Forced oscillation detection and positioning method and system for power distribution network

Abstract

The invention relates to the technical field of power distribution networks, in particular to a forced oscillation detection and positioning method and system for a power distribution network. Comprising the following steps: performing cut set division on a power grid by adopting a spectrogram theory, and repeatedly performing cut set division based on a sub-grid energy transfer value; and analyzing a key line of the cut set line based on the betweenness centrality, the power fluctuation value and the frequency and phase angle data of the cut set. According to the method, binary segmentation is carried out through a Fiedler vector, a natural cut set structure in the power grid can be effectively identified, the power grid is divided into two sub-grids, nodes in each sub-grid have higher connectivity, key line candidates are determined through three dimensions of betweenness centrality, power fluctuation, frequency deviation and phase angle deviation respectively, and the power grid is divided into two sub-grids. The method is advantaged in that identification accuracy can be substantially improved, through combination of IMF decomposed by EMD and Hilbert transform, instantaneous frequency and amplitude can be extracted, and oscillation intensity and frequency on each key line can be monitored in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network, and in particular to a forced oscillation detection and positioning method and system for power distribution network. Background Art

[0002] With the continuous expansion of the scale of power grids and the increasing complexity of the operating environment, in large-scale transmission and distribution systems, forced oscillations often lead to frequency fluctuations, power imbalances, and even system failures in power grid operation. Traditional oscillation detection methods are mostly based on time-frequency analysis and linear modeling to detect forced oscillations. With the increase of grid load fluctuations and distributed energy access, the oscillation mode often shows high nonlinearity and non-stationarity.

[0003] However, most of the existing oscillation mode detection methods rely on static data analysis and lack the ability to dynamically process complex time series data. They are unable to accurately identify the patterns of forced oscillations and locate nodes. At the same time, traditional oscillation energy calculations fail to fully utilize the topological structure and energy transfer characteristics of the power grid for multi-level analysis, which can easily lead to inaccurate identification of key lines and make it difficult to locate oscillation sources in a complex power grid in a timely and effective manner. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to solve the problem that most of the existing oscillation mode detection methods rely on static data analysis, lack the ability to dynamically process complex time series data, and cannot accurately identify the mode of forced oscillation and locate the nodes. At the same time, traditional oscillation energy calculation fails to fully utilize the topological structure and energy transfer characteristics of the power grid for multi-level analysis, which easily leads to inaccurate identification of key lines and difficulty in timely and effective location of oscillation sources in complex power grids.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a forced oscillation detection and positioning method for a distribution network, which includes obtaining real-time operation data of the distribution network and using spectral graph theory to divide the power grid into cut sets;

[0008] Analyze and screen key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set;

[0009] Perform empirical mode decomposition and Hilbert transform on the selected key lines to construct time-frequency domain distribution features;

[0010] The LSTM deep learning model is used to identify oscillation patterns, calculate line oscillation energy, and locate the maximum disturbance source.

[0011] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, the method of using spectral graph theory to divide the power grid into cut sets includes:

[0012] Construct a graph theory model of the distribution network, with power stations, substations and load centers as nodes and lines and transformers as edges;

[0013] Construct an adjacency matrix based on node connection relationships, and calculate the degree matrix and Laplace matrix;

[0014] The eigenvector corresponding to the second smallest eigenvalue is selected as the Fiedler vector, and the nodes are assigned to two subnetworks according to whether they are greater than or equal to 0 or less than 0.

[0015] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, wherein: the subnet energy transfer threshold is calculated based on historical data;

[0016] When the subnet energy transfer is greater than or equal to the threshold, the cut set segmentation continues;

[0017] When the subnet energy transfer is less than the threshold, the segmentation is stopped.

[0018] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, the analysis and screening of key lines include: calculating the line betweenness centrality, selecting the ranking 20% ​​as the first category candidate lines; calculating the power fluctuation value, selecting the ranking 25% as the second category candidate lines; calculating the frequency and phase angle deviation, selecting the ranking 20% ​​as the third category candidate lines; determining the intersection of any two categories of candidate lines as the final key line.

