A Power Grid Fault Inversion Method and System Based on Multi-Source Data Fusion

Abstract

This invention relates to the field of power grid fault inversion technology, specifically disclosing a power grid fault inversion method and system based on multi-source data fusion. The method includes creating a power grid map based on the power grid system, matching time and location labels, and inserting stable and instantaneous quantities into the power grid map as power grid features with time labels; performing self-identification on the power grid features with time labels to locate fault regions; performing fault inversion on the power grid system based on the fault regions to obtain fault inversion results; training a neural network model from the power grid features to the fault inversion results; judging the accuracy of the neural network model based on the self-identification process; and adjusting the application probability of the neural network model according to the accuracy. This invention performs two-dimensionalization of multi-source data, identifies the two-dimensional data, performs fault inversion, and trains the neural network model simultaneously during fault inversion, thereby improving fault inversion efficiency.

Description

Technical Field

[0001] This invention relates to the field of power grid fault inversion technology, specifically a power grid fault inversion method and system based on multi-source data fusion. Background Technology

[0002] The composition of PMU / WAMS and SCADA data, as well as the data storage and management methods of different platforms, still differ significantly, making integration difficult. This results in current measured data reuse primarily relying on a single monitoring platform. A single monitoring platform may suffer interference and contamination during data acquisition and storage, leading to errors and omissions. Considering the multiple types and levels of monitoring involved in power grid operation, fully integrating relevant monitoring data would enable cross-temporal and spatial verification and correction of operational data, effectively improving the quality of measured operational data and facilitating in-depth analysis of the multi-timescale dynamic characteristics of the power grid. Therefore, breakthroughs in the key technologies for PMU / WAMS and SCADA data integration are crucial for comprehensively characterizing the multi-timescale dynamic behavior of high-proportion renewable energy power grids in typical scenarios, and for accurately analyzing the control and operation characteristics and key mechanisms of new power systems.

[0003] Existing multi-source data fusion solutions often employ complex transcoding models to convert all data into data with the same structure for analysis. While this approach is feasible, the readability of the transcoding model is low, making it difficult for managers to assess the accuracy of fault inversion during the fault inversion phase. Summary of the Invention

[0004] The purpose of this invention is to provide a power grid fault inversion method and system based on multi-source data fusion to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A power grid fault inversion method based on multi-source data fusion, the method comprising:

[0007] The system acquires steady-state quantities containing time and location tags in the power grid system based on RTU units, and instantaneous quantities containing time and location tags based on PMU units; wherein, the data acquisition period of the RTU unit is on the order of seconds; the data acquisition period of the PMU unit is on the order of milliseconds; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit;

[0008] A power grid map is created based on the power grid system. Time and location labels are matched, and stable and instantaneous quantities are inserted into the power grid map as power grid features with time labels. The value of each point in the power grid map is related to the stable and instantaneous quantities.

[0009] The power grid features containing time tags are self-identified to locate fault areas. Based on the fault areas, fault inversion is performed on the power grid system to obtain fault inversion results.

[0010] A neural network model is trained from power grid characteristics to fault inversion results. The accuracy of the neural network model is judged based on the self-identification process, and the application probability of the neural network model is adjusted according to the accuracy.

[0011] For each inversion task, the power grid fault inversion results are output based on the application of a probabilistic random neural network model and a self-identification process.

[0012] As a further aspect of the present invention: the steps of acquiring steady-state quantities containing time and location tags in the power grid system based on RTU units, and acquiring instantaneous quantities containing time and location tags based on PMU units include:

[0013] The installation coordinates of the RTU unit are retrieved from the pre-defined facility registration database, and the installation coordinates are used as location tags.

[0014] The collected data of each RTU unit is obtained based on a preset data structure, and the collection time is recorded as a time stamp;

[0015] The installation coordinates of the PMU unit are retrieved from the pre-defined facility registration database, and the installation coordinates are used as location labels.

[0016] The collected data of each PMU unit is obtained based on a preset data structure, and the collection time is recorded as a time stamp.

[0017] The data acquisition cycle of the RTU unit is 3s, and the data acquisition cycle of the PMU unit is 20ms.

