Power distribution network risk early warning method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of power distribution network operation risk early warning, in particular to a power distribution network risk early warning method and device, electronic equipment and a storage medium, and the method comprises the steps: firstly obtaining a plurality of first feeder current data queues, a plurality of first feeder voltage data queues and a bus voltage data queue; performing frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues and the bus voltage data queue to obtain a plurality of first current vectors, a plurality of first voltage vectors and a bus voltage vector; constructing the plurality of first current vectors, the plurality of first voltage vectors and the bus voltage vector into a first matrix, and performing dimension conversion on the first matrix through a first dimension reduction matrix to obtain a second matrix; and finally, substituting the second matrix into the first hyperplane equation, and carrying out risk early warning according to an output result of the first hyperplane equation. Early warning is carried out based on the hyperplane, and the probability of early warning omission is reduced.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation risk early warning technology, and in particular to a power distribution network risk early warning method, device, electronic equipment and storage medium. Background Technology

[0002] Distribution network risk early warning is a management system that monitors and analyzes the operating status of the distribution network, identifies potential faults or safety hazards in advance, and provides early warning information for operation and maintenance decisions. Its core objective is to reduce power outage time and reduce fault losses.

[0003] Currently, commonly used early warning methods include rule-based methods. The core logic of these methods is based on industry standards and operational experience to formulate clear "if-then" rules (e.g., "line load rate > 90% for 1 hour → yellow alert"). This method is typically applicable to simple, quantifiable risk scenarios such as equipment overload and common environmental impacts (heavy rain, icing). Its advantages include low development costs, transparent logic, ease of understanding for operations personnel, and no need for complex algorithm training.

[0004] At the same time, the rule-based approach has the drawback of requiring manual rule updates and being unable to adapt to changes in distribution network topology and new risks. This is because the rule-based approach cannot exhaust all complex scenarios (such as risks coupled with multiple factors), and is prone to "missed warnings".

[0005] Therefore, it is necessary to develop a method for early warning of risks in power distribution networks. Summary of the Invention

[0006] The present invention provides a method, device, electronic device and storage medium for early warning of distribution network risks, which solves the problem that existing methods for early warning of distribution network risks are prone to missing early warnings.

[0007] In a first aspect, embodiments of the present invention provide a method for early warning of risks in a power distribution network, comprising: Acquire multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues; Frequency domain feature extraction is performed on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues, and the bus voltage data queue respectively to obtain a plurality of first current vectors, a plurality of first voltage vectors, and a bus voltage vector; A first matrix is ​​constructed by multiple first current vectors, multiple first voltage vectors, and a bus voltage vector. The first matrix is ​​then transformed in dimension by a first dimension reduction matrix to obtain a second matrix. The second matrix is ​​substituted into the first hyperplane equation, and risk warning is performed based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

[0008] In one possible implementation, the step of performing frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues, and the bus voltage data queue to obtain a plurality of first current vectors, a plurality of first voltage vectors, and a bus voltage vector includes: Autocorrelation analysis was performed on the bus voltage data queue, and the fundamental frequency was determined based on the analysis results; Multiple harmonic frequencies are generated based on multiple harmonic values ​​and the fundamental frequency; For each data queue, frequency domain features are extracted using the multiple harmonic frequencies to obtain multiple frequency amplitudes, where each frequency amplitude corresponds to a harmonic frequency. For each data queue, the multiple frequency amplitudes obtained from the frequency domain feature analysis are arranged in a preset order of harmonic values ​​to obtain the first current vector, the first voltage vector, or the bus voltage vector.

[0009] In one possible implementation, performing autocorrelation analysis on the bus voltage data queue and determining the fundamental frequency based on the analysis results includes: Obtain multiple first hysteresis values; For each first lag value, the autocorrelation coefficient is calculated based on the bus voltage data queue and the first formula, wherein the first formula is:

[0010] In the formula, The correlation coefficient is... The first bus voltage data queue One data point, This represents the average value of the bus voltage data queue. The variance of the bus voltage data queue. The first lag value, This represents the total number of data entries in the bus voltage data queue. The first lag value with the largest autocorrelation coefficient is taken as the target lag value; The product of the target hysteresis value and the bus voltage data sampling interval is taken as the fundamental period, and the reciprocal of the fundamental period is taken as the fundamental frequency.

[0011] In one possible implementation, for each data queue, frequency domain feature extraction is performed using the multiple harmonic frequencies to obtain multiple frequency amplitudes, including: For each data queue, frequency domain features are extracted using the second formula and the multiple harmonic frequencies to obtain multiple frequency amplitudes, wherein the second formula is:

[0012] In the formula, For the first Each frequency amplitude, The total number of data in the queue. For the first in the queue One data point, It is a natural constant. Pi The imaginary unit, For the harmonic coefficient, This is the fundamental frequency.

