. Compared with the traditional fixed model evaluation method, the variable structure factor graph reasoning can more quickly respond to the changes brought by the extreme event, reduce the calculation redundancy in the reasoning process, and improve the evaluation efficiency. At the same time, by comprehensively considering the real-time network model and the extreme event index, the estimated value is closer to the actual operation state, thereby improving the accuracy of the evaluation.In summary, the application significantly improves the evaluation accuracy of the power distribution network in extreme conditions by actively monitoring extreme event indicators, data quality grading and weighted tensor completion, dynamic topology updating and real-time network model construction, variable structure factor graph reasoning, and detailed and accurate state evaluation result generation.

[0041] Optionally, the extreme event indicator of the power distribution network in the current time period is obtained according to the measurement data stream, comprising: The measurement data stream is feature extracted to obtain meteorological disturbance feature quantity and topology change feature quantity; The meteorological disturbance feature quantity is compared with a preset meteorological threshold to obtain a meteorological disturbance intensity index; The topology change feature quantity is compared with a preset topology threshold to obtain a topology change rate index; The meteorological disturbance intensity index and the topology change rate index are weighted and fused to generate the extreme event indicator.

[0042] Specifically, first, meteorological disturbance feature quantity and topology change feature quantity are obtained from the measurement data stream through feature extraction. The meteorological disturbance feature quantity includes meteorological data such as wind speed and rainfall, while the topology change feature quantity includes changes in the physical structure of the power distribution network, such as changes in switch state and switching of lines. These feature quantities can directly reflect the changes in the external environment and internal structure of the power distribution network. Next, the extracted feature quantities are compared with the preset meteorological threshold and topology threshold respectively to obtain the meteorological disturbance intensity index and the topology change rate index. The preset threshold can be set based on analysis of historical data of the power distribution network and understanding of the characteristics of extreme events. By comparison, the influence degree of the current meteorological disturbance and topology change on the power distribution network is quantified. Finally, by weighted fusion method, the meteorological disturbance intensity index and the topology change rate index are combined to generate a comprehensive extreme event indicator. The process of weighted fusion needs to consider the importance of the two indicators, usually according to historical data and expert experience to determine the weight. This method can comprehensively reflect the influence of multiple factors in a single indicator, providing a scientific basis for subsequent judgment of whether the power distribution network exists extreme event.

[0043] In a preferred embodiment of the present application, various monitoring devices are equipped in the power distribution network to collect data such as wind speed, rainfall, switch state, etc. in real time. Meteorological thresholds (e.g. wind speed exceeding 30 m / s) and topological thresholds (e.g. switch switching frequency exceeding 5 times / hour) are set in advance. During a strong typhoon, the monitoring devices detect that the wind speed reaches 35 m / s, the rainfall increases sharply, and multiple switch switching occurs in the power distribution network due to the influence of the typhoon. After extracting these feature quantities, the meteorological disturbance intensity index and the topological change rate index are obtained by comparing with the preset thresholds. According to historical data and expert experience, the meteorological disturbance is given a greater weight (e.g. 0.7), and the topological change is given a smaller weight (e.g. 0.3). Through weighted fusion, an extreme event index is generated, and compared with the preset extreme event threshold to determine whether the power distribution network is in an extreme event state. This determination triggers the subsequent data processing and state evaluation process, enabling the operation and maintenance personnel to take timely measures to ensure the safe and stable operation of the power distribution network.

[0044] In the embodiments of the present application, the method of feature extraction and weighted fusion realizes accurate monitoring and rapid response of the power distribution network under extreme conditions. First, the feature extraction technology can extract key information from massive measurement data, reducing the complexity of data processing and improving the processing efficiency. Second, by comparing with the preset threshold, the influence of meteorological disturbance and topological change can be quantified, making the evaluation process more objective and operable. Finally, the weighted fusion method integrates different types of feature quantities into a comprehensive index, not only improving the accuracy of judgment, but also enhancing the adaptability and robustness of the method.

[0045] Optionally, the determining whether the power distribution network has an extreme event according to the extreme event index comprises: comparing the extreme event index with a preset threshold range of an extreme event; if the extreme event index is in any of the threshold ranges, the preset extreme event corresponding to the threshold range is taken as the extreme event of the power distribution network; if the extreme event index is not in any of the threshold ranges of the preset extreme events, it is determined that the power distribution network is in a normal operating state.

