Distribution network load transaction early warning method and device

Abstract

The invention provides a distribution network load transaction early warning method and device, and relates to the technical field of load transaction early warning, and the method comprises the steps: collecting the power utilization parameters of each node in a distribution network in real time, forming a whole network node state matrix at the current moment, carrying out the sliding updating of the node state matrix in a continuous time window, and calculating the track entropy of different nodes; and constructing a state confluence area, identifying a complete abnormal event propagation chain map, and carrying out cross validation on the abnormal event propagation chain map and a preset impact load characteristic rule. Space-time correlation characteristics of load evolution can be dynamically captured based on a whole network node state matrix and trajectory entropy calculation; through state confluence area identification and abnormal propagation chain atlas backtracking, a source node of electric vehicle load aggregation can be accurately positioned, and an abnormal diffusion path and an influence range can be tracked; a set cross validation mechanism can effectively distinguish different types of impact load conditions such as electric vehicle charging and capacitor switching.

Description

Technical Field

[0001] The present invention relates to the field of load abnormal movement warning, and specifically to a method and device for warning of abnormal movement of distribution network load. Background Art

[0002] With the large-scale access of electric vehicles, the distribution network faces severe challenges from new impact loads. Traditional load anomaly detection methods mainly rely on the threshold judgment of single-node voltage and current, and it is difficult to capture the multi-node spatio-temporal coupling power mutation caused by centralized charging of electric vehicles. Therefore, timely and effectively identifying load abnormal movements in the distribution network has become an urgent problem to be solved in the current electric vehicle industry.

[0003] In the prior art, the publication number CN106849356A discloses a method of compiling a distribution network load abnormal movement determination module based on the OPEN3000 main network control system to automatically search and determine the distribution network load abnormal movement circuit. However, this method does not consider the application scenario of electric vehicle connection. The spatio-temporal randomness of electric vehicle access leads to highly dynamic load impact characteristics. The existing steady-state threshold rules cannot distinguish normal fluctuations from real anomalies. The electrical coupling relationship between distribution network nodes is complex, and local anomalies are prone to spread through topological paths, lacking the ability to completely trace the abnormal propagation chain; impact load events are often accompanied by active-reactive collaborative fluctuations, and traditional single-index detection of current or power is prone to false alarms due to noise interference. The above problems make it difficult for the distribution network to quickly locate and accurately classify abnormal events in the high-penetration scenario of electric vehicles, restricting the active warning and safety control capabilities. Therefore, it is necessary to consider multi-dimensional parameters to dynamically capture load abnormal movement situations.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and device for warning of abnormal movement of distribution network load to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: A method for warning of abnormal movement of distribution network load, the specific steps include: S1: Real-time collect the power consumption parameters of each node in the distribution network, the power consumption parameters include voltage, active power, reactive power and current, construct a multi-dimensional state vector for each node, and form a state matrix of all network nodes at the current moment with the node topological number as the index, and represent the overall operation state of the distribution system at a specific moment through the state matrix; S2: Sliding update of the node state matrix in a continuous time window, generating the state evolution trajectory of each node in the time window to construct a node trajectory set, and calculating the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the load state of the distribution network; S3: Performing aggregation analysis on the state trajectory set to identify the aggregation of multiple node trajectories in space and time. When a preset aggregation threshold is met, the corresponding area is determined as a state confluence area, where the state confluence area is an abnormal load aggregation area where electric vehicles are concentrated. S4: trace back each node in the state confluence area and its time evolution trajectory, identify the starting node, expansion path and impact boundary of the abnormal evolution, and build a complete abnormal event propagation chain map, which includes the abnormal starting node, abnormal impact chain and event propagation duration; S5: Cross-validate the abnormal event propagation chain map with the preset impact load characteristic rules. When the rule matching conditions are met, confirm that the abnormal type is the impact load abnormality caused by the concentrated access of electric vehicles.

[0007] Furthermore, the node status matrix of the entire network at the current moment is formed. The specific process is: by The power consumption parameters are sampled at the sampling interval, and the number of sampling points is set to , continuously collect the power consumption parameters of each node, where the topology number of each node According to the step-by-step increasing coding from the power supply end to the load end; each sampling point in the time window , the node Voltage , Active Power , reactive power And current The four-dimensional state vectors are stacked vertically in the order of node numbers to form the node state matrix of the entire network. , whose dimensions are ,The rows correspond to nodes, and the columns represent the voltage, active power, reactive power and current parameters respectively.

[0008] Furthermore, generating the state evolution trajectory of each node in the time window to construct a node trajectory set includes the following steps: Sliding update of the node state matrix of the entire network is performed to construct the power coefficients of current and voltage as well as the power coefficients of active power and reactive power, and the dimension reduction of the state evolution trajectory equation is performed to construct a node trajectory set: ; ; in, represents the steepness coefficient of the electric power coefficient; Indicates voltage and current electric quantity coefficient; Indicates active power and reactive power power coefficient; The set of node trajectories after dimensionality reduction is: ; wherein, Indicates the node trajectory matrix.