[0013] In one possible implementation, the first dimensionality reduction matrix is ​​obtained through multiple first history matrices, including: Multiple second-dimensionality-reduced matrices and multiple first-historical matrices are obtained. Each first-historical matrix corresponds to an abnormal state label. The first-historical matrix is ​​constructed by extracting frequency domain features from multiple historical first feeder current data queues, multiple historical first feeder voltage data queues, and historical bus voltage data queues. For each second dimension-reduced matrix, the plurality of first historical matrices are multiplied by the second dimension-reduced matrix respectively, and the resulting plurality of second historical matrices are constructed into a set of historical matrices; Cluster each historical matrix set and perform label consistency verification on the obtained historical matrix classes to obtain a consistency indicator value, wherein the number of historical matrix classes is the same as the number of types of abnormal state labels; If the number of iterations is reached, the second dimension reduction matrix corresponding to the smallest consistency indicator value is used as the first dimension reduction matrix. Otherwise, the second dimension-reduced matrix corresponding to the smallest consistency indicator value is taken as the globally optimal matrix; Add each consistency indicator value to the indicator queue of the corresponding second dimension reduction matrix; Select the historical best matrix based on each indicator queue; Based on the global optimal matrix and multiple historical optimal matrices, the multiple second dimensionality reduction matrices are adjusted, and then the process jumps to the step of multiplying the multiple first historical matrices with the second dimensionality reduction matrix for each second dimensionality reduction matrix to obtain multiple second historical matrices and constructing a historical matrix set.

[0014] In one possible implementation, the step of performing label consistency verification on the plurality of historical matrix classes to obtain a consistency indicator value includes: For each historical matrix class, multiple first matrix labels are extracted and the multiple first matrix labels are grouped into a first label set, wherein the first matrix label is the label of the first historical matrix corresponding to the second historical matrix in the class; Extract the number of majority tags in each first tag set, and use that as the first count; The ratio of the sum of multiple first quantities to the total number of first matrix labels is used as the consistency indicator value.

[0015] In one possible implementation, adjusting the plurality of second dimensionality-reduced matrices based on the globally optimal matrix and a plurality of historical optimal matrices includes: Each second dimensionality reduction matrix is ​​adjusted according to the third formula, which is:

[0016] In the formula, The adjusted second dimensionality reduction matrix Line number Column elements, As the first coefficient, As the second coefficient, The bias constant is The first of the globally optimal matrix Line number Column elements, The historical best matrix of the th Line number Column elements, The second dimension reduction matrix before adjustment Line number The elements of the column.

[0017] In one possible implementation, the first hyperplane equation is constructed based on a plurality of second history matrices, including: Multiple second historical matrices are obtained, wherein the second historical matrices are obtained by reducing the dimensionality of the first historical matrices using the first dimensionality reduction matrix, and each second historical matrix corresponds to a label characterizing whether there is a risk in the distribution network; Each element in the second history matrix is ​​standardized, and the resulting third history matrix is ​​used to construct the first history vector. The label of the second history matrix is ​​used as the label of the first history vector, where the elements in the first history vector are distributed in the interval between 0 and 1. The process hyperplane equation is constructed based on the number of elements in the first history vector, wherein the process hyperplane equation is:

[0018] In the formula, For the first The first coefficient, The first history vector is the first One element, The constant coefficients, The number of elements in the first history vector; Based on the process hyperplane equation, the first history vector, and the label of the first history vector, a distance equation is constructed to represent the distance from the first history vector to the process hyperplane, wherein the distance equation is:

[0019] In the formula, Let be the distance from the first history vector to the hyperplane. The label for the first historical vector; Construct a target equation based on the distance equation, with the objective of maximizing the distance to the hyperplane of the vector that minimizes the distance to the hyperplane. The objective equation is solved based on the first historical vector and the label of the first historical vector to obtain a solution with multiple coefficients; Substituting the solutions of the plurality of coefficients into the process hyperplane, the equation of the first hyperplane is obtained.

[0020] In one possible implementation, substituting the second matrix into the first hyperplane equation and performing risk warning based on the output of the first hyperplane equation includes: Standardize each element in the second matrix and construct the first vector from the resulting third matrix. Substitute the first vector into the first hyperplane equation, and perform risk warning based on the output of the hyperplane equation.