[0018] As a further aspect of the present invention: the step of creating a power grid map based on the power grid system, matching time and location labels, and inserting stable and instantaneous quantities into the power grid map as power grid features containing time labels includes:

[0019] Obtain information on the power grid system's line layout and create a power grid map based on a preset scale;

[0020] The standard time point is determined based on the data acquisition cycle, and the time labels of stable quantities and instantaneous quantities are registered based on the standard time point;

[0021] For data to be inserted into the power grid diagram, query the data insertion point corresponding to the location label in the power grid diagram based on the same scale;

[0022] Data is inserted into the power grid diagram based on the data insertion point, serving as a power grid feature.

[0023] The data insertion rules are as follows:

[0024] In the formula, Z(i,j) is the value at point (i,j) in the power grid diagram, (x k ,y k () represents the coordinates of the k-th data insertion point, dis 2 [(x k ,y k ),(i,j)] means (x k ,y k The square of the distance between (i,j) and (i,j), S(x) k ,y k α represents the rate of change of the data to be inserted into the power grid diagram corresponding to the k-th data insertion point. k The weight is the weight of the data to be inserted into the power grid corresponding to the kth data insertion point. The weight has only two values: one for instantaneous quantities and the other for stable quantities. The magnitude of the data insertion weight for instantaneous quantities is greater than the magnitude of the data insertion weight for stable quantities.

[0025] As a further aspect of the present invention: the steps of self-identifying the power grid features containing time tags, locating the fault region, and performing fault inversion on the power grid system based on the fault region to obtain the fault inversion result include:

[0026] Read the power grid features containing time tags, traverse the power grid features, and calculate the difference at each point in the power grid features;

[0027] Based on the differential marker boundary points, the power grid features are divided into regions;

[0028] Calculate the mean value for each region, obtain the mean threshold reached by the mean value, and query the fault type in the preset fault table;

[0029] The fault types of each region are statistically analyzed and used as the results of fault inversion.

[0030] The fault table includes a mean threshold item and a fault type item.

[0031] As a further aspect of the present invention: the operator for calculating the difference is:

[0032]

[0033] As a further aspect of the present invention: the neural network model for training power grid characteristics to fault inversion results, the step of judging the accuracy of the neural network model based on a self-identification process, and adjusting the application probability of the neural network model according to the accuracy includes:

[0034] Read the power grid characteristics and fault inversion results to create a sample set;

[0035] The sample set is divided into a training set and a test set according to a preset ratio. The neural network model from power grid characteristics to fault inversion results is trained based on the training set, and the accuracy of the neural network model is judged based on the test set.

[0036] The probability of applying a neural network model is determined based on its accuracy.

[0037] The present invention also provides a power grid fault inversion system based on multi-source data fusion, the system comprising:

[0038] The power grid data acquisition module is used to acquire steady-state quantities containing time and location tags in the power grid system based on RTU units, and to acquire instantaneous quantities containing time and location tags based on PMU units; wherein, the data acquisition cycle of the RTU unit is on the order of seconds; the data acquisition cycle of the PMU unit is on the order of milliseconds; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit;

[0039] The power grid feature generation module is used to create a power grid map based on the power grid system, match time labels and location labels, and insert stable and instantaneous quantities into the power grid map as power grid features with time labels; the value of each point in the power grid map is related to the stable and instantaneous quantities.

[0040] The fault self-identification module is used to self-identify the power grid characteristics containing time tags, locate the fault area, perform fault inversion on the power grid system based on the fault area, and obtain the fault inversion result.

[0041] The model training application module is used to train a neural network model from power grid characteristics to fault inversion results, judge the accuracy of the neural network model based on the self-identification process, and adjust the application probability of the neural network model according to the accuracy.

[0042] The inversion execution module is used to output the grid fault inversion results for each inversion task based on the application of a probabilistic random neural network model and a self-identification process.

[0043] As a further aspect of the present invention: the power grid data acquisition module includes:

[0044] The first location determination unit is used to query the installation coordinates of the RTU unit in the preset facility registration database, and use the installation coordinates as location labels;

[0045] The first unit is identified, which is used to acquire the data collected by each RTU unit based on a preset data structure, and record the acquisition time as a time stamp;

[0046] The second location determination unit is used to query the installation coordinates of the PMU unit in the preset facility registration database, and use the installation coordinates as location labels.

[0047] The second time determination unit is used to acquire the data collected by each PMU unit based on a preset data structure, and record the acquisition time as a time tag.

[0048] The data acquisition cycle of the RTU unit is 3s, and the data acquisition cycle of the PMU unit is 20ms.