[0009] Furthermore, calculating the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the distribution network load state includes the following steps: For each parameter column in, calculate the mean and standard deviation, and the standardized trajectory matrix is: ; wherein, Indicates the standardized electric quantity coefficient; Indicates the standardized power coefficient; Calculate the covariance matrix: ; wherein, Indicates the transpose matrix of; Indicates the th node's covariance matrix; Solve the eigenvalues and eigenvectors of the covariance matrix: ; wherein, Indicates the th eigenvalue of the electrical parameter; Indicates the th eigenvector of the electrical parameter; Calculate the trajectory entropy: ; wherein, ; wherein, Indicates the parameter variable index; Indicates the node trajectory entropy; Indicates the th normalized weight of the eigenvalue; If , it indicates that the electrical parameters tend to a steady state; If It indicates that the electrical parameters tend to be abnormal.

[0010] Furthermore, identifying the aggregation situation of multiple node trajectories in space and time includes the following steps: Obtain the trajectory entropy sequences of all nodes within the time window and standardize all sequences: ; where represents the mean value of the trajectory entropy of node within the time window; represents the standard deviation of the trajectory entropy of node within the time window; represents the value after standardization; Calculate the dynamic similarity score of different node and sequences using the mutual information entropy method : ; where, if the score tends to 1, it indicates that the evolution patterns of the two nodes are highly synchronized; if the score tends to 0, it indicates that the evolution patterns are irrelevant; represents the mutual information between and represents the information entropy; represents the information entropy; represents the value after standardization; Based on the dynamic similarity score construct a matrix: and, based on the power grid topology, define the spatial proximity relationship of nodes and construct an adjacency matrix where represents that node is directly connected to otherwise ; Combine the similarity matrix with the adjacency matrix to construct a fused spatio-temporal feature matrix: ; Divide the node groups through the spectral clustering algorithm: , represents candidate confluence regions; represents element-wise multiplication; represents node and node The spatio-temporal fusion matrix; Indicates the serial number of the candidate confluence area; For each cluster , calculate the comprehensive score: ; Where, ; ; ; Where, Indicates the comprehensive score of the th cluster; Indicates the time consistency strength; Indicates the spatial aggregation degree; Indicates the weight of the time consistency strength; Indicates the weight of the spatial aggregation degree; Indicates the total number of nodes included in the group ; Indicates the spatial distance between nodes; If , then mark as the status confluence area, where Indicates the preset comprehensive score threshold.

[0011] Furthermore, construct a complete abnormal event propagation chain graph, which specifically includes the following steps: For each node in the status confluence area , locate the abnormal starting point through the following steps: Calculate the local mutation index of node in the th window: ; Where, Indicates the local mutation index of node in the th window; Indicates the mean trajectory entropy of node in the th window; Indicates the trajectory entropy standard deviation of the sequence composed of all trajectory entropies from the th window to the th window to the th window of node Indicates the window number index; If , then mark as triggering an abnormality in the th window; Denote the preset local mutation threshold and verify it using Granger causality test of whether the time series significantly affects other nodes; Establish dynamic propagation rules for extended path tracing. The propagation from node to node needs to satisfy: ; Denote the number of windows passed from node to node ; Denote the power grid fluctuation propagation speed; Denote node and node 's minimum topological path length; Denote the window time normalization factor; Denote the sampling interval; Denote the number of sampling points; If node is marked as abnormal within the window, and the trajectory entropy correlation coefficient between windows satisfies: ; Then mark as an effective propagation path, Denote the mean value of the normalized trajectory entropy of node within the th window; Denote the normalized trajectory entropy of node within the th time window; Denote the Pearson correlation coefficient; Iteratively expand until no new nodes meet the conditions; If node meets the following conditions during the extended path tracing: ; where, Denote the peak value of the trajectory entropy of node in all windows; Denote the preset trajectory entropy boundary threshold; Denote the total number of time windows.

[0012] Furthermore, cross-validate the abnormal event propagation chain map with the preset impact load characteristic rules, including the following steps: Check from the abnormal propagation chain map whether the preset impact load characteristic rules are triggered. The preset impact load characteristic rules include multi-node power surge, voltage linkage, and double-index jump; Set the set of propagation chain nodes as: , and the set of propagation chain paths as: ; is the total number of nodes in the propagation chain; The multi-node power surge includes counting the number of active power surges of the nodes in the propagation chain: ; Among them, ; represents the minimum number of nodes triggering the active power surge rule; represents the indicative function; represents the node at the th sampling point, the difference between the active power and the active power of the previous sampling point; represents the node at the th sampling point, the active power; represents the point at the th sampling point, the active power; represents the multi-node surge evaluation index value; The voltage linkage includes, for the propagation path , calculating the correlation coefficient and the phase difference of the node voltages: ; represents the Pearson correlation coefficient between the node and the node ; represents the voltage fluctuation phase difference; represents the path ratio threshold; represents the voltage linkage evaluation index value; The double-index jump includes, for each node in the propagation chain, detecting its active power and reactive power , ; Among them, represents the proportion threshold of active power and reactive power; represents the node proportion threshold; represents the double-index jump evaluation index value; Count the number of nodes and the path ratio in the propagation chain that meet the above preset impact load characteristic rules, and perform confidence calculation. The calculation formula is as follows: ; Among them, represents the index mean; if , it is determined that the propagation chain is triggered by an impact load, and a warning is triggered.