[0046] Specifically, the extreme event indicator is first compared with preset threshold ranges. These threshold ranges are preset according to historical data and expert experience, and can reflect the influence degree of different types of extreme events on the power distribution network. If the extreme event indicator falls within a certain threshold range, it indicates that the current state of the power distribution network meets the characteristics of the extreme event corresponding to the threshold range. This method can classify and identify different degrees of extreme events by setting multiple threshold ranges, thereby providing more detailed information for subsequent response measures. If the extreme event indicator is not within any preset threshold range, it is determined that the power distribution network is in a normal operating state. The threshold comparison-based judgment method is simple and direct, easy to implement, and can quickly give a judgment result, suitable for application in a real-time monitoring system. In the preferred embodiment of the present application, the first threshold range (0-0.3) corresponds to a slight extreme event, the second threshold range (0.3-0.6) corresponds to a moderate extreme event, and the third threshold range (0.6-1.0) corresponds to a severe extreme event. During a strong typhoon, the monitoring device detects that the extreme event indicator of the power distribution network is 0.75. The system compares this indicator with the preset threshold ranges and finds that it is within the third threshold range (0.6-1.0), so it determines that the power distribution network is in a severe extreme event state.

[0047] In the embodiment of the present application, the threshold range is set to determine whether an extreme event exists in the power distribution network, achieving fast and accurate judgment of the operating state of the power distribution network. It can timely identify whether the power distribution network is in an extreme event state, thereby providing a timely trigger signal for subsequent response measures. By setting multiple threshold ranges, not only can the existence of an extreme event be identified, but also the severity of the event can be classified, which helps the operation and maintenance personnel to take appropriate response measures according to the severity of the event.

[0048] Optionally, the classifying the measurement data stream by the data quality grading mechanism to obtain the complete data segment and the missing data segment in the measurement data stream comprises: obtaining a data quality level of each measurement point in the measurement data stream according to data integrity, timestamp continuity and value rationality of the measurement point; determining whether there is data in the measurement point according to the data quality level; when there is data in the measurement point, the measurement point is regarded as a complete data point; when there is no data in the measurement point, the measurement point is regarded as a complete data point; determining the complete data segment and the missing data segment in the measurement data stream according to the complete data point and the missing data point.

[0049] Specifically, according to the data integrity, timestamp continuity and numerical rationality of each measurement point in the measurement data stream, the data quality level of the measurement point is obtained. The data integrity mainly checks whether the data is missing or incorrect; the timestamp continuity ensures that the data is continuous and uninterrupted in the time dimension; the numerical rationality verifies whether the data is within the expected physical range, such as whether the voltage is within the normal range, whether the current exceeds the device capacity, etc. According to the comprehensive evaluation of these dimensions, each measurement point is assigned a data quality level, which reflects the credibility and availability of the data. Next, according to the data quality level, it is determined whether there is valid data in the measurement point. If the measurement point has data and the data quality level meets the preset standard, it is considered that the measurement point contains complete data, otherwise it is considered as missing data. By traversing all measurement points, the measurement data stream can be divided into complete data segments and missing data segments, for example, continuous complete data points form a complete data segment, and continuous missing data points form a missing data segment. This process not only identifies the problems in the data, but also provides a clear classification basis for subsequent data processing, ensuring the accuracy and reliability of subsequent analysis. For example, if the distribution network system operates in heavy rain, some smart meters are interrupted due to water ingress, while other devices are operating normally. By analyzing the measurement data stream through the data quality grading mechanism, for each measurement point, check its data integrity, timestamp continuity and numerical rationality. For example, the data of a certain smart meter is missing for a period of time, the timestamp is not continuous, and the value is outside the normal range, so it is rated as low-quality data level. Mark these low-quality data points as missing data points, and cluster continuous missing data points into missing data segments. At the same time, identify another part of the data that is complete, continuous in time stamp and reasonable in value, and mark it as complete data point and cluster it into complete data segment. Thus, it is successfully distinguished which data can be used for subsequent analysis and which data needs to be completed or corrected, thereby ensuring the accuracy and reliability of the state evaluation.

[0050] In the embodiments of the present application, by classifying each measurement point in the measurement data stream in detail, the pertinence and effectiveness of data processing are ensured. Complete data segments can be directly used for subsequent analysis and modeling, while missing data segments can be completed or corrected in a targeted manner. Not only improves the efficiency of data processing, but also enhances the robustness of the entire evaluation system. By identifying and classifying data quality problems, reducing false positives caused by data abnormalities, improving the accuracy and reliability of state evaluation.