[0049] As a further aspect of the present invention: the power grid feature generation module includes:

[0050] The power grid diagram creation unit is used to obtain the line layout information of the power grid system and create a power grid diagram based on a preset scale.

[0051] The time registration unit is used to determine the standard time point based on the data acquisition cycle, and to register the time labels of stable quantities and instantaneous quantities based on the standard time point;

[0052] The location matching unit is used to query the data insertion point corresponding to the location label in the power grid map based on the same scale for the data to be inserted into the power grid map.

[0053] The data insertion unit is used to insert data into the power grid diagram based on the data insertion point, as a power grid feature.

[0054] The data insertion rules are as follows:

[0055] In the formula, Z(i,j) is the value at point (i,j) in the power grid diagram, (x k ,y k () represents the coordinates of the k-th data insertion point, dis 2 [(x k ,y k ),(i,j)] means (x k ,y k The square of the distance between (i,j) and (i,j), S(x) k ,y k α represents the rate of change of the data to be inserted into the power grid diagram corresponding to the k-th data insertion point. k The weight is the weight of the data to be inserted into the power grid corresponding to the kth data insertion point. The weight has only two values: one for instantaneous quantities and the other for stable quantities. The magnitude of the data insertion weight for instantaneous quantities is greater than the magnitude of the data insertion weight for stable quantities.

[0056] As a further aspect of the present invention: the fault self-identification module includes:

[0057] The differential calculation unit is used to read the power grid features containing time tags, traverse the power grid features, and calculate the differences at each point in the power grid features.

[0058] Boundary point marking unit, used to perform regional segmentation of power grid features based on the differentially marked boundary points;

[0059] The type query unit is used to calculate the numerical mean of each region, obtain the mean threshold reached by the numerical mean, and query the fault type in the preset fault table.

[0060] The fault statistics unit is used to statistically analyze each region and its fault type, as the result of fault inversion.

[0061] The fault table includes a mean threshold item and a fault type item.

[0062] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention converts multi-source data into two dimensions, identifies the two-dimensional data, performs fault inversion, and trains a neural network model during fault inversion to improve the efficiency of fault inversion. Since the two-dimensional data can be integrated with the display process, it has high readability. After multiple applications, managers can easily judge whether the fault inversion results are correct, thereby indirectly improving the accuracy of the fault inversion process. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0064] Figure 1 This is a flowchart of a power grid fault inversion method based on multi-source data fusion.

[0065] Figure 2 This is the first sub-flowchart of the power grid fault inversion method based on multi-source data fusion.

[0066] Figure 3 This is the second sub-flowchart of the power grid fault inversion method based on multi-source data fusion.

[0067] Figure 4 This is the third sub-flowchart of the power grid fault inversion method based on multi-source data fusion.

[0068] Figure 5 This is the fourth sub-flowchart of the power grid fault inversion method based on multi-source data fusion.

[0069] Figure 6 This is a structural diagram of a power grid fault inversion system based on multi-source data fusion. Detailed Implementation

[0070] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0071] Figure 1 The flowchart below illustrates a power grid fault retrieval method based on multi-source data fusion. In this embodiment of the invention, a power grid fault retrieval method based on multi-source data fusion includes:

[0072] Step S100: Based on the RTU unit, steady-state quantities containing time and location tags in the power grid system are collected, and based on the PMU unit, instantaneous quantities containing time and location tags are obtained; wherein, the data acquisition period of the RTU unit is in the second range; the data acquisition period of the PMU unit is in the millisecond range; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit;

[0073] Step S200: Create a power grid map based on the power grid system, match time labels and location labels, and insert stable quantities and instantaneous quantities into the power grid map as power grid features with time labels; the value of each point in the power grid map is related to the stable quantity and instantaneous quantity;

[0074] Step S300: Self-identify the power grid features containing time tags, locate the fault area, perform fault inversion on the power grid system based on the fault area, and obtain the fault inversion result;

[0075] Step S400: Train a neural network model from power grid characteristics to fault inversion results, determine the accuracy of the neural network model based on the self-identification process, and adjust the application probability of the neural network model according to the accuracy.

[0076] Step S500: For each inversion task, based on the application of a probabilistic random neural network model and a self-identification process, output the power grid fault inversion results.

[0077] SCADA (Supervisory Control And Data Acquisition) system is a data acquisition and monitoring system. It is the foundation and core of the power grid dispatch automation system. It is responsible for collecting and processing various real-time and non-real-time data in the operation of the power system and is the main data source for various application software in the power grid dispatch center.