[0013] The present invention further provides a warning device for abnormal load changes in a distribution network. The warning device is used to execute the above warning method, including: A data acquisition unit, configured to collect the power consumption parameters of each node in the distribution network in real time. The power consumption parameters include voltage, active power, reactive power, and current, construct a multi-dimensional state vector for each node, and form a state matrix of all network nodes at the current moment with the node topology number as the index, and represent the overall operating state of the distribution system at a specific moment through the state matrix; A trajectory identification unit, configured to perform sliding updates on the node state matrix within a continuous time window, generate the state evolution trajectory of each node within the time window, construct a node trajectory set, and calculate the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the load state of the distribution network; A node aggregation unit, configured to perform aggregation analysis on the node trajectory set, identify the aggregation situation of multiple node trajectories in space and time, and when the preset aggregation degree threshold is met, determine the corresponding area as a state confluence area, and the state confluence area is an abnormal load aggregation area where electric vehicles are concentratedly connected; An abnormal backtracking unit, configured to backtrack each node and its time evolution trajectory in the state confluence area, identify the starting node, expansion path, and influence boundary of abnormal evolution, and construct a complete abnormal event propagation chain map, and the map includes an abnormal starting node, an abnormal influence chain, and an event propagation duration; A mobile analysis unit, which cross-verifies the abnormal event propagation chain map with the preset impact load characteristic rules, and when the rule matching condition is met, confirms that the abnormal type is an impact load abnormality caused by the concentrated access of electric vehicles.

[0014] Compared with the prior art, the beneficial effects of the present invention are: Through the construction of a multi-dimensional state matrix, trajectory entropy analysis, and abnormal propagation chain modeling, the present invention systematically solves the problem of detecting the impact load of electric vehicles. Based on the state matrix of all network nodes and the calculation of trajectory entropy in a sliding time window, the spatio-temporal correlation characteristics of load evolution can be dynamically captured, significantly improving the sensitivity to the disordered charging - orderly propagation mode of electric vehicle clusters; through the identification of the state confluence area and the backtracking of the abnormal propagation chain map, the source node of the electric vehicle load aggregation can be accurately located, and the abnormal diffusion path and influence range can be traced; by integrating the jump rules of active and reactive power double indicators, voltage linkage characteristics, and the cross-verification mechanism of sudden power increase of multiple nodes, different types of impact loads such as electric vehicle charging and capacitor switching can be effectively distinguished, and the risk of misoperation of protection caused by the disordered access of electric vehicles can be reduced. Description of the Drawings

[0015] Figure 1 This is a schematic flow chart of the distribution network load abnormal movement warning method in the present invention; Figure 2 This is a schematic diagram of multi-node power sudden increase data in the distribution network load abnormal movement warning in the present invention; Figure 3 This is a schematic diagram of voltage linkage data in the distribution network load abnormal movement warning in the present invention; Figure 4 This is a schematic diagram of double-index jump data in the distribution network load abnormal movement warning in the present invention; Figure 5 This is a structural block diagram of the distribution network load abnormal movement warning device in the present invention. Detailed implementation manners

[0016] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0017] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right" are only used to represent relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0018] Embodiment: Please refer to Figures 1-4 , the present invention provides a technical solution: A distribution network load abnormal movement warning method, the specific steps include: S1: Real-time collect the power consumption parameters of each node in the distribution network, the power consumption parameters include voltage, active power, reactive power and current, construct a multi-dimensional state vector for each node, and form a network-wide node state matrix at the current moment with the node topology number as the index, and represent the overall operating state of the distribution system at a specific moment through the state matrix; The specific process of forming the network-wide node state matrix at the current moment is: At sampling interval, sample the power consumption parameters, and set the number of sampling points to , continuously collect the power consumption parameters of each node, where the topological number of each node is encoded incrementally from the power supply end to the load end; at each sampling point within the time window , the voltage , active power , reactive power , and current constitute a four-dimensional state vector and are stacked vertically in the order of node numbers to form a network-wide node state matrix , whose dimension is , with its rows corresponding to nodes and its columns representing voltage, active power, reactive power, and current parameters in sequence.

[0019] The topological number of each node is encoded according to its spatial position in the power grid hierarchy, specifically: The topological number of each node can be sequentially encoded according to the physical hierarchy structure of the power grid from the power supply end to the load end, that is, starting from the main substation or power generation node on the power supply side, and sequentially assigning increasing integer numbers to the physical positions of downstream distribution transformers, branch lines, and finally the end-user load along the electrical connection path, so that nodes with smaller numbers are closer to the upstream of the power supply hierarchy, nodes with larger numbers are located downstream of the load hierarchy, and adjacent feeder nodes at the same level use consecutive numbers.