[0021] In a second aspect, embodiments of the present invention provide a distribution network risk early warning device for implementing the distribution network risk early warning method as described in the first aspect or any possible implementation thereof, the distribution network risk early warning device comprising: The power grid operation data acquisition module is used to acquire multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues; The feature extraction module is used to perform frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues and the bus voltage data queue respectively to obtain a plurality of first current vectors, a plurality of first voltage vectors and a bus voltage vector. The dimension transformation module is used to construct a first matrix from multiple first current vectors, multiple first voltage vectors and bus voltage vectors, and to perform dimension transformation on the first matrix through a first dimension reduction matrix to obtain a second matrix. as well as, The risk warning module is used to substitute the second matrix into the first hyperplane equation and perform risk warning based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

[0022] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention discloses a risk early warning method for distribution networks. First, it acquires multiple first feeder current data queues, multiple first feeder voltage data queues, and a bus voltage data queue. Then, it performs frequency domain feature extraction on each of these data queues to obtain multiple first current vectors, multiple first voltage vectors, and a bus voltage vector. Next, it constructs a first matrix from these vectors and performs a dimensionality transformation using a first dimensionality reduction matrix to obtain a second matrix. Finally, it substitutes the second matrix into a first hyperplane equation and performs risk early warning based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state into two parts: one representing the normal state and the other representing the abnormal state. This invention reduces the information dimensionality through feature extraction and dimensionality transformation while retaining categorizable features. It classifies information based on a hyperplane and performs early warning based on the classification results. Compared to non-rule-based methods, this invention reduces the probability of missed warnings and has higher reliability. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the power distribution network risk early warning method provided by the embodiments of the present invention; Figure 2 This is a schematic diagram of the risk warning principle based on the first hyperplane equation provided by an embodiment of the present invention; Figure 3 This is a functional block diagram of the power distribution network risk early warning device provided in the embodiments of the present invention. Detailed Implementation

[0025] In the following description, specific details such as particular system structures and techniques are set forth for illustrative purposes and not for limitation, so as to provide a thorough understanding of embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0026] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0027] The embodiments of the present invention will be described in detail below. This example is implemented based on the technical solution of the present invention, and provides detailed implementation methods and specific operation processes. However, the protection scope of the present invention is not limited to the following embodiments.

[0028] Figure 1 A flowchart of a power distribution network risk early warning method provided for an embodiment of the present invention.

[0029] like Figure 1 As shown, a flowchart illustrating the implementation of the power distribution network risk early warning method provided by an embodiment of the present invention is presented, and is described in detail below: In step 101, multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues are acquired.

[0030] In step 102, frequency domain features are extracted from the plurality of first feeder current data queues, the plurality of first feeder voltage data queues, and the bus voltage data queue to obtain a plurality of first current vectors, a plurality of first voltage vectors, and a bus voltage vector.

[0031] In some implementations, the step of performing frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues, and the bus voltage data queue to obtain a plurality of first current vectors, a plurality of first voltage vectors, and a bus voltage vector includes: Autocorrelation analysis was performed on the bus voltage data queue, and the fundamental frequency was determined based on the analysis results; Multiple harmonic frequencies are generated based on multiple harmonic values ​​and the fundamental frequency; For each data queue, frequency domain features are extracted using the multiple harmonic frequencies to obtain multiple frequency amplitudes, where each frequency amplitude corresponds to a harmonic frequency. For each data queue, the multiple frequency amplitudes obtained from the frequency domain feature analysis are arranged in a preset order of harmonic values ​​to obtain the first current vector, the first voltage vector, or the bus voltage vector.

[0032] In some implementations, performing autocorrelation analysis on the bus voltage data queue and determining the fundamental frequency based on the analysis results includes: Obtain multiple first hysteresis values; For each first lag value, the autocorrelation coefficient is calculated based on the bus voltage data queue and the first formula, wherein the first formula is:

[0033] In the formula, The correlation coefficient is... The first bus voltage data queue One data point, This represents the average value of the bus voltage data queue. The variance of the bus voltage data queue. The first lag value, This represents the total number of data entries in the bus voltage data queue. The first lag value with the largest autocorrelation coefficient is taken as the target lag value; The product of the target hysteresis value and the bus voltage data sampling interval is taken as the fundamental period, and the reciprocal of the fundamental period is taken as the fundamental frequency.

[0034] In some implementations, for each data queue, frequency domain feature extraction is performed using the multiple harmonic frequencies to obtain multiple frequency amplitudes, including: For each data queue, frequency domain features are extracted using the second formula and the multiple harmonic frequencies to obtain multiple frequency amplitudes, wherein the second formula is:

[0035] In the formula, For the first Each frequency amplitude, The total number of data in the queue. For the first in the queue One data point, It is a natural constant. Pi The imaginary unit, For the harmonic coefficient, This is the fundamental frequency.

[0036] Exemplarily, this invention aims to provide a method for risk warning by extracting features from power grid operating status data—including a first feeder current data queue, multiple first feeder voltage data queues, and a bus voltage data queue—through frequency domain analysis, and then classifying these features. Specifically, this invention uses a first hyperplane equation to classify the extracted features, and risk classification and warning are achieved based on whether the features fall above or below the first hyperplane equation. The advantage of this approach is that it maintains good classification and warning performance even in unknown scenarios. Furthermore, the equation expression is relatively simple, making it suitable for scenarios requiring high real-time performance.