[0078] The WAMS system, or Wide Area Measurement System, employs synchronous phase angle measurement technology. By gradually deploying synchronous phase angle measurement units (PMUs) at key measurement points across the entire network, it achieves real-time, high-speed acquisition of the synchronous phase angle and other key power grid data. The acquired data is transmitted in real-time to the wide area monitoring master station system via the power dispatch data network, thereby providing real-time monitoring and analysis of the power grid under normal operation and accident disturbance conditions, and enabling timely acquisition and understanding of the dynamic process of power grid operation.

[0079] The differences between the two systems are actually quite significant, including:

[0080] The monitoring scope differs: SCADA uses RTU units that can only collect steady-state system quantities (such as V, I, P, etc.), and cannot collect phasor information of each bus or the power angle of the generators. WAMS, on the other hand, uses PMU units that utilize GPS signals to achieve synchronous data acquisition. In addition to steady-state quantities, the acquired signals include the phase angles of the voltage and current of each bus, and can also measure the power angle of each generator.

[0081] The monitoring of dynamic processes differs: SCADA systems collect data with long sampling periods (typically refreshed every 3 seconds) and the data is not synchronized, making it impossible for the dispatch center to track the dynamic status of the system's operation through this information. In contrast, PMUs can directly measure the amplitude and phase angle of node voltages and the power angle of generators at a more granular time scale (milliseconds). Substations equipped with PMUs have complete observability and can better monitor the dynamic processes of the power system.

[0082] The technical solution of this invention adopts an architecture in which the WAMS system is installed only in the most critical part of the power grid system, so as to accurately monitor this part of the system with fine WAMS measurement data, and use this fine data to improve the accuracy of SCADA data in other parts of the system.

[0083] In one example of the technical solution of this invention, a specific data acquisition and application scheme is provided. The steady-state quantities containing time and location tags in the power grid system are acquired based on the RTU unit, and the instantaneous quantities containing time and location tags are acquired based on the PMU unit. These two types of data have different data structures. This application adopts a two-dimensional form to unify them. Whether it is a steady-state quantity or an instantaneous quantity, the rate of change of the data acquired at adjacent time moments is calculated, and the fill color value is determined by the rate of change and filled into the power grid map determined by the power grid lines. At this time, all data are integrated into a single image.

[0084] Using existing image recognition algorithms, such as image partitioning algorithms, regions can be segmented in the image. Since the values ​​of each point in the image are related to the rate of change, by analyzing the values ​​of the points in each region, it can be determined whether there is a risk in each region. Regions with risks are regarded as fault regions. Based on the value of the points in the fault regions, the fault type can be determined, thereby obtaining the judgment status of all regions. The judgment status of all regions is used as the fault inversion result.

[0085] Based on this, the present invention records the fault inversion results and the data-containing power grid map (power grid features) used in the self-identification process (step S300) in real time, trains a neural network model from the power grid features to the fault inversion results, judges the accuracy of the neural network model based on the self-identification process, and adjusts the application probability of the neural network model according to the accuracy. In practical applications, each time a choice is made between the neural network model and the self-identification process. As the accuracy of the neural network model increases, the selection probability of the neural network model (the application probability mentioned above) will increase. At this time, the entire fault inversion process is almost completed based on the trained neural network model, which is extremely efficient.

[0086] Figure 2 The first sub-flowchart of the power grid fault inversion method based on multi-source data fusion includes the following steps: acquiring steady-state quantities containing time and location tags in the power grid system based on RTU units, and acquiring instantaneous quantities containing time and location tags based on PMU units.

[0087] Step S101: Query the installation coordinates of the RTU unit in the preset facility registration database, and use the installation coordinates as location labels;

[0088] Step S102: Acquire the collected data of each RTU unit based on the preset data structure, and record the acquisition time as a time stamp;

[0089] Step S103: Query the installation coordinates of the PMU unit in the preset facility registration database, and use the installation coordinates as location labels;

[0090] Step S104: Acquire the data collected by each PMU unit based on the preset data structure, record the acquisition time as a time stamp.

[0091] In one example of the technical solution of the present invention, the acquisition process of steady-state and instantaneous quantities is specifically defined. Steps S101 and S102 are used to acquire steady-state quantities, and steps S103 and S104 are used to acquire instantaneous quantities. During the acquisition process, the acquisition time is used as a time label and the installation coordinates are used as a position label. It is worth mentioning that the data acquisition period is specifically defined. The data acquisition period of the RTU unit is 3s and the data acquisition period of the PMU unit is 20ms.