[0019] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, the empirical mode decomposition includes: extracting the intrinsic mode function of power fluctuation and frequency deviation signals; calculating the mean of the upper and lower envelopes as the local trend; iteratively removing the local trend until the IMF definition is met.

[0020] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, the LSTM deep learning model includes: an input layer receives time-frequency domain feature data;

[0021] A two-layer LSTM network processes time series features; the fully connected layer outputs oscillation frequency and amplitude prediction values; and the mean square error is used as the loss function to optimize the model parameters.

[0022] As a preferred solution of the forced oscillation detection and positioning method for distribution network described in the present invention, the positioning of the maximum disturbance source includes: calculating the oscillation energy of the key line; determining the maximum disturbance line by comparing the oscillation energy; calculating the change rate of node frequency deviation; and determining the node with the largest frequency change rate as the oscillation source point.

[0023] In a second aspect, an embodiment of the present invention provides a forced oscillation detection and positioning system for a distribution network, which includes a data acquisition module, which acquires real-time operation data of the distribution network and uses spectral graph theory to divide the power grid into cut sets;

[0024] The key line identification module analyzes and screens key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set;

[0025] The oscillation mode detection module performs empirical mode decomposition and Hilbert transform on the selected key lines to construct the time-frequency domain distribution characteristics;

[0026] The oscillation source location module uses the LSTM deep learning model to identify the oscillation pattern, calculate the line oscillation energy, and locate the maximum disturbance source.

[0027] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of a forced oscillation detection and positioning method for a distribution network as described in the first aspect of the present invention are implemented.

[0028] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of a forced oscillation detection and positioning method for a distribution network as described in the first aspect of the present invention are implemented.

[0029] The beneficial effects of the present invention are as follows: by performing binary segmentation using the Fiedler vector, the natural cut set structure in the power grid can be effectively identified, the power grid can be divided into two subnets, and the nodes in each subnet have stronger connectivity. The key line candidates are respectively determined by the three dimensions of betweenness centrality, power fluctuation, frequency deviation and phase angle deviation, which can significantly improve the accuracy of identification. By combining the IMF decomposed by the EMD with the Hilbert transform, the instantaneous frequency and amplitude can be extracted, and the intensity and frequency of oscillations on each key line can be monitored in real time. By constructing and using the LSTM model, dynamic analysis can be performed based on time-frequency graph information to identify the frequency change trend over time and the change in oscillation intensity. By calculating the frequency deviation change rate of the nodes on each line, the node where the oscillation frequency changes earliest can be accurately identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 A flow chart of a forced oscillation detection and positioning method and system for a distribution network;

[0032] Figure 2 A computer device diagram of a forced oscillation detection and location method for distribution networks. DETAILED DESCRIPTION

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0034] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0035] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0036] Example 1

[0037] Reference Figure 1-2 , which is the first embodiment of the present invention, and provides a forced oscillation detection and positioning method for a distribution network, comprising:

[0038] S100: Acquire real-time operation data of the distribution network and use spectral graph theory to divide the power grid into cut sets;

[0039] S101: Use spectral graph theory to divide the power grid into cut sets, including:

[0040] Construct a graph theory model of the distribution network, with power stations, substations and load centers as nodes and lines and transformers as edges;

[0041] Construct an adjacency matrix based on node connection relationships, and calculate the degree matrix and Laplace matrix;

[0042] The eigenvector corresponding to the second smallest eigenvalue is selected as the Fiedler vector, and the nodes are assigned to two subnetworks according to whether they are greater than or equal to 0 or less than 0.