[0051] Optionally, constructing the observation weight matrix of the extreme event according to the data in the complete data segment comprises: Determining a first weight component negatively correlated with the extreme event indicator according to the numerical credibility, temporal confidence, and spatial correlation of each measurement point in the complete data segment; determining, according to the data quality level of each of the measurement points in the complete data segment, a second weight component positively correlated with the data quality level; Fusing the first weight component and the second weight component to obtain a comprehensive weight for each of the measurement points; The comprehensive weights are arranged according to the time dimension, the node dimension and the data type dimension to obtain the observation weight matrix.

[0052] Specifically, first, according to the numerical credibility, time confidence and spatial correlation of each measurement point in the complete data segment, a first weight component negatively correlated with the extreme event index is determined. The numerical credibility reflects the rationality of the data in the physical sense, the time confidence ensures the continuity of the data in the time series, and the spatial correlation considers the correlation of the data in the geographical distribution. These factors jointly affect the reliability of the data under the extreme event, so the first weight component is negatively correlated with the extreme event index, that is, the more serious the extreme event, the lower the credibility of the data, and the weight is also correspondingly reduced. Secondly, according to the data quality level of each measurement point, a second weight component positively correlated with the data quality level is determined. The data quality level is obtained through multi-dimensional evaluation of data integrity, timestamp continuity and numerical rationality, and can directly reflect the availability and accuracy of the data. Therefore, the second weight component is positively correlated with the data quality level, and the higher the data quality, the greater the weight. Finally, the first weight component and the second weight component are fused to obtain the comprehensive weight of each measurement point. This fusion process is usually realized by weighted summation or other mathematical methods, aiming to comprehensively consider the influence of the extreme event and the reliability of the data itself. The comprehensive weight not only reflects the credibility of the data under the extreme event, but also embodies its importance in the overall data set. The comprehensive weight is arranged according to the time dimension, node dimension and data type dimension to finally form an observation weight matrix. The observation weight matrix provides a scientific weight basis for the subsequent weighted tensor completion operation, ensuring the scientificity and accuracy of the completion process. In the preferred embodiment of the present application, it is assumed that the power distribution network operates under typhoon weather, and some monitoring devices have data transmission interruption or data anomaly due to strong wind and heavy rain. The complete data segment and the missing data segment are identified through the data quality grading mechanism. For each measurement point in the complete data segment, further analyze its numerical credibility, time confidence and spatial correlation. For example, the data of a certain smart meter is reasonable in value, the timestamp is continuous, and it has strong spatial correlation with the data of other adjacent meters, so the first weight component of the smart meter is high. At the same time, the data quality level of the smart meter is also high, so the second weight component is also high. The two weight components are fused to obtain the comprehensive weight of the measurement point, and a higher weight value is assigned to the corresponding position in the observation weight matrix. For measurement points with lower data quality or greater typhoon impact, their comprehensive weight will be correspondingly reduced. In this way, high-quality data can be more accurately identified and utilized to provide a reliable basis for subsequent weighted tensor completion operation and state evaluation.

[0053] In the embodiments of the present application, by comprehensively considering the numerical credibility, time credibility, spatial correlation degree and data quality level, the observation weight matrix can more accurately reflect the reliability and importance of each measurement point under extreme events. This weight matrix not only can reduce the influence of data anomalies caused by extreme events on the evaluation results, but also can improve the credibility of the completed data, thereby improving the robustness and accuracy of the entire evaluation system. By reasonably allocating the weight, it ensures that high-quality data plays a greater role in the completion and evaluation process, further enhancing the credibility of the evaluation results.

[0054] Optionally, the weighted tensor completion operation on the missing data segment according to the observation weight matrix comprises: mapping the complete data segment and the missing data segment into an initial tensor, wherein the initial tensor is set to zero at the missing position of the missing data; applying a weighted mask to the initial tensor according to the observation weight matrix to obtain a weighted tensor completion target function; solving the global optimal solution of the tensor completion target function by a convex optimization iterative algorithm to obtain a completed tensor; taking the numerical value corresponding to the missing position in the completed tensor as the missing data of the missing data segment.