[0092] Figure 3 The second sub-flowchart of the power grid fault inversion method based on multi-source data fusion includes the following steps: creating a power grid map based on the power grid system, matching time and location labels, and inserting stable and instantaneous quantities into the power grid map as power grid features with time labels.

[0093] Step S201: Obtain the power grid system's line layout information and create a power grid map based on a preset scale;

[0094] Step S202: Determine the standard time point based on the data acquisition cycle, and register the time labels of the stable quantity and the instantaneous quantity based on the standard time point;

[0095] Step S203: For the data to be inserted into the power grid diagram, query the data insertion point corresponding to the location label in the power grid diagram based on the same scale;

[0096] Step S204: Insert the data into the power grid diagram based on the data insertion point as a power grid feature.

[0097] The system acquires information on the power grid system's line layout and creates a power grid map based on a preset scale. The power grid map can be likened to a line drawing with black lines on a white background. Standard time points are determined based on the data acquisition cycle, such as 0s, 3s, 6s, and so on. Time tags are registered based on these standard time points because there may be some delays during actual data acquisition. The acquisition time may not be a standard time, and the difference between the acquisition times of two adjacent acquisition actions is not the data acquisition cycle (which should theoretically be the data acquisition cycle). This delay can be eliminated by registering the data tags using standard time points.

[0098] Furthermore, all data acquired at the same time are inserted into the power grid diagram to obtain the power grid diagram at that time. The time is included in the standard time point. For the data to be inserted into the power grid diagram, the data insertion point corresponding to the location label is queried in the power grid diagram based on the same scale. The data is inserted into the power grid diagram based on the data insertion point as a power grid feature. In layman's terms, the power grid diagram after inserting data into the power grid feature.

[0099] The data insertion rules are as follows:

[0100] In the formula, Z(i,j) is the value at point (i,j) in the power grid diagram, (x k ,y k () represents the coordinates of the k-th data insertion point, dis 2 [(x k ,y k ),(i,j)] means (x k ,y k The square of the distance between (i,j) and (i,j), S(y k ,y k α represents the rate of change of the data to be inserted into the power grid diagram corresponding to the k-th data insertion point. k The weight is the weight of the data to be inserted into the power grid corresponding to the kth data insertion point. The weight has only two values: one for instantaneous quantities and the other for stable quantities. The magnitude of the data insertion weight for instantaneous quantities is greater than the magnitude of the data insertion weight for stable quantities.

[0101] The principle of the data insertion rule is very simple. For each point, the influence of all data to be inserted on that point is calculated and accumulated. The influence is proportional to the rate of change of the data. The closer the data to be inserted is to the point, the greater the influence.

[0102] Figure 4 The third sub-flowchart of the power grid fault inversion method based on multi-source data fusion includes the following steps: self-identifying the power grid features containing time labels, locating the fault region, performing fault inversion on the power grid system based on the fault region, and obtaining the fault inversion result.

[0103] Step S301: Read the power grid features containing time tags, traverse the power grid features, and calculate the difference at each point in the power grid features;

[0104] Step S302: Divide the power grid features into regions based on the differential marker boundary points;

[0105] Step S303: Calculate the mean value of each region, obtain the mean threshold value reached by the mean value, and query the fault type in the preset fault table;

[0106] Step S304: Statistically analyze each region and its fault type to obtain the fault inversion result;

[0107] The fault table includes a mean threshold item and a fault type item.

[0108] The above content provides a specific self-identification scheme for power grid features. For each moment of power grid features, the system traverses each point in the power grid features and calculates the difference at each point in the power grid features. The difference represents the difference between a certain point and its surrounding points. When the difference reaches a preset threshold, the point is marked as a boundary point, thereby partitioning the power grid features.

[0109] After partitioning, the average value of all points in each region is calculated. Combined with the data filling process of the power grid diagram mentioned above, the average value reflects the degree of change of the collected data (including steady-state and instantaneous quantities) within a region. The average value is sufficient as a parameter for assessing the fault type. Staff can pre-calculate the average values ​​under different fault types to obtain the relationship between the average value and the fault type. This relationship is represented by a fault table. In the actual application stage, the fault type of each region can be obtained by querying the fault table based on the average value. It should be noted that the application of the fault table occurs in all regions, and the fault type in most regions is no fault.