[0037] In the power distribution network described in this invention, the main power output end is connected to the busbar, while one end of the feeder is connected to the busbar and the other end is connected to the load. Since the power distribution network has multiple feeders, when this invention provides early warning for the power distribution network, it obtains feeder current data queues, feeder voltage data queues, and busbar voltage data queues. These data queues correspond to the same time period and are obtained by the monitoring equipment collecting data based on this time period and arranging it in chronological order.

[0038] In terms of frequency domain feature extraction, this invention first performs autocorrelation analysis on the bus voltage data queue, determines the fundamental frequency based on the analysis results, and then generates multiple harmonics based on the fundamental frequency. Note that these harmonics must satisfy the condition of being positive integers, for example, positive integers such as 1, 2, 3, 4, etc. are used as harmonic coefficients. These harmonic coefficients are multiplied by the fundamental frequency to obtain the harmonic frequencies. Frequency domain features are extracted from the data queue based on these harmonic frequencies. The extracted features are arranged in a predetermined order to obtain a vector.

[0039] Regarding the extraction of the fundamental frequency, this invention first initializes multiple first hysteresis values, and then, based on these hysteresis values, calculates the autocorrelation coefficient according to the bus voltage data queue and a first formula, wherein the first formula is:

[0040] In the formula, The correlation coefficient is... The first bus voltage data queue One data point, This represents the average value of the bus voltage data queue. The variance of the bus voltage data queue. The first lag value, This represents the total number of data entries in the bus voltage data queue.

[0041] In other words, using the first formula, each first lag value yields an autocorrelation coefficient. We select the largest first lag value from the autocorrelation coefficients as the target lag value. The product of the target lag value and the bus voltage data sampling interval is the fundamental period, and the reciprocal of the fundamental period is the fundamental frequency.

[0042] Regarding frequency domain feature extraction, this invention employs the second formula to process each data queue one by one:

[0043] In the formula, For the first Each frequency amplitude, The total number of data in the queue. For the first in the queue One data point, It is a natural constant. Pi The imaginary unit, For the harmonic coefficient, This is the fundamental frequency.

[0044] Using the above formula, we obtain multiple frequency amplitudes. These frequency amplitudes are arranged in order of the harmonic coefficients to obtain the corresponding vectors: the first current vector, the first voltage vector, or the bus voltage vector.

[0045] In step 103, a first matrix is ​​constructed by multiple first current vectors, multiple first voltage vectors, and bus voltage vector, and a second matrix is ​​obtained by performing a dimensionality transformation on the first matrix through a first dimensionality reduction matrix.

[0046] In some implementations, the first dimensionality reduction matrix is ​​obtained through multiple first history matrices, including: Multiple second-dimensionality-reduced matrices and multiple first-historical matrices are obtained. Each first-historical matrix corresponds to an abnormal state label. The first-historical matrix is ​​constructed by extracting frequency domain features from multiple historical first feeder current data queues, multiple historical first feeder voltage data queues, and historical bus voltage data queues. For each second dimension-reduced matrix, the plurality of first historical matrices are multiplied by the second dimension-reduced matrix respectively, and the resulting plurality of second historical matrices are constructed into a set of historical matrices; Cluster each historical matrix set and perform label consistency verification on the obtained historical matrix classes to obtain a consistency indicator value, wherein the number of historical matrix classes is the same as the number of types of abnormal state labels; If the number of iterations is reached, the second dimension reduction matrix corresponding to the smallest consistency indicator value is used as the first dimension reduction matrix. Otherwise, the second dimension-reduced matrix corresponding to the smallest consistency indicator value is taken as the globally optimal matrix; Add each consistency indicator value to the indicator queue of the corresponding second dimension reduction matrix; Select the historical best matrix based on each indicator queue; Based on the global optimal matrix and multiple historical optimal matrices, the multiple second dimensionality reduction matrices are adjusted, and then the process jumps to the step of multiplying the multiple first historical matrices with the second dimensionality reduction matrix for each second dimensionality reduction matrix to obtain multiple second historical matrices and constructing a historical matrix set.

[0047] In some implementations, the step of performing label consistency verification on the plurality of historical matrix classes to obtain a consistency indication value includes: For each historical matrix class, multiple first matrix labels are extracted and the multiple first matrix labels are grouped into a first label set, wherein the first matrix label is the label of the first historical matrix corresponding to the second historical matrix in the class; Extract the number of majority tags in each first tag set, and use that as the first count; The ratio of the sum of multiple first quantities to the total number of first matrix labels is used as the consistency indicator value.