[0020] The node is specifically a topological unit in the distribution network that can independently measure power consumption parameters, and the power consumption parameters are specifically the current passing through the node, the voltage at both ends, active power, and reactive power; The formation of the network-wide node state matrix at the current moment, the specific logic is: At time , at each monitoring point, 4 power consumption parameters are collected, and the node state vector can be expressed as: ; Among them, represents the node topological number; represents the th sampling point voltage of the th node; represents the th node th sampling point active power; represents the th sampling point reactive power of the The state vector of the th sampling point of the node; represents the node index; represents the sampling point serial number; Based on the node state vector, construct the state matrix of all nodes in the ; where represents the state matrix of all nodes; represents the total number of nodes; represents the th sampling point of the th node; represents the th sampling point of the th node; represents the th sampling point of the th node; represents the th sampling point of the th node; represents the current of the

[0021] S2: Slide and update the node state matrix within a continuous time window, generate the state evolution trajectory of each node within this time window, construct a node trajectory set, and calculate the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the distribution network load state; The step of generating the state evolution trajectory of each node within this time window and constructing a node trajectory set includes the following steps: The number of sampling points is , slide and update the state matrix of all nodes. For any node , its state evolution trajectory equation within the time window is: ; represents the trajectory matrix of node ; represents the th sampling point voltage of node ; represents the th sampling point active power of node ; represents the th sampling point reactive power of node ; The th sampling point voltage of node ; Construct the quantity coefficients of current and voltage, as well as the power coefficients of active power and reactive power, reduce the dimension of the state evolution trajectory equation, and construct the node trajectory set: ; ; Among them, represents the steepness coefficient of the quantity coefficient; represents the voltage and current quantity coefficients; represents the power coefficients of active power and reactive power ; In the above formula, the dependent variable reflects the quantity coefficient of the th node at the th sampling point. Its essence is to divide the instantaneous product of and current by a non-linear adjustment factor based on the ratio. Thus, when the voltage-current ratio is abnormal, such as a voltage dip or a current surge, it suppresses the distortion of power calculation and improves the robustness of the result; the independent variables and directly affect the dependent variable: on the one hand, the increase of and will directly increase through the numerator term, showing a positive correlation; on the other hand, the ratio of the two adjusts the overall result through the exponential term in the denominator. When the voltage increases or the current decreases, the exponential term approaches 0 and the denominator approaches 1, and the power calculation remains the original value; conversely, if the current surges, it will cause the ratio to decrease, the exponential term to increase significantly, and the denominator to expand, so that is actively suppressed, avoiding numerical overflow or misjudgment of traditional algorithms in extreme cases; The dependent variable in the above formula represents the power coefficient of the th node at the th sampling point. It is obtained by dividing the active power of this node by the magnitude of its apparent power , so as to map the absolute value of to the interval from 0 to 1, reflecting the relative proportion of active power in the total apparent power. By normalization, the influence of different equipment capacities or load levels on power analysis is eliminated, facilitating horizontal comparison of power utilization efficiency; the independent variables and jointly determine the dependent variable: the increase of will directly increase through the same-direction changes of the numerator and denominator, but because the denominator contains and , is positively correlated with the influence on ; while the increase of will lead to the inflation of the denominator, resulting in the decrease of

[0022] The set of node trajectories after dimensionality reduction is: ; where represents the trajectory matrix of node .

[0023] The steps of calculating the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the load state of the distribution network include the following: For each parameter column in , calculate the mean and standard deviation. The standardized trajectory matrix is: ; where represents the standardized electricity coefficient; represents the standardized power coefficient; Calculate the covariance matrix: ; where represents the transpose matrix of represents the covariance matrix of the th node; Solve the eigenvalues and eigenvectors of the covariance matrix: ; where represents the eigenvalue of the th electrical parameter; represents the eigenvector of the th electrical parameter; Calculate the trajectory entropy: ; where ; where represents the parameter variable index; represents the trajectory entropy of node ; represents the normalized weight of the th eigenvalue; If , it means that the electrical parameters tend to a steady state; If , it indicates that the electrical parameters tend to be abnormal.