[0110] Specifically, the operator for calculating the difference is:

[0111]

[0112] The operator is a Gaussian operator used to compute second-order differences.

[0113] Figure 5 The fourth sub-flowchart of the power grid fault inversion method based on multi-source data fusion includes the following steps: training the neural network model from power grid features to fault inversion results; judging the accuracy of the neural network model based on a self-identification process; and adjusting the application probability of the neural network model according to the accuracy.

[0114] Step S401: Read the power grid characteristics and fault inversion results, and create a sample set;

[0115] Step S402: Divide the sample set into a training set and a test set according to a preset ratio, train a neural network model from power grid characteristics to fault inversion results based on the training set, and judge the accuracy of the neural network model based on the test set;

[0116] Step S403: Determine the application probability of the neural network model based on the direct proportion of accuracy.

[0117] In one example of the technical solution of this invention, power grid characteristics and fault inversion results are read, a sample set is created, and the sample set is split into a training set and a test set in a 9:1 ratio. A neural network model from power grid characteristics to fault inversion results is trained based on the training set, and the accuracy of the neural network model is judged based on the test set. The application probability of the neural network model is determined by the accuracy. The simplest way is to directly use the accuracy as the application probability.

[0118] Figure 6 The diagram shows the structural composition of a power grid fault inversion system based on multi-source data fusion. In this embodiment of the invention, a power grid fault inversion system based on multi-source data fusion, system 10, includes:

[0119] The power grid data acquisition module 11 is used to acquire steady-state quantities containing time and location tags in the power grid system based on the RTU unit, and to acquire instantaneous quantities containing time and location tags based on the PMU unit; wherein, the data acquisition cycle of the RTU unit is on the order of seconds; the data acquisition cycle of the PMU unit is on the order of milliseconds; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit;

[0120] The power grid feature generation module 12 is used to create a power grid map based on the power grid system, match time labels and location labels, and insert stable quantities and instantaneous quantities into the power grid map as power grid features with time labels; the value of each point in the power grid map is related to the stable quantity and instantaneous quantity.

[0121] The fault self-identification module 13 is used to self-identify the power grid characteristics containing time tags, locate the fault area, perform fault inversion on the power grid system based on the fault area, and obtain the fault inversion result.

[0122] The model training application module 14 is used to train a neural network model from power grid characteristics to fault inversion results, judge the accuracy of the neural network model based on the self-identification process, and adjust the application probability of the neural network model according to the accuracy.

[0123] The inversion execution module 15 is used to output the power grid fault inversion results for each inversion task based on the application of a probabilistic random application neural network model and a self-identification process.

[0124] Furthermore, the power grid data acquisition module 11 includes:

[0125] The first location determination unit is used to query the installation coordinates of the RTU unit in the preset facility registration database, and use the installation coordinates as location labels;

[0126] The first unit is identified, which is used to acquire the data collected by each RTU unit based on a preset data structure, and record the acquisition time as a time stamp;

[0127] The second location determination unit is used to query the installation coordinates of the PMU unit in the preset facility registration database, and use the installation coordinates as location labels.

[0128] The second time determination unit is used to acquire the data collected by each PMU unit based on a preset data structure, and record the acquisition time as a time tag.

[0129] The data acquisition cycle of the RTU unit is 3s, and the data acquisition cycle of the PMU unit is 20ms.

[0130] Specifically, the power grid feature generation module 12 includes:

[0131] The power grid diagram creation unit is used to obtain the line layout information of the power grid system and create a power grid diagram based on a preset scale.

[0132] The time registration unit is used to determine the standard time point based on the data acquisition cycle, and to register the time labels of stable quantities and instantaneous quantities based on the standard time point;

[0133] The location matching unit is used to query the data insertion point corresponding to the location label in the power grid map based on the same scale for the data to be inserted into the power grid map.

[0134] The data insertion unit is used to insert data into the power grid diagram based on the data insertion point, as a power grid feature.