[0048] In some implementations, adjusting the plurality of second dimensionality-reduced matrices based on the global optimal matrix and a plurality of historical optimal matrices includes: Each second dimensionality reduction matrix is ​​adjusted according to the third formula, which is:

[0049] In the formula, The adjusted second dimensionality reduction matrix Line number Column elements, As the first coefficient, As the second coefficient, The bias constant is The first of the globally optimal matrix Line number Column elements, The historical best matrix of the th Line number Column elements, The second dimension reduction matrix before adjustment Line number The elements of the column.

[0050] For example, although the data obtained from frequency domain feature extraction is significantly reduced compared to the original data, the total amount of data is still large due to the numerous feeders in the power grid. This large amount of data contains some noisy and redundant data, which can pose challenges to classification.

[0051] Therefore, after the aforementioned steps are completed, the present invention will construct a first matrix from multiple vectors. In one scenario, the vectors obtained in the aforementioned steps will be used as row vectors to construct the first matrix. When using this matrix for dimensionality reduction, the first matrix will be multiplied by the first dimensionality reduction matrix to obtain a second matrix. Due to the dimensionality reduction effect of the first dimensionality reduction matrix, the data will be greatly reduced. At the same time, the key features of the data need to be preserved to ensure the accuracy of subsequent classification.

[0052] We can see that the first dimensionality reduction matrix is ​​an important step in realizing data dimensionality transformation and key feature preservation. In fact, the first dimensionality reduction matrix of this invention is obtained through multiple first historical matrices. Each of these first historical matrices corresponds to an abnormal state label (such as a label indicating whether there is an anomaly in the distribution network). The first historical matrix is ​​constructed by extracting frequency domain features from multiple historical first feeder current data queues, multiple historical first feeder voltage data queues, and historical bus voltage data queues, and then constructing multiple historical first current vectors, multiple historical first voltage vectors, and multiple bus voltage vectors. In other words, the construction process of the first historical matrix is ​​the same as that of the first matrix.

[0053] When constructing the matrix, multiple second-dimensionality-reduced matrices are first initialized. For each of these second-dimensionality-reduced matrices, multiple first-historical matrices are multiplied by the second-dimensionality-reduced matrix respectively, and multiple second-historical matrices are constructed into a set of historical matrices.

[0054] Then, each historical matrix set is clustered, and the label consistency of the obtained historical matrix classes is verified to obtain a consistency indicator value. The number of historical matrix classes is the same as the number of abnormal state label types.

[0055] If the number of iterations is reached, the second dimension reduction matrix corresponding to the smallest consistency indicator value is used as the first dimension reduction matrix.

[0056] Otherwise, the second dimension-reduced matrix corresponding to the smallest consistency indicator value is taken as the global optimal matrix. Each consistency indicator value is added to the indicator queue of the corresponding second dimension-reduced matrix. The historical optimal matrix is ​​selected according to each indicator queue. Finally, the multiple second dimension-reduced matrices are adjusted according to the global optimal matrix and multiple historical optimal matrices. After the adjustment is completed, the above steps of multiplying multiple first historical matrices with the second dimension-reduced matrix respectively are repeated to obtain multiple second historical matrices to construct a historical matrix set.

[0057] The consistency indicator value in the above process is used to ensure that the data after dimensionality reduction has good separability. In calculating this value, this invention first extracts multiple first matrix labels for each historical matrix class and aggregates these multiple first matrix labels into a first label set. Then, it extracts the number of majority labels in each first label set as the first quantity. Finally, it calculates the ratio of the sum of multiple first quantities to the total number of first matrix labels, which is the consistency indicator value.

[0058] Regarding the optimization of the dimensionality reduction matrix, this invention utilizes the third formula:

[0059] In the formula, The adjusted second dimensionality reduction matrix Line number Column elements, As the first coefficient, As the second coefficient, The bias constant is The first of the globally optimal matrix Line number Column elements, The historical best matrix of the th Line number Column elements, The second dimension reduction matrix before adjustment Line number The elements of the column.

[0060] Through the above steps, we have reduced the dimensionality of the data while ensuring that the data is separable, thus greatly reducing the amount of data.