[0043] Preferably, the spectral graph theory is used to divide the power grid into cut sets, including:

[0044] Based on the physical topology of the power grid, the power stations, substations and load centers in the power grid are taken as nodes, and the lines and transformers are taken as edges to construct a graph model of the distribution network. The weight of each edge can be defined according to the physical parameters of the line (such as impedance or conductance), which can be expressed as:

[0045] G = Graph(V, E, ω);

[0046] or ij =G ij ;

[0047] Where G represents the graph structure of the power grid, including the node set V, the edge set E and the weight set ω, ω ij represents the weight between node i and node j, Z ij represents the impedance between node i and node j, G ij represents the conductance value between node i and node j;

[0048] The spectral graph theory is used to divide the power grid into cut sets, where the adjacency matrix A is constructed according to whether nodes i and j are directly connected through lines, which is expressed as:

[0049]

[0050] Among them A ij Represents the elements of node i and node j in the adjacency matrix, representing the connection weight between node i and node j;

[0051] Based on the sum of each row in the adjacency matrix A, the weights of the edges directly connected to the node are calculated and used as the diagonal elements of the corresponding nodes, and the degree matrix D is constructed, which is expressed as:

[0052]

[0053] Where D ii represents the sum of the weights of the edges directly connected to node i;

[0054] The Laplacian matrix L is calculated by subtracting the adjacency matrix A from the degree matrix D, expressed as:

[0055] L = DA;

[0056] Use the QR decomposition algorithm to perform eigenvalue decomposition on the Laplace matrix L, and iterate and update it step by step until it converges to a diagonal matrix, and use the elements of the diagonal matrix as eigenvalues ​​and columns as eigenvectors;

[0057] The eigenvector corresponding to the second smallest eigenvalue is selected as the Fiedler vector, where each component of the Fiedler vector corresponds to a node in the distribution network and is split into two. If the component is greater than or equal to 0, the node is assigned to subnet 1; if the component is less than 0, the node is assigned to subnet 2.

[0058] Based on the allocation of nodes to subnet 1 and subnet 2, two cut sets of nodes are determined, which represent the set of edges that are cut between the two subnets when the power grid is divided into two parts. The total weight A' of the cut edges is calculated and expressed as:

[0059]

[0060] The sum of the average value and double standard deviation of the total weight calculated based on history is used as the upper limit of the cut set weight. If the calculated total weight A' is greater than or equal to the cut set weight, the eigenvalue decomposition is performed again.

[0061] By taking power stations, substations and load centers as nodes and using lines and transformers as edges, the physical topology of the power grid can be accurately reflected. The physical parameters of the power grid lines (such as impedance or conductance) are used to define the edge weights, ensuring that the edge weights in the model contain the physical properties of the power grid lines, improving the physical accuracy of power grid modeling, and using the Fiedler vector for binary segmentation to effectively identify the natural cut set structure in the power grid, dividing the power grid into two subnets, and the nodes in each subnet have stronger connectivity, while the connection between subnets is weaker, improving the rationality of the cut set division, and through QR decomposition and iterative update until convergence to a diagonal matrix, the accuracy of the final eigenvalues ​​and eigenvectors can be ensured. The second smallest eigenvalue and the corresponding eigenvector (Fiedler vector) reflect the path that is most easily cut off in the power grid. Through this eigenvalue, the natural cut set of the power grid can be obtained. The Fiedler vector provides a clear division basis for each node, helping to divide the power grid into subnets with strong internal connections and determine the lines that are easily cut off.

[0062] S102: Calculating a subnet energy transfer threshold based on historical data;

[0063] When the subnet energy transfer is greater than or equal to the threshold, the cut set segmentation continues;

[0064] When the subnet energy transfer is less than the threshold, the segmentation is stopped.

[0065] Furthermore, the cut set segmentation is repeated based on the subnet energy transfer value, including:

[0066] Based on the energy transfer value of the stable operation cut set in the historical data as a reference, the average value of the energy transfer value is calculated as the energy threshold of the cut set. Based on the historical stable operation cut set data, the cut set segmentation is repeated, and the power flow of each cut set is calculated, which is expressed as:

[0067]

[0068] Where E z (t) represents the total energy transfer of the cut set z at time t, P ij (t) represents the transmission power of cut set node i and node j at time t;

[0069] If the energy transfer of the cut set is greater than or equal to the energy threshold, the cut set needs to be further segmented. If the energy transfer of the network is less than the energy threshold, the segmentation of the cut set is stopped.