[0055] Specifically, time, node and data type are taken as three-dimensional coordinates, and the complete data segment and the missing data segment are mapped into an initial tensor. In this tensor, the position of the missing data is set to zero, so as to clearly distinguish the complete data and the missing data in subsequent operations. Next, a weighted mask is applied to the initial tensor according to the observation weight matrix to generate a weighted target function. This target function aims to minimize the difference between the completed data and the known data, while maximizing the integrity and consistency of the data. By a convex optimization iterative algorithm, the global optimal solution of the target function can be solved to obtain the completed tensor. Finally, according to the numerical value of the missing position in the completed tensor, the estimated value of the missing data is obtained. In the preferred embodiments of the present application, time, node and data type are taken as three-dimensional coordinates to construct the initial tensor wherein is the number of time points, is the number of nodes, is the number of data types.

[0056] For the position of the missing data, the corresponding position in the initial tensor is set to zero. That is: ; wherein, is the time , node , data type observed values at the missing positions.

[0057] According to the observation weight matrix , the initial tensor is applied with a weighted mask, resulting in a weighted tensor completion objective function: ; where is the tensor to be solved, represents element-wise multiplication, is the Frobenius norm, is the kernel norm, is the regularization parameter.

[0058] The above objective function is solved by a convex optimization iterative algorithm (such as the Alternating Direction Method of Multipliers, ADMM), resulting in the completion tensor . The specific iterative steps are as follows: Initialization , , Lagrange multiplier .

[0059] Iterative update: ; ; ; where is the soft threshold operator, is the threshold parameter, : the completion tensor estimate value of the th iteration. is the observed initial tensor, where the missing positions are zero. is the Lagrange multiplier of the th iteration, used to balance and , is the observation weight matrix, representing the credibility of each observed value, is the element-wise multiplication, used to apply the weight matrix to the tensor , emphasizing the difference between the observed value and the current estimate. is the auxiliary variable update value of the th iteration. is the soft threshold operator, used for thresholding the tensor, promoting sparse solutions. The threshold parameter controls the size of the regularization term, thereby affecting the sparsity of the solution. is the current completion tensor estimate value minus the Lagrange multiplier, used as input for the soft threshold operator. is the Lagrange multiplier update value of the first iteration, is the Lagrange multiplier update value of the first iteration, is the difference between the current completion tensor and the auxiliary variable, used to update the Lagrange multiplier to adjust the balance of the two in the subsequent iteration. is the estimated value of the missing data at time , node , and data type .

[0060] According to the numerical value of the missing position in the completion tensor , the estimated value of the missing data is obtained. For each missing position , the estimated value is: ; wherein, is the value of the corresponding position in the completion tensor , as the estimate of the missing data.

[0061] In the embodiments of the present application, by utilizing the three-dimensional tensor structure, the potential relationship between time, node and data type can be fully utilized, thereby improving the quality of the completed data. In addition, the introduction of the observation weight matrix ensures that high-quality data plays a greater role in the completion process, reducing the influence of low-quality data or noise. In addition, the use of convex optimization iterative algorithm ensures the stability and convergence of the solving process, making the completion result more reliable. This not only improves the accuracy of the power distribution network state evaluation, but also enhances the adaptability and robustness of the power distribution network state evaluation under extreme events.

[0062] Optionally, the method further comprises: fusing the complete data segment and the missing data to obtain fused measurement data stream; extracting data from the fused measurement data stream to generate a real-time topology of the current power grid connection relationship of the power distribution network; comparing the real-time topology with a pre-stored reference topology, and synchronously correcting the line parameters, distributed power supply access points and load model of the power distribution network according to the comparison result to obtain the real-time network model corresponding to the extreme event.

[0063] In particular, first, by fusing the complete data segment and the completed missing data, a complete measurement data stream is obtained, thereby ensuring the continuity and integrity of the data and providing a reliable data basis for subsequent topology update. Next, the fused measurement data stream is subjected to data extraction to generate a real-time topology of the distribution network. The generation of the real-time topology requires the extraction of switch states, protection signals and other key information from the data, which can reflect the connection relationship and structural changes of the distribution network under extreme events. Finally, the generated real-time topology is compared with the pre-stored reference topology to identify newly added, disconnected or reconstructed branches and nodes. According to the differences, the line parameters, distributed power access points and load models of the distribution network are simultaneously corrected, and finally a real-time network model corresponding to the extreme event is obtained. This process not only can timely capture the changes in the structure of the distribution network, but also can ensure the accuracy and real-time performance of the network model, thereby providing a reliable basis for subsequent state assessment. In the preferred embodiment of the present application, during a strong typhoon, part of the lines of a certain distribution network fail due to excessive wind force, resulting in changes in the topology structure. The topology information is dynamically updated and a real-time network model is constructed by the following steps: the complete data segment and the completed missing data are fused to obtain a fused measurement data stream. Assuming that the complete data segment is , the completed missing data segment is , and the fused measurement data stream is: ; The switch states, protection signals and other key information are extracted from the fused measurement data stream to generate a real-time topology. The real-time topology can be represented by an adjacency matrix , where is the number of nodes, indicates that there is a connection between node and node , and indicates no connection.