[0061] In step 104, the second matrix is ​​substituted into the first hyperplane equation, and a risk warning is given based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

[0062] In some implementations, the first hyperplane equation is constructed based on a plurality of second history matrices, including: Multiple second historical matrices are obtained, wherein the second historical matrices are obtained by reducing the dimensionality of the first historical matrices using the first dimensionality reduction matrix, and each second historical matrix corresponds to a label characterizing whether there is a risk in the distribution network; Each element in the second history matrix is ​​standardized, and the resulting third history matrix is ​​used to construct the first history vector. The label of the second history matrix is ​​used as the label of the first history vector, where the elements in the first history vector are distributed in the interval between 0 and 1. The process hyperplane equation is constructed based on the number of elements in the first history vector, wherein the process hyperplane equation is:

[0063] In the formula, For the first The first coefficient, The first history vector is the first One element, The constant coefficients, The number of elements in the first history vector; Based on the process hyperplane equation, the first history vector, and the label of the first history vector, a distance equation is constructed to represent the distance from the first history vector to the process hyperplane, wherein the distance equation is:

[0064] In the formula, Let be the distance from the first history vector to the hyperplane. The label for the first historical vector; Construct a target equation based on the distance equation, with the objective of maximizing the distance to the hyperplane of the vector that minimizes the distance to the hyperplane. The objective equation is solved based on the first historical vector and the label of the first historical vector to obtain a solution with multiple coefficients; Substituting the solutions of the plurality of coefficients into the process hyperplane, the equation of the first hyperplane is obtained.

[0065] In some implementations, substituting the second matrix into the first hyperplane equation and performing risk warning based on the output of the first hyperplane equation includes: Standardize each element in the second matrix and construct the first vector from the resulting third matrix. Substitute the first vector into the first hyperplane equation, and perform risk warning based on the output of the hyperplane equation.

[0066] For example, this invention classifies data indicating the presence or absence of risk using a first hyperplane. In application, the second matrix obtained in the aforementioned steps is substituted into the hyperplane equation, and warnings are issued based on the output data of the hyperplane equation.

[0067] In fact, each element in the second matrix needs to be standardized (after standardization, the elements are distributed in the range of 0-1). The elements in the obtained third matrix are arranged in a predetermined order to construct the first vector. These vectors are substituted into the first hyperplane equation. Based on the output of the first hyperplane equation, the risk is determined. The first hyperplane equation of this invention will eventually output a value with a sign. Usually, the sign indicates whether there is a risk. For example, a value less than 0 indicates that there is a risk, otherwise it indicates that it is relatively safe.

[0068] As we can see, the first hyperplane equation is key to achieving risk early warning. The first hyperplane equation of this invention is constructed through multiple second historical matrices. These second historical matrices are obtained by reducing the dimensionality of the first historical matrix using the first dimensionality reduction matrix. Each second historical matrix corresponds to a label characterizing whether there is a risk in the distribution network. In other words, the process of obtaining the second historical matrix is ​​the same as the process of obtaining the second matrix.

[0069] These second history matrices are first standardized, and the standardized third history matrix is ​​used to construct the first history vector. It can be seen that there is a one-to-one correspondence, and the label of the second history matrix is ​​the label of the first history vector. Similarly, due to the standardization, the elements in the first history vector are distributed in the range of 0 to 1.

[0070] Then, the process hyperplane equation is constructed based on the number of elements in the first history vector, wherein the process hyperplane equation is:

[0071] In the formula, For the first The first coefficient, The first history vector is the first One element, The constant coefficients, The number of elements in the first history vector.

[0072] In practice, when we use the above hyperplane equation, we only substitute the second vector into the part on the left side of the equals sign and determine whether there is a risk based on the output data.

[0073] Next, based on the process hyperplane equation, the first history vector, and the label of the first history vector, we construct a distance equation for the distance from the first history vector to the process hyperplane:

[0074] In the formula, Let be the distance from the first history vector to the hyperplane. The label for the first historical vector.

[0075] Next, we construct an objective equation based on the distance equation, aiming to maximize the distance to the hyperplane from the vector with the minimum distance to the hyperplane. We then solve the objective equation using the first historical vector and its label, obtaining solutions with multiple coefficients. Substituting these solutions into the process hyperplane yields the first hyperplane equation.

[0076] Figure 2 A schematic diagram illustrating the principle of hyperplane classification of two different types of data is shown. In the diagram, the first historical vector 203 labeled as risk and the first historical vector 202 not labeled as risk are separated by the first hyperplane 201. Regarding the construction of the objective equation, the objective equation of this invention strives to maximize the distance from the vectors on both sides of the diagram to the first hyperplane 201. One possible objective equation is:

[0077] In the above formula, For the value maximization function, It is a minimum value function.

[0078] The present invention provides an implementation method for a distribution network risk early warning method. First, it acquires multiple first feeder current data queues, multiple first feeder voltage data queues, and a bus voltage data queue. Then, it performs frequency domain feature extraction on each of these data queues to obtain multiple first current vectors, multiple first voltage vectors, and a bus voltage vector. Next, it constructs a first matrix from these vectors and performs a dimensionality transformation using a first dimensionality reduction matrix to obtain a second matrix. Finally, it substitutes the second matrix into a first hyperplane equation and performs risk early warning based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state into two parts: one representing the normal state and the other representing the abnormal state. This invention reduces the information dimensionality through feature extraction and dimensionality transformation while retaining categorizable features. It classifies information based on a hyperplane and performs early warning based on the classification results. Compared to non-rule-based methods, this invention reduces the probability of missed warnings and has higher reliability.