[0024] S3: Conduct an aggregation analysis on the state trajectory set, identify the aggregation situations of multiple node trajectories in space and time, and when the preset aggregation degree threshold is met, determine the corresponding area as the state confluence area, where the state confluence area is an abnormal load aggregation area with concentrated access of electric vehicles; The identification of the aggregation situations of multiple node trajectories in space and time includes the following steps: Obtain the trajectory entropy sequences of all nodes within the time window and standardize all sequences: ; Among them, represents the mean value of the trajectory entropy of node within the time window; represents the standard deviation of the trajectory entropy of node within the time window; represents the value after standardization; Use the mutual information entropy method to calculate the dynamic similarity score and sequences of different nodes : ; Among them, if the score tends to 1, it indicates that the evolution patterns of the two nodes are highly synchronized; if the score tends to 0, it indicates that the evolution patterns are irrelevant; represents and 's mutual information; represents information entropy; represents information entropy; represents the value after standardization; The specific mutual information entropy method is as follows: Set , , among which, and are class symbols, and respectively represent the symbol sets of the two sequences, , ; represents the symbol of the th type; represents the symbol of the ; Calculate univariate probabilities and joint probabilities: ; ; ; Denote the probability of the symbol appearing; Denote the probability of the symbol appearing; Denote the joint probability; Calculate the information entropy: ; ; Denote the information entropy; Denote the information entropy; Calculate the mutual information: ; Denote the and mutual information; Based on the dynamic similarity score , construct a matrix: , and define the spatial proximity relationship of nodes based on the power grid topology to construct an adjacency matrix , where denotes that node is directly connected to , otherwise ; Combine the similarity matrix with the adjacency matrix to construct a fused spatio-temporal feature matrix: ; Divide the node groups by the spectral clustering algorithm: , , denoting candidate confluence regions; Denote element-wise multiplication; Denote the spatio-temporal fusion matrix between node and node ; The specific spectral clustering algorithm is as follows: Obtain the fused spatio-temporal feature matrix: , construct a degree matrix, the diagonal elements of which are the node degrees , and perform Laplacian normalization on the degree matrix: , where represents the degree matrix; represents the normalized Laplacian matrix; For solve the eigenvalue equation: , select the first non-zero eigenvalues: , and the corresponding eigenvectors are: , based on the eigenvectors, construct a low-dimensional embedding matrix : ; Take each row of the matrix as the low-dimensional representation of the node. The clustering process is as follows: randomly initialize cluster centers, where is the preset number of confluence regions; iteratively update the center points, assign the nodes to the nearest center until convergence, and output the set of node groups

[0025] For each cluster , calculate the comprehensive score: ; where ; ; ; where represents the comprehensive score of the th cluster; represents the time consistency strength; represents the spatial aggregation degree; represents the weight of the time consistency strength; represents the weight of the spatial aggregation degree; represents the th group represents the total number of nodes included; If , then mark as a state confluence region, where represents the preset comprehensive score threshold.

[0026] S4: Trace back the nodes and their time evolution trajectories in the state confluence region, identify the starting nodes, expansion paths, and influence boundaries of abnormal evolution, and construct a complete abnormal event propagation chain map. The map includes abnormal starting nodes, abnormal influence chains, and event propagation durations; The construction of the complete abnormal event propagation chain map specifically includes the following steps: For each node in the state confluence area , locate the abnormal starting point through the following steps: Compute Node In the Local mutation index within a window: ; in, Representation Node In the The local mutation index within a window; Representation Node In the The mean entropy of the trajectories in the window; Representation Node No. Window to The standard deviation of the trajectory entropy of the sequence composed of all trajectory entropies of windows; Indicates the window number index; if , then mark In the An exception is triggered within a window; Represents the preset local mutation threshold, which is verified by Granger causality test of Whether the timing significantly affects other nodes; The Granger causality test is specifically: Set the target node to , requires verification Whether there is a causal impact on it, the steps are as follows: Construct unconstrained models, including Historical information: ; represents the intercept term; Representation Node The regression coefficient of its own historical value; Representation Node In the Trajectory entropy of a time window; Representation Node Regression coefficients of historical values; Construct a constrained model without Historical information: ; in, represents the maximum lag order; represents the residual term; use test: ; Among them, represents the total number of time windows; represents the sum of squared residuals of the constrained model; represents the sum of squared residuals of the unconstrained model; If the value exceeds the critical value, then it is considered that is the Granger cause of Establish dynamic propagation rules for extended path tracking. The propagation from node to node needs to satisfy: ; represents the number of windows passed from node to node ; represents the power grid fluctuation propagation speed; represents node and node the minimum topological path length between; represents the window time normalization factor; represents the sampling interval; represents the number of sampling points; If node is marked as abnormal within the th window, and the trajectory entropy correlation coefficient between windows satisfies: ; Then mark as a valid propagation path, represents the mean value of the standardized trajectory entropy of node within the th window; represents the standardized trajectory entropy of node within the th time window; represents the Pearson correlation coefficient; Iteratively expand until no new nodes meet the conditions; The calculation of the said correlation coefficient is as follows: Collect the mean values of the standardized trajectory entropy of node and node in multiple consecutive time windows to form two groups of sequences: Sequence : The mean value of the standardized trajectory entropy of node in the time window : ; Sequence D: Node In the window Mean of the normalized trajectory entropy: ; Wherein, Is the number of data points; Global normalization: ; Wherein, Represents the mean of the sequence ; Represents the mean of the sequence ; If the node Meets in the extended path tracing: ; Wherein, Represents the peak of the trajectory entropy of the node In all windows; Represents a preset trajectory entropy boundary threshold; Represents the total number of time windows.