[0135] The data insertion rules are as follows:

[0136] In the formula, Z(i,j) is the value at point (i,j) in the power grid diagram, (x k ,y k () represents the coordinates of the k-th data insertion point, dis 2 [(x k ,y k ),(i,j)] means (x k ,y k The square of the distance between (i,j) and (i,j), S(x) k ,y k α represents the rate of change of the data to be inserted into the power grid diagram corresponding to the k-th data insertion point. k The weight is the weight of the data to be inserted into the power grid corresponding to the kth data insertion point. The weight has only two values: one for instantaneous quantities and the other for stable quantities. The magnitude of the data insertion weight for instantaneous quantities is greater than the magnitude of the data insertion weight for stable quantities.

[0137] In addition, the fault self-identification module 13 includes:

[0138] The differential calculation unit is used to read the power grid features containing time tags, traverse the power grid features, and calculate the differences at each point in the power grid features.

[0139] Boundary point marking unit, used to perform regional segmentation of power grid features based on the differentially marked boundary points;

[0140] The type query unit is used to calculate the numerical mean of each region, obtain the mean threshold reached by the numerical mean, and query the fault type in the preset fault table.

[0141] The fault statistics unit is used to statistically analyze each region and its fault type, as the result of fault inversion.

[0142] The fault table includes a mean threshold item and a fault type item.

[0143] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power grid fault inversion method based on multi-source data fusion, characterized in that, The method includes: The system acquires steady-state quantities containing time and location tags in the power grid system based on RTU units, and instantaneous quantities containing time and location tags based on PMU units; wherein, the data acquisition period of the RTU unit is on the order of seconds; the data acquisition period of the PMU unit is on the order of milliseconds; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit; A power grid map is created based on the power grid system. Time and location labels are matched, and stable and instantaneous quantities are inserted into the power grid map as power grid features with time labels. The value of each point in the power grid map is related to the stable and instantaneous quantities. The power grid features containing time tags are self-identified to locate fault areas. Based on the fault areas, fault inversion is performed on the power grid system to obtain fault inversion results. A neural network model is trained from power grid characteristics to fault inversion results. The accuracy of the neural network model is judged based on the self-identification process, and the application probability of the neural network model is adjusted according to the accuracy. For each inversion task, the power grid fault inversion results are output based on the application of a probabilistic random neural network model and a self-identification process. The steps of creating a power grid map based on the power grid system, matching time and location labels, and inserting stable and instantaneous quantities into the power grid map as power grid features with time labels include: Obtain information on the power grid system's line layout and create a power grid map based on a preset scale; The standard time point is determined based on the data acquisition cycle, and the time labels of stable quantities and instantaneous quantities are registered based on the standard time point; For data to be inserted into the power grid diagram, query the data insertion point corresponding to the location label in the power grid diagram based on the same scale; Data is inserted into the power grid diagram based on the data insertion point, serving as a power grid feature. The data insertion rules are as follows: In the formula, Points in the power grid diagram The value at that location, For the first The coordinates of the data insertion points express and The square of the distance, For the first The rate of change of the data to be inserted into the power grid diagram corresponding to each data insertion point. For the first The weight of the data to be inserted into the power grid diagram corresponding to each data insertion point has only two values: one corresponding to an instantaneous quantity and the other to a stable quantity. The magnitude of the data insertion weight for an instantaneous quantity is greater than the magnitude of the data insertion weight for a stable quantity.

2. The power grid fault inversion method based on multi-source data fusion according to claim 1, characterized in that, The steps of acquiring steady-state quantities containing time and location tags in the power grid system based on RTU units and acquiring instantaneous quantities containing time and location tags based on PMU units include: The installation coordinates of the RTU unit are retrieved from the pre-defined facility registration database, and the installation coordinates are used as location tags. The collected data of each RTU unit is obtained based on a preset data structure, and the collection time is recorded as a time stamp; The installation coordinates of the PMU unit are retrieved from the pre-defined facility registration database, and the installation coordinates are used as location labels. The collected data of each PMU unit is obtained based on a preset data structure, and the collection time is recorded as a time stamp. The data acquisition cycle of the RTU unit is 3s, and the data acquisition cycle of the PMU unit is 20ms.

3. The power grid fault inversion method based on multi-source data fusion according to claim 1, characterized in that, The steps of self-identifying the power grid features containing time tags, locating the fault region, and performing fault inversion on the power grid system based on the fault region to obtain the fault inversion result include: Read the power grid features containing time tags, traverse the power grid features, and calculate the difference at each point in the power grid features; Based on the differential marker boundary points, the power grid features are divided into regions; Calculate the mean value for each region, obtain the mean threshold reached by the mean value, and query the fault type in the preset fault table; The fault types of each region are statistically analyzed and used as the results of fault inversion. The fault table includes a mean threshold item and a fault type item.