[0064] The real-time topology is compared with the pre-stored reference topology to identify newly added, disconnected or reconstructed branches and nodes. The comparison formula is: ; where represents the exclusive or operation, represents that the connection state between node and node has changed.

[0065] According to the difference result, the line parameters, distributed power access points and load models of the distribution network are simultaneously corrected. For example, if a branch is disconnected in the real-time topology, the admittance of the branch is set to zero, and the corresponding node parameters are updated.

[0066] Based on the revised topology information and parameters, a real-time network model corresponding to extreme events is constructed. The real-time network model can be represented as a graph structure ,in is a collection of nodes, is a set of branches, each branch Have corresponding parameters (such as resistance, reactance, etc.).

[0067] In an embodiment of the present invention, by dynamically updating the distribution network topology information and constructing a real-time network model, the accuracy and efficiency of distribution network status assessment can be significantly improved. The generation and updating of the real-time topology ensures that the assessment model accurately reflects the actual operating status of the distribution network under extreme events, avoiding assessment errors caused by model lag. By comparing differences with the baseline topology, changes in the distribution network structure can be quickly identified, and line parameters, distributed power access points, and load models can be simultaneously corrected, making the assessment model more closely aligned with actual operating conditions.

[0068] Optionally, obtaining estimated values ​​of node voltage and branch power of the distribution network by variable structure factor graph reasoning based on the real-time network model and the extreme event index includes: generating an initial factor graph including node variables, measurement factors, pseudo-measurement factors, and event-topology coupling factors according to the node-branch relationship of the real-time network model; determining an adaptive weight of the event-topology coupling factor according to the extreme event indicator; Adjusting node variables and factor nodes in the initial factor graph according to the adaptive weights to obtain a variable structure factor graph corresponding to the real-time topology; Belief propagation inference is performed on the variable structure factor graph to obtain estimated values ​​of the node voltage and the branch power of the power distribution network.

[0069] Specifically, an initial factor graph is generated according to the node-branch relationship of the real-time network model, wherein the initial factor graph contains node variables, measurement factors, pseudo-measurement factors and event-topology coupling factors. The node variables represent the state quantities such as voltages of nodes in the power distribution network; the measurement factors reflect the observed values of node voltages and branch powers based on actual measurement data; the pseudo-measurement factors are used to supplement the case of insufficient measurement data and are generated by historical data or prediction data; and the event-topology coupling factors capture the influence of extreme events on the topology structure and operating state of the power distribution network. Then, the adaptive weight of the event-topology coupling factor is determined according to the extreme event index, which reflects the severity of the influence of the extreme event on the power distribution network. The higher the extreme event index, the greater the influence of the event-topology coupling factor on the reasoning process. Then, the node variables and factor nodes in the initial factor graph are adjusted according to the adaptive weight to obtain a variable structure factor graph corresponding to the real-time topology. By adjusting the connection relationship and weight of the node variables and factor nodes, the factor graph can adapt to the topology change and operating state change of the power distribution network under the extreme event. Finally, belief propagation reasoning is performed on the variable structure factor graph. The belief propagation reasoning is a distributed reasoning algorithm based on a probabilistic graph model, which gradually converges to the estimated values of node voltages and branch powers by transmitting messages in the factor graph.

[0070] In a preferred embodiment of the application, in a flood disaster caused by a heavy rain, part of the lines of the power distribution network are submerged, resulting in changes in the topology structure and abnormal node voltages and branch powers. First, an initial factor graph is generated according to the node-branch relationship of the real-time network model. The real-time network model reflects the connection relationship and parameters of the current power distribution network, and the initial factor graph constructed based thereon contains voltage variables of nodes, power variables of branches and corresponding measurement factors, pseudo-measurement factors and event-topology coupling factors. Next, the calculated extreme event index reflects the influence degree of the heavy rain flood on the power distribution network, and a large adaptive weight is assigned to the event-topology coupling factor according to the index, highlighting its importance in reasoning. Subsequently, the initial factor graph is adjusted according to the adaptive weight, changing the connection strength of the node variables and the event-topology coupling factors to form a variable structure factor graph. Finally, the belief propagation reasoning algorithm is used to perform iterative calculation on the variable structure factor graph. At the initial moment, the messages of node voltages are initialized according to the prior probability and measurement data, and the messages are constantly updated and transmitted as the iteration proceeds. After multiple rounds of iteration, the estimated values of the node voltages and branch powers are obtained, which present the actual operating state of the power distribution network under the heavy rain flood disaster, providing a key basis for subsequent fault diagnosis and recovery strategy formulation.