[0079] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0080] The following are embodiments of the apparatus of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0081] Figure 3This is a functional block diagram of the power distribution network risk early warning device provided in the embodiments of the present invention, with reference to... Figure 3 The power distribution network risk early warning device includes: a power grid operation data acquisition module 301, a feature extraction module 302, a dimension conversion module 303, and a risk early warning module 304, wherein: The power grid operation data acquisition module 301 is used to acquire multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues. Feature extraction module 302 is used to perform frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues and the bus voltage data queue respectively to obtain a plurality of first current vectors, a plurality of first voltage vectors and a bus voltage vector; The dimension transformation module 303 is used to construct a first matrix from multiple first current vectors, multiple first voltage vectors and bus voltage vectors, and to perform dimension transformation on the first matrix through a first dimension reduction matrix to obtain a second matrix. The risk warning module 304 is used to substitute the second matrix into the first hyperplane equation and perform risk warning based on the output of the first hyperplane equation, wherein the first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

[0082] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the aforementioned method embodiments, and will not be repeated here.

[0083] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0084] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0085] In the embodiments provided by this invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0086] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0087] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0088] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-described embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various methods and apparatus embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0089] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for early warning of risks in a power distribution network, characterized in that, include: Acquire multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues; Frequency domain feature extraction is performed on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues and the bus voltage data queue respectively to obtain a plurality of first current vectors, a plurality of first voltage vectors and a bus voltage vector; A first matrix is ​​constructed by multiple first current vectors, multiple first voltage vectors, and a bus voltage vector. The first matrix is ​​then transformed in dimension by a first dimension reduction matrix to obtain a second matrix. The second matrix is ​​substituted into the first hyperplane equation, and risk warning is performed based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

2. The distribution network risk early warning method according to claim 1, characterized in that, The step of performing frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues, and the bus voltage data queue to obtain a plurality of first current vectors, a plurality of first voltage vectors, and a bus voltage vector includes: Autocorrelation analysis was performed on the bus voltage data queue, and the fundamental frequency was determined based on the analysis results; Multiple harmonic frequencies are generated based on multiple harmonic values ​​and the fundamental frequency; For each data queue, frequency domain features are extracted using the multiple harmonic frequencies to obtain multiple frequency amplitudes, where each frequency amplitude corresponds to a harmonic frequency. For each data queue, the multiple frequency amplitudes obtained from the frequency domain feature analysis are arranged in a preset order of harmonic values ​​to obtain the first current vector, the first voltage vector, or the bus voltage vector.

3. The distribution network risk early warning method according to claim 2, characterized in that, The step of performing autocorrelation analysis on the bus voltage data queue and determining the fundamental frequency based on the analysis results includes: Obtain multiple first hysteresis values; For each first lag value, the autocorrelation coefficient is calculated based on the bus voltage data queue and the first formula, wherein the first formula is: In the formula, The correlation coefficient, The first bus voltage data queue One data point, This represents the average value of the bus voltage data queue. The variance of the bus voltage data queue. The first lag value, This represents the total number of data entries in the bus voltage data queue. The first lag value with the largest autocorrelation coefficient is taken as the target lag value; The product of the target hysteresis value and the bus voltage data sampling interval is taken as the fundamental period, and the reciprocal of the fundamental period is taken as the fundamental frequency.

4. The distribution network risk early warning method according to claim 3, characterized in that, For each data queue, frequency domain feature extraction is performed using the multiple harmonic frequencies to obtain multiple frequency amplitudes, including: For each data queue, frequency domain features are extracted using the second formula and the multiple harmonic frequencies to obtain multiple frequency amplitudes, wherein the second formula is: In the formula, For the first Each frequency amplitude, The total number of data in the queue. For the first in the queue One data point, It is a natural constant. Pi The imaginary unit, For the harmonic coefficient, This is the fundamental frequency.