[0027] S5: Cross-validate the abnormal event propagation chain spectrum with the preset impact load characteristic rules. When the rule matching condition is met, confirm that the abnormal type is the impact load abnormality caused by the centralized access of electric vehicles.

[0028] The preset impact load characteristic rules include multi-node power surge, voltage linkage, and double-index jump; Set the propagation chain node set as: , and the propagation chain path set as: ; Represents the total number of nodes in the propagation chain; The multi-node power surge includes counting the number of active power surges of the nodes in the propagation chain: ; Wherein, ; Represents the minimum number of nodes triggering the active power surge rule; Represents the indicative function; Represents the node At the Difference between the active power of the sampling point and the active power of the previous sampling point; Represents the node At the Sampling point active power; Represents the point At the The active power of each sampling point; Indicates the multi-node sudden increase evaluation index value; In this embodiment, 10 time points and the active power of 5 groups of nodes are selected to simulate the multi-node power sudden increase. The experimental measurement data is shown in Table 1: Table 1: Multi-node power sudden increase data detection table

[0029] In Table 1, it can be seen that the active power of multiple nodes is collected and divided into 10 groups of data according to time nodes. Among them, the active power of nodes 1 to 5 shows a sudden increase trend, which conforms to the abnormal distribution network load caused by the connection of electric vehicles, that is, the multi-node power sudden increase; The voltage linkage includes for the propagation path , calculate the correlation coefficient of the node voltage and the phase difference: ; Indicates node and node 's Pearson correlation coefficient; Indicates the voltage fluctuation phase difference; Indicates the path ratio threshold; Indicates node 's voltage; Indicates node 's voltage; Indicates the voltage linkage evaluation index value; ; Among them, ; ; Calculate the covariance: ; Indicates and 's covariance; Indicates the sampling point index variable; Indicates 's mean value; Indicates 's mean value; Calculate the standard deviation: ; ; Indicates 's standard deviation; Indicates Standard deviation; Calculate the mean value: ; ; Phase difference Indicates a node And the node The time lag of voltage fluctuation can be calculated by the cross - correlation function method: ; Wherein, Indicates the time offset; Indicates the sampling interval; In this embodiment, 10 time points are selected, and 5 groups of node voltages are used for simulating voltage linkage. The data measurement is shown in Table 2: Table 2: Voltage linkage data measurement table

[0030] In Table 2, it can be seen that the voltages of multiple nodes are collected. According to the time nodes, they are divided into 10 groups of data. Among them, the voltages of nodes 1 to 5 show the same degree of upward trend, and the voltage values of each node show similar values at the same time point, which is in line with the abnormal change of the distribution network load caused by the connection of electric vehicles, that is, voltage linkage; The double - index jump includes each node in the propagation chain , detecting its active power And reactive power , ; Wherein, Indicates the ratio threshold of active power and reactive power; Indicates the node ratio threshold; Indicates the double - index jump evaluation index value; In this embodiment, 10 time points are selected, and 5 groups of node power ratios are used to simulate the double - index jump. The measurement data is shown in Table 3: Table 3: Double - index jump data measurement table

[0031] In Table 3, it can be seen that the voltages of multiple nodes are collected. According to the time nodes, they are divided into 10 groups of data. Among them, the power ratios of active power and reactive power of nodes 1 to 5 show an upward trend, and it is in line with the abnormal change of the distribution network load caused by the connection of electric vehicles, that is, the double - index jump; Count the number of nodes and the path ratio in the propagation chain that meet the above - mentioned preset impact load characteristic rules, and perform confidence calculation. The calculation formula is as follows: ; Among them, represents the index mean value; if , it is determined that the propagation chain is triggered by the impact load, and a warning is triggered.

[0032] Please refer to Figure 5 , the present invention further provides a warning device for abnormal load changes in a distribution network. The warning device is used to execute the above warning method, including: A data acquisition unit, configured to collect the power consumption parameters of each node in the distribution network in real time. The power consumption parameters include voltage, active power, reactive power, and current, construct a multi-dimensional state vector for each node, and form a state matrix of all network nodes at the current moment with the node topology number as the index, and represent the overall operating state of the distribution system at a specific moment through the state matrix; A trajectory identification unit, configured to perform sliding updates on the node state matrix within a continuous time window, generate the state evolution trajectory of each node within the time window, construct a node trajectory set, and calculate the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the load state of the distribution network; A node aggregation unit, configured to perform aggregation analysis on the node trajectory set, identify the aggregation situations of multiple node trajectories in space and time, and when the preset aggregation degree threshold is met, determine the corresponding area as the state confluence area, and the state confluence area is an abnormal load aggregation area where electric vehicles are concentratedly connected; An abnormal backtracking unit, configured to backtrack each node and its time evolution trajectory in the state confluence area, identify the starting node, expansion path, and influence boundary of abnormal evolution, and construct a complete abnormal event propagation chain map, where the map includes the abnormal starting node, abnormal influence chain, and event propagation duration; A mobile analysis unit, which cross-verifies the abnormal event propagation chain map with the preset impact load characteristic rules, and when the rule matching condition is met, confirms that the abnormal type is an impact load abnormality caused by the concentrated access of electric vehicles.