[0071] In the embodiments of the present application, by constructing a variable structure factor graph containing event-topology coupling factors, the influence of extreme events on the distribution network can be fully considered, making the state estimation more close to the actual operation. The generation of the initial factor graph provides a basic framework for reasoning, while the introduction of adaptive weights enhances the sensitivity to extreme events, enabling the reasoning process to dynamically respond to extreme events of different intensities. The formation of the variable structure factor graph further improves the adaptability of the model, enabling it to accurately reflect the real-time topology and operating state of the distribution network.

[0072] Optionally, the state evaluation result of the distribution network in the current time period is obtained according to the estimated values of the node voltages and the branch powers, comprising: According to the preset voltage safety interval and the branch power current carrying limit, the estimated values are scanned for over-limit, to generate node voltage deviation indicators and branch power over-limit indicators; The node voltage deviation indicators and the branch power over-limit indicators are mapped to a risk quantification model to obtain a comprehensive risk score of the distribution network in the current time period; According to the comprehensive risk score, the state evaluation result is obtained.

[0073] Specifically, first, according to the preset voltage safety interval and the branch power current carrying limit, the estimated values are scanned for over-limit. By comparing the voltage estimated value of each node with the preset voltage safety interval, the node voltage deviation indicator is calculated; at the same time, by comparing the power estimated value of each branch with the branch power current carrying limit, the branch power over-limit indicator is calculated. This step can accurately identify the nodes and branches in the distribution network where voltage and power appear abnormal, providing quantitative basis for subsequent risk assessment. Next, the node voltage deviation indicators and the branch power over-limit indicators are mapped to a risk quantification model. This model comprehensively considers the influence of voltage deviation and power over-limit on the safe operation of the distribution network, and calculates the comprehensive risk score of the distribution network in the current time period through specific mapping relationship and weight distribution. The mapping process usually involves normalization processing and weighted summation of different indicators to ensure that the comprehensive risk score can fully reflect the overall operation risk of the distribution network. Finally, according to the size of the comprehensive risk score, the state evaluation result of the distribution network is determined. Different risk level intervals are usually set, such as low risk, medium risk and high risk. The state evaluation result of the distribution network is the corresponding risk level when the comprehensive risk score falls in which interval, thereby providing intuitive and explicit decision basis for operation and maintenance personnel.

[0074] In a preferred embodiment of the present application, it is assumed that the estimated values of node voltages and branch powers of a power distribution network after a gale weather have been obtained through variable structure factor graph reasoning. The power distribution network has 100 nodes and 80 branches. First, according to the preset voltage safety interval (0.9 p.u.-1.1 p.u.) and the branch power current-carrying limit (based on the branch thermal limit power), the estimated values of all node voltages and branch powers are scanned for over-limit. It is found through scanning that the estimated values of voltages of 5 nodes are outside the voltage safety interval, of which the voltages of 3 nodes are too low (lower than 0.9 p.u.) and the voltages of 2 nodes are too high (higher than 1.1 p.u.); at the same time, the estimated values of powers of 3 branches exceed 80% of the branch power current-carrying limit. For these over-limit conditions, the node voltage deviation index and the branch power over-limit index are calculated. For the voltage deviation index, the following formula can be used for calculation: ; It is assumed that the total sum of the calculated voltage deviation index is 0.12. For the branch power over-limit index, the following formula can be used for calculation: ; wherein m is the total number of branches, and it is assumed that the total sum of the calculated branch power over-limit index is 0.08. Next, the two indexes are mapped to a risk quantification model. The risk quantification model can be set as: ; wherein and are the weights of the voltage deviation index and the branch power over-limit index, respectively, and it is assumed that , . Substituting the calculation obtains: .

[0075] According to the preset risk level interval (assuming low risk: 0-0.1, medium risk: 0.1-0.3, high risk: 0.3-1.0), the comprehensive risk score 0.128 falls in the medium risk interval. Therefore, the state evaluation result of the power distribution network in the current time period is medium risk. This result prompts the operation and maintenance personnel to further check and handle the over-limit nodes and branches to prevent the occurrence of potential faults and ensure the safe and stable operation of the power distribution network.