5. The distribution network risk early warning method according to claim 1, characterized in that, The first dimensionality reduction matrix is ​​obtained through multiple first historical matrices, including: Multiple second-dimensionality-reduced matrices and multiple first-historical matrices are obtained. Each first-historical matrix corresponds to an abnormal state label. The first-historical matrix is ​​constructed by extracting frequency domain features from multiple historical first feeder current data queues, multiple historical first feeder voltage data queues, and historical bus voltage data queues. For each second dimension-reduced matrix, the plurality of first historical matrices are multiplied by the second dimension-reduced matrix respectively, and the resulting plurality of second historical matrices are constructed into a set of historical matrices; Cluster each historical matrix set and perform label consistency verification on the obtained historical matrix classes to obtain a consistency indicator value, wherein the number of historical matrix classes is the same as the number of types of abnormal state labels; If the number of iterations is reached, the second dimension reduction matrix corresponding to the smallest consistency indicator value is used as the first dimension reduction matrix. Otherwise, the second dimension-reduced matrix corresponding to the smallest consistency indicator value is taken as the globally optimal matrix; Add each consistency indicator value to the indicator queue of the corresponding second dimension reduction matrix; Select the historical best matrix based on each indicator queue; Based on the global optimal matrix and multiple historical optimal matrices, the multiple second dimensionality reduction matrices are adjusted, and then the process jumps to the step of multiplying the multiple first historical matrices with the second dimensionality reduction matrix for each second dimensionality reduction matrix to obtain multiple second historical matrices and constructing a historical matrix set.

6. The distribution network risk early warning method according to claim 5, characterized in that, The step of performing label consistency verification on the multiple historical matrix classes to obtain consistency indication values ​​includes: For each historical matrix class, multiple first matrix labels are extracted and the multiple first matrix labels are grouped into a first label set, wherein the first matrix label is the label of the first historical matrix corresponding to the second historical matrix in the class; Extract the number of majority tags in each first tag set, and use that as the first count; The ratio of the sum of multiple first quantities to the total number of first matrix labels is used as the consistency indicator value.

7. The distribution network risk early warning method according to claim 5, characterized in that, The step of adjusting the plurality of second dimensionality reduction matrices based on the global optimal matrix and the plurality of historical optimal matrices includes: Each second dimensionality reduction matrix is ​​adjusted according to the third formula, which is: In the formula, The adjusted second dimensionality reduction matrix Line number Column elements, As the first coefficient, As the second coefficient, The bias constant is The first of the globally optimal matrix Line number Column elements, The historical best matrix of the th Line number Column elements, The second dimensionality reduction matrix before adjustment Line number The elements of the column.

8. The distribution network risk early warning method according to any one of claims 1-7, characterized in that, The first hyperplane equation is constructed based on multiple second history matrices, including: Multiple second historical matrices are obtained, wherein the second historical matrices are obtained by reducing the dimensionality of the first historical matrices using the first dimensionality reduction matrix, and each second historical matrix corresponds to a label characterizing whether there is a risk in the distribution network; Each element in the second history matrix is ​​standardized, and the resulting third history matrix is ​​used to construct the first history vector. The label of the second history matrix is ​​used as the label of the first history vector, where the elements in the first history vector are distributed in the interval between 0 and 1. The process hyperplane equation is constructed based on the number of elements in the first history vector, wherein the process hyperplane equation is: In the formula, For the first The first coefficient, The first history vector is the first One element, The constant coefficients, The number of elements in the first history vector; Based on the process hyperplane equation, the first history vector, and the label of the first history vector, a distance equation is constructed to represent the distance from the first history vector to the process hyperplane, wherein the distance equation is: In the formula, Let be the distance from the first history vector to the hyperplane. The label for the first historical vector; Construct a target equation based on the distance equation, with the objective of maximizing the distance to the hyperplane of the vector that minimizes the distance to the hyperplane. The objective equation is solved based on the first historical vector and the label of the first historical vector to obtain a solution with multiple coefficients; Substituting the solutions of the plurality of coefficients into the process hyperplane, the equation of the first hyperplane is obtained.

9. The distribution network risk early warning method according to claim 8, characterized in that, The step of substituting the second matrix into the first hyperplane equation and performing risk warning based on the output of the first hyperplane equation includes: Standardize each element in the second matrix and construct the first vector from the resulting third matrix. Substitute the first vector into the first hyperplane equation, and perform risk warning based on the output of the hyperplane equation.

10. A power distribution network risk early warning device, characterized in that, For implementing the distribution network risk early warning method as described in any one of claims 1-9, the distribution network risk early warning device comprises: The power grid operation data acquisition module is used to acquire multiple first feeder current data queues, multiple first feeder voltage data queues, and bus voltage data queues; The feature extraction module is used to perform frequency domain feature extraction on the plurality of first feeder current data queues, the plurality of first feeder voltage data queues and the bus voltage data queue respectively, to obtain a plurality of first current vectors, a plurality of first voltage vectors and a bus voltage vector. The dimension transformation module is used to construct a first matrix from multiple first current vectors, multiple first voltage vectors and bus voltage vectors, and to perform dimension transformation on the first matrix through a first dimension reduction matrix to obtain a second matrix. as well as, The risk warning module is used to substitute the second matrix into the first hyperplane equation and perform risk warning based on the output of the first hyperplane equation. The first hyperplane equation divides the matrix representing the normal state and the matrix representing the abnormal state into two parts.

Images