[0033] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0034] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0035] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They may be located in one place or distributed over multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0036] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application.

Claims

1. A method for early warning of abnormal changes in distribution network load, characterized in that, Including: S1: Real-time collect the power consumption parameters of each node in the distribution network. The power consumption parameters include voltage, active power, reactive power, and current. Construct a multi-dimensional state vector for each node, and use the node topology number as an index to form the network-wide node state matrix at the current moment. The overall operating state of the distribution system at a specific moment is characterized by the state matrix; S2: Slide and update the node state matrix within a continuous time window, generate the state evolution trajectory of each node within this time window, construct a node trajectory set, and calculate the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the distribution network load state; S3: Conduct an aggregation analysis on the node trajectory set to identify the aggregation situations of multiple node trajectories in space and time. When the preset aggregation degree threshold is met, determine the corresponding area as the state confluence area. The state confluence area is an abnormal load aggregation area where electric vehicles are concentratedly connected; S4: Trace back the nodes and their time evolution trajectories in the state confluence area, identify the starting node, expansion path, and influence boundary of abnormal evolution, and construct a complete abnormal event propagation chain graph. The graph includes the abnormal starting node, abnormal influence chain, and event propagation duration; S5: Cross-validate the abnormal event propagation chain graph with the preset impact load characteristic rules. When the rule matching condition is met, confirm that the abnormal type is the impact load abnormality caused by the concentrated access of electric vehicles.

2. The method for early warning of abnormal distribution network load according to claim 1, characterized in that, The specific process of forming the network-wide node state matrix at the current moment is as follows: Sampling is performed on the electrical parameters at the sampling interval of , and the number of sampling points is set to . The electrical parameters of each node are continuously collected, where the topological numbers of each node are encoded incrementally from the power supply end to the load end; at each sampling point within the time window , the voltage , active power , reactive power , and current of the node are longitudinally stacked in the order of node numbers to form a network-wide node status matrix , whose dimension is . Its rows correspond to nodes, and the columns represent voltage, active power, reactive power, and current parameters in sequence.

3. The method for early warning of abnormal load change in a distribution network according to claim 2, characterized in that, The generation of the state evolution trajectory of each node within this time window to construct a node trajectory set includes the following steps: Slide and update the network-wide node state matrix, construct the power quantity coefficients of current and voltage, and the power coefficients of active power and reactive power. Reduce the dimension of the state evolution trajectory equation to construct a node trajectory set: ； ； Among them, represents the steepness coefficient of the power coefficient; represents voltage and current power coefficient; represents the power coefficient of active power and reactive power ; The node trajectory set after dimension reduction is: ； Among them, represents the trajectory matrix of the node.

4. The method for early warning of abnormal load variation in a distribution network according to claim 1, wherein, The calculation of the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the distribution network load state includes the following steps: For calculate the mean and standard deviation for each parameter column, and the standardized trajectory matrix is: ； Among them, represents the normalized electricity quantity coefficient; represents the normalized power coefficient; Calculate the covariance matrix: ； Among them, represents the transpose matrix of; represents the covariance matrix of the Solve the eigenvalues and eigenvectors of the covariance matrix: ； Among them, represents the eigenvalue of the th electrical parameter; represents the eigenvector of the th electrical parameter; Calculate the trajectory entropy: ； Wherein, ； Among them, represents the parameter variable index; represents the node trajectory entropy; represents the normalized weight of the eigenvalue; If , it means that the electrical parameters tend to be stable; If , it indicates that the electrical parameters tend to be abnormal.

5. The method for early warning of abnormal distribution network load according to claim 1, characterized in that The identification of the aggregation situations of multiple node trajectories in space and time includes the following steps: Obtain the trajectory entropy sequences of all nodes within the time window and standardize all sequences: ； Among them, represents the mean value of the trajectory entropy of the node within the time window; represents the node standard deviation of the trajectory entropy within the time window; represents the value after normalization; Calculate the dynamic similarity scores of different nodes using the mutual information entropy method and sequences : ； Among them, if the score tends to 1, it means that the evolution patterns of the two nodes are highly synchronized; if the score tends to 0, it means that the evolution patterns are irrelevant; represents and the mutual information of; represents the information entropy; represents the information entropy; represents the value after standardization; Based on the dynamic similarity score , construct a matrix: , and define the node spatial proximity relationship based on the power grid topology to construct an adjacency matrix , where represents node is directly connected to , otherwise ; Combine the similarity matrix with the adjacency matrix , and construct a fused spatio-temporal feature matrix: ； Dividing node groups through spectral clustering algorithm: , , representing candidate confluence regions; represents element-wise multiplication; represents node and node spatio-temporal fusion matrix; represents the serial number of the candidate confluence region; For each cluster , calculate the comprehensive score: ； Wherein, ； ； ； Among them, represents the comprehensive score of the th cluster; represents the time consistency strength; represents the spatial aggregation degree; represents the weight of the time consistency strength; represents the weight of the spatial aggregation degree; represents the total number of nodes included in the group ; represents the spatial distance between nodes; If , then mark as the state confluence area, where represents the preset comprehensive score threshold.