4. The power grid fault inversion method based on multi-source data fusion according to claim 3, characterized in that, The operator for calculating the difference is: 。 5. The power grid fault inversion method based on multi-source data fusion according to claim 1, characterized in that, The steps of training the neural network model from power grid characteristics to fault inversion results, determining the accuracy of the neural network model based on a self-identification process, and adjusting the application probability of the neural network model according to the accuracy include: Read the power grid characteristics and fault inversion results to create a sample set; The sample set is divided into a training set and a test set according to a preset ratio. The neural network model from power grid characteristics to fault inversion results is trained based on the training set, and the accuracy of the neural network model is judged based on the test set. The probability of applying a neural network model is determined based on its accuracy.

6. A power grid fault inversion system based on multi-source data fusion, characterized in that, The system includes: The power grid data acquisition module is used to acquire steady-state quantities containing time and location tags in the power grid system based on RTU units, and to acquire instantaneous quantities containing time and location tags based on PMU units; wherein, the data acquisition cycle of the RTU unit is on the order of seconds; the data acquisition cycle of the PMU unit is on the order of milliseconds; the RTU unit is a remote terminal unit, and the PMU unit is a synchronous phasor measurement unit; The power grid feature generation module is used to create a power grid map based on the power grid system, match time labels and location labels, and insert stable and instantaneous quantities into the power grid map as power grid features with time labels; the value of each point in the power grid map is related to the stable and instantaneous quantities. The fault self-identification module is used to self-identify the power grid characteristics containing time tags, locate the fault area, perform fault inversion on the power grid system based on the fault area, and obtain the fault inversion result. The model training application module is used to train a neural network model from power grid characteristics to fault inversion results, judge the accuracy of the neural network model based on the self-identification process, and adjust the application probability of the neural network model according to the accuracy. The inversion execution module is used to output the power grid fault inversion results for each inversion task based on the application of a probabilistic random neural network model and a self-identification process. The power grid feature generation module includes: The power grid diagram creation unit is used to obtain the line layout information of the power grid system and create a power grid diagram based on a preset scale. The time registration unit is used to determine the standard time point based on the data acquisition cycle, and to register the time labels of stable quantities and instantaneous quantities based on the standard time point; The location matching unit is used to query the data insertion point corresponding to the location label in the power grid map based on the same scale for the data to be inserted into the power grid map. The data insertion unit is used to insert data into the power grid diagram based on the data insertion point, as a power grid feature. The data insertion rules are as follows: In the formula, Points in the power grid diagram The value at that location, For the first The coordinates of the data insertion points express and The square of the distance, For the first The rate of change of the data to be inserted into the power grid diagram corresponding to each data insertion point. For the first The weight of the data to be inserted into the power grid diagram corresponding to each data insertion point has only two values: one corresponding to an instantaneous quantity and the other to a stable quantity. The magnitude of the data insertion weight for an instantaneous quantity is greater than the magnitude of the data insertion weight for a stable quantity.

7. The power grid fault inversion system based on multi-source data fusion according to claim 6, characterized in that, The power grid data acquisition module includes: The first location determination unit is used to query the installation coordinates of the RTU unit in the preset facility registration database, and use the installation coordinates as location labels; The first unit is identified, which is used to acquire the data collected by each RTU unit based on a preset data structure, and record the acquisition time as a time stamp; The second location determination unit is used to query the installation coordinates of the PMU unit in the preset facility registration database, and use the installation coordinates as location labels. The second time determination unit is used to acquire the data collected by each PMU unit based on a preset data structure, and record the acquisition time as a time tag. The data acquisition cycle of the RTU unit is 3s, and the data acquisition cycle of the PMU unit is 20ms.

8. The power grid fault inversion system based on multi-source data fusion according to claim 6, characterized in that, The fault self-identification module includes: The differential calculation unit is used to read the power grid features containing time tags, traverse the power grid features, and calculate the differences at each point in the power grid features. Boundary point marking unit, used to perform regional segmentation of power grid features based on the differentially marked boundary points; The type query unit is used to calculate the numerical mean of each region, obtain the mean threshold reached by the numerical mean, and query the fault type in the preset fault table. The fault statistics unit is used to statistically analyze each region and its fault type, as the result of fault inversion. The fault table includes a mean threshold item and a fault type item.

Images