[0076] In the embodiment of the present application, through over-limit scanning, the voltage deviation and power over-limit problems existing in the power distribution network can be found in time to provide early warning for the prevention and handling of potential faults. Mapping the over-limit indexes to the risk quantification model enables different types of indexes to be integrated in a unified risk evaluation framework, avoiding the possible one-sidedness caused by single index evaluation.

[0077] In combination withFigure 2 As shown in the embodiments of the present application, the power distribution network evaluation system comprises: a data monitoring unit configured to obtain a measurement data stream of an operation state of a power distribution network in a current time period; an extreme event judging unit configured to obtain an extreme event index of the power distribution network in the current time period according to the measurement data stream, and judge whether the power distribution network has an extreme event according to the extreme event index; a data classification unit configured to, if yes, classify the measurement data stream through a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream; a matrix construction unit configured to construct an observation weight matrix of the extreme event according to data in the complete data segments; a completion operation unit configured to perform a weighted tensor completion operation on the missing data segments according to the observation weight matrix to obtain missing data of the missing data segments; a model establishing unit configured to dynamically update topology information of the power distribution network according to the complete data segments and the missing data, and obtain a real-time network model corresponding to the extreme event according to the topology information; an estimation unit configured to obtain estimation values of node voltages and branch powers of the power distribution network through a variable structure factor graph reasoning according to the real-time network model and the extreme event index; an evaluation unit configured to obtain a state evaluation result of the power distribution network in the current time period according to the estimation values of the node voltages and the branch powers.

[0078] The power distribution network evaluation system of the present application has the same advantages as the power distribution network evaluation method of the present application, and will not be repeated here.

[0079] Although the present application is disclosed as above, the protection scope of the present application is not limited to this. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application, and these changes and modifications will fall within the protection scope of the present application.

Claims

1. A distribution network evaluation method, characterized in that: include: Obtain the measurement data stream of the distribution network's operating status in the current time period; Obtaining an extreme event index of the distribution network in the current time period according to the measurement data stream, and determining whether an extreme event occurs in the distribution network according to the extreme event index; If yes, classify the measurement data stream by using a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream; constructing an observation weight matrix of the extreme event based on the data in the complete data segment; Performing a weighted tensor completion operation on the missing data segment according to the observation weight matrix to obtain missing data of the missing data segment; Dynamically updating topology information of the distribution network according to the complete data segment and the missing data, and obtaining a real-time network model corresponding to the extreme event according to the topology information; Obtaining estimated values ​​of node voltage and branch power of the distribution network through variable structure factor graph reasoning based on the real-time network model and the extreme event index; A state assessment result of the distribution network in a current time period is obtained according to the estimated values ​​of the node voltage and the branch power.

2. The distribution network evaluation method according to claim 1, characterized in that: Obtaining, based on the measurement data stream, an extreme event indicator of the distribution network in the current time period, includes: Extracting features from the measurement data stream to obtain meteorological disturbance feature quantities and topological change feature quantities; Comparing the meteorological disturbance characteristic quantity with a preset meteorological threshold value to obtain a meteorological disturbance intensity index; Comparing the topology change characteristic with a preset topology threshold to obtain a topology change rate indicator; The extreme event index is generated by weighted fusion of the meteorological disturbance intensity index and the topology change rate index.

3. The distribution network evaluation method according to claim 1, characterized in that: The determining, based on the extreme event indicator, whether an extreme event occurs in the distribution network includes: Comparing the extreme event indicator with a preset extreme event threshold range; If the extreme event indicator is within any of the threshold ranges, the preset extreme event corresponding to the threshold range is used as the extreme event of the distribution network; If the extreme event indicator is not within the threshold range of any of the preset extreme events, it is determined that the power distribution network is in a normal operating state.

4. The distribution network evaluation method according to claim 1, characterized in that: The method of classifying the measurement data stream by a data quality grading mechanism to obtain complete data segments and missing data segments in the measurement data stream includes: Obtaining a data quality level of each measurement point in the measurement data stream based on data integrity, timestamp continuity, and numerical rationality; Determining whether data exists in the measurement point according to the data quality level; wherein, when data exists in the measurement point, the measurement point is regarded as a complete data point; when data does not exist in the measurement point, the measurement point is regarded as a complete data point; The complete data segments and the missing data segments in the measurement data stream are determined according to the complete data points and the missing data points.