6. The method for early warning of abnormal load change in a distribution network according to claim 1, characterized in that, The construction of a complete abnormal event propagation chain graph specifically includes the following steps: For each node in the status confluence area , locate the abnormal starting point through the following steps: Computing node In the local mutation index within the window: ； Among them, represents the local mutation index of the node in the th window; represents the average trajectory entropy of the node in the th window; represents the standard deviation of the trajectory entropy of the sequence composed of all trajectory entropies from the th window to the th window of the node; represents the window number index; If , then mark to trigger an exception within the th window; represents a preset local mutation threshold, and uses Granger causality test to verify of whether the time series significantly affects other nodes; Establish dynamic propagation rules for extended path tracing. The propagation from node to node needs to satisfy: ； Indicates the slave node to the node The number of windows passed; Indicates the propagation speed of grid fluctuations; Indicates the node and the node The minimum topological path length; Indicates the window time normalization factor; Indicates the sampling interval; Indicates the number of sampling points; If a node is marked as abnormal within the window, and the trajectory entropy correlation coefficient between windows satisfies: ； Then the label is an effective propagation path, indicating the node in the mean of the normalized trajectory entropy within the window; indicating the node in the normalized trajectory entropy within the time window; indicating the Pearson correlation coefficient; iteratively expand until no new nodes meet the criteria; If a node satisfies in the extended path tracing: ； Among them, represents the peak of the trajectory entropy of the node in all windows; represents a preset trajectory entropy boundary threshold; represents the total number of time windows.

7. The method for early warning of abnormal load in a distribution network according to claim 1, characterized in that The cross-validation of the abnormal event propagation chain graph with the preset impact load characteristic rules includes the following steps: Check whether the preset impact load characteristic rules are triggered from the abnormal propagation chain graph. The preset impact load characteristic rules include multi-node power sudden increase, voltage linkage, and double-index jump; Set the set of propagation chain nodes as: , and the set of propagation chain paths as: ; is the total number of nodes in the propagation chain; The multi-node power sudden increase includes counting the number of times of active power sudden increase of the nodes in the propagation chain: ； Wherein, ； Indicates the minimum number of nodes that trigger the active power sudden increase rule; Indicates the indicative function; Indicates the node At the Difference in active power between the sampling point and the previous sampling point; Indicates the node At the Sampling point of active power; Indicates the point At the Sampling point of active power; Indicates the multi-node sudden increase evaluation index value; The voltage linkage includes, for the propagation path , calculating the correlation coefficient of the node voltage and the phase difference: ； Represents a node With the node Pearson correlation coefficient; Represents the voltage fluctuation phase difference; Represents the path ratio threshold; Represents the voltage linkage evaluation index value; The double-index jump includes each node in the propagation chain , detecting its active power and reactive power , ； in, Indicates the ratio threshold of active power to reactive power; Indicates the node ratio threshold; Indicates the value of the double index jump evaluation index; Count the number of nodes and the path ratio in the propagation chain that meet the above preset impact load characteristic rules, and conduct confidence calculation. The calculation formula is as follows: ； Among them, represents the index mean; if , it is determined that the propagation chain is triggered by an impact load, and a warning is triggered.

8. A distribution network load abnormal movement warning device, characterized in that: The warning device is used to execute the warning method according to any one of claims 1-7, including: A data acquisition unit, which is used to collect the power consumption parameters of each node in the distribution network in real time. The power consumption parameters include voltage, active power, reactive power, and current. A multi-dimensional state vector of each node is constructed, and with the node topology number as the index, a node state matrix of the entire network at the current moment is formed, and the overall operating state of the distribution system at a specific moment is characterized by the state matrix; A trajectory identification unit, which is used to slide and update the node state matrix within a continuous time window, generate the state evolution trajectory of each node within this time window, construct a node trajectory set, and calculate the trajectory entropy of different nodes to capture the dynamic evolution characteristics of the load state of the distribution network; A node aggregation unit, which is used to perform an aggregation analysis on the node trajectory set, identify the aggregation situations of multiple node trajectories in space and time, and when the preset aggregation degree threshold is met, determine the corresponding area as the state confluence area. The state confluence area is an abnormal load aggregation area where electric vehicles are concentratedly connected; An abnormal backtracking unit, which is used to backtrack each node and its time evolution trajectory in the state confluence area, identify the starting node, expansion path, and influence boundary of the abnormal evolution, and construct a complete abnormal event propagation chain map. The map includes the abnormal starting node, abnormal influence chain, and event propagation duration; A mobile analysis unit, which is used to cross-validate the abnormal event propagation chain map with the preset impact load characteristic rules. When the rule matching condition is met, confirm that the abnormal type is the impact load abnormality caused by the concentrated connection of electric vehicles.

Images