5. The distribution network evaluation method according to claim 4, characterized in that: The step of constructing an observation weight matrix of the extreme event based on the data in the complete data segment includes: Determining a first weight component negatively correlated with the extreme event indicator according to the numerical credibility, temporal confidence, and spatial correlation of each measurement point in the complete data segment; determining, according to the data quality level of each of the measurement points in the complete data segment, a second weight component positively correlated with the data quality level; Fusing the first weight component and the second weight component to obtain a comprehensive weight for each of the measurement points; The comprehensive weights are arranged according to the time dimension, the node dimension and the data type dimension to obtain the observation weight matrix.

6. The distribution network evaluation method according to claim 5, characterized in that: The performing a weighted tensor completion operation on the missing data segment according to the observation weight matrix to obtain missing data of the missing data segment includes: Taking time, node, and data type as three-dimensional coordinates, and mapping the complete data segment and the missing data segment into an initial tensor, wherein the initial tensor is set to zero at the missing position of the missing data; Applying a weighted mask to the initial tensor according to the observation weight matrix to obtain a weighted tensor completion objective function; Solving the global optimal solution of the tensor completion objective function through a convex optimization iterative algorithm to obtain a completed tensor; The missing data of the missing data segment is determined based on the value corresponding to the missing position in the completed tensor.

7. The distribution network evaluation method according to claim 1, characterized in that: The dynamically updating the topology information of the distribution network according to the complete data segment and the missing data, and obtaining a real-time network model corresponding to the extreme event according to the topology information, includes: fusing the complete data segment with the missing data to obtain a fused measurement data stream; Performing data extraction on the fused measurement data stream to generate a real-time topology of the current grid connection relationship of the distribution network; The real-time topology is compared with a pre-stored reference topology, and the line parameters, distributed power access points and load model of the distribution network are synchronously corrected according to the comparison results to obtain the real-time network model corresponding to the extreme event.

8. The distribution network evaluation method according to claim 7, characterized in that: The step of obtaining estimated values ​​of node voltage and branch power of the distribution network by using a variable structure factor graph inference based on the real-time network model and the extreme event index includes: generating an initial factor graph including node variables, measurement factors, pseudo-measurement factors, and event-topology coupling factors according to the node-branch relationship of the real-time network model; determining an adaptive weight of the event-topology coupling factor according to the extreme event indicator; Adjusting node variables and factor nodes in the initial factor graph according to the adaptive weights to obtain a variable structure factor graph corresponding to the real-time topology; Belief propagation inference is performed on the variable structure factor graph to obtain estimated values ​​of the node voltage and the branch power of the power distribution network.

9. The distribution network evaluation method according to claim 7, characterized in that: Obtaining a state assessment result of the distribution network in a current time period based on the estimated values ​​of the node voltage and the branch power includes: Performing an over-limit scan on the estimated value according to a preset voltage safety interval and a branch power carrying current limit to generate a node voltage deviation index and a branch power over-limit index; Mapping the node voltage deviation index and the branch power limit exceeding index to a risk quantification model to obtain a comprehensive risk score of the distribution network in the current time period; The status assessment result is obtained according to the comprehensive risk score.

10. A distribution network evaluation system, characterized in that: include: A data monitoring unit is used to obtain a measurement data stream of the operating status of the distribution network in the current time period; an extreme event judgment unit, configured to obtain an extreme event index of the distribution network in the current time period based on the measurement data stream, and judge whether an extreme event occurs in the distribution network based on the extreme event index; a data classification unit, configured to, if yes, classify the measurement data stream by a data quality classification mechanism to obtain complete data segments and missing data segments in the measurement data stream; A matrix construction unit, configured to construct an observation weight matrix of the extreme event based on the data in the complete data segment; a completion operation unit, configured to perform a weighted tensor completion operation on the missing data segment according to the observation weight matrix to obtain missing data of the missing data segment; a model building unit, configured to dynamically update topology information of the distribution network according to the complete data segments and the missing data, and obtain a real-time network model corresponding to the extreme event according to the topology information; an estimating unit, configured to obtain estimated values ​​of node voltage and branch power of the distribution network through variable structure factor graph reasoning based on the real-time network model and the extreme event index; An evaluation unit is used to obtain a state evaluation result of the distribution network in a current time period based on the estimated values ​​of the node voltage and the branch power.

Images