[0070] By using the energy transfer value of the cut set during historical stable operation to set the energy threshold, the rationality and adaptability of the threshold can be ensured, and the set energy threshold can be ensured to be highly consistent with the actual system operation status, thereby improving the reliability of cut set segmentation. Through multiple segmentation and energy transfer calculations of the cut set, when the energy transfer value of a cut set is too large, it will be further divided into sub-cut sets, thereby reducing the power flow between subnets, optimizing the power load distribution, and avoiding power overload in local areas. At the same time, by continuously calculating the power flow between cut sets, the load status of each area in the power grid can be monitored in real time, and the step of stopping the segmentation based on the energy threshold prevents excessive segmentation from increasing the calculation burden and destroying the integrity and coordination of the power grid, thereby reducing unnecessary segmentation processes and ensuring the improvement of calculation efficiency.

[0071] S200: Analyze and screen key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set;

[0072] S201: Analysis and screening of key lines include: calculating line betweenness centrality, selecting lines ranked 20% as first-category candidate lines; calculating power fluctuation values, selecting lines ranked 25% as second-category candidate lines; calculating frequency and phase angle deviations, selecting lines ranked 20% as third-category candidate lines; and determining the intersection of any two categories of candidate lines as the final key line.

[0073] Preferably, the key circuits of the cut set circuits are analyzed based on the betweenness centrality, power fluctuation value, frequency and phase angle data of the cut set, including:

[0074] Based on the segmented set, the betweenness centrality is calculated for each line between two nodes, expressed as:

[0075]

[0076] Among them C B (e ij ) indicates line e ij The betweenness centrality of It means to sum all the different node pairs (s, x) in the node set V in the cut set, and s≠x, σ(s, x|e ij ) indicates line e ij The number of shortest paths between the node pair s and x, σ(s,x) represents the number of shortest paths between node s and node x;

[0077] Based on historical operation data, 20% of the distribution betweenness centrality ranking is selected as the betweenness centrality threshold, and when the betweenness centrality of a line is greater than or equal to the betweenness centrality threshold, it is judged as the first key line candidate;

[0078] The sliding window analysis is performed based on the active power data collected in real time by PMU, which is expressed as:

[0079]

[0080] Where ΔP ij (t) represents the power fluctuation value of node i and node j at time t, T represents the window time, τ represents the integral variable, represents the average power of lines i and j in the time window T before time t;

[0081] Based on historical operation data, 25% of the power fluctuation value ranking is selected as the fluctuation value threshold, and when the power fluctuation value of the line is greater than or equal to the fluctuation value threshold, it is determined to be a second key line candidate;

[0082] PMU phase measurement units are deployed in the power grid to collect real-time active power data of each line. ij (t);

[0083] Use the PMU device to collect the frequency and phase angle data on the line, and calculate the frequency deviation and phase angle deviation respectively, which are expressed as:

[0084] Δθ i (t) = θ i (t)-θ ref ;

[0085] Δω i (t) = ω i (t)-ω ref ;

[0086] where Δω i (t) represents the frequency deviation between node i and node j, ω i (t) represents the frequency of node i at time t, ωref represents the steady-state reference frequency (determined based on historical data), Δθ i (t) represents the phase angle deviation between node i and node j, θ i (t) represents the phase angle of node i at time t, θ ref Represents the steady-state reference phase angle (determined based on historical data);

[0087] Based on historical operation data, 20% of the frequency deviation and phase angle deviation rankings are selected as the first deviation threshold and the second deviation threshold, respectively, and when the frequency deviation and phase angle deviation of the line are both greater than or equal to the first deviation threshold and the second deviation threshold, it is determined to be a third key line candidate;

[0088] If the route belongs to the intersection of any two key route candidates among the three key route candidates, it will be regarded as a key route;

[0089] If fast analysis and calculation are required, the intersection of the three key line candidates can be selected as the key line, which can further reduce the amount of calculation. Only the key lines are passed, the number of key lines is further reduced, and subsequent frequency decomposition and other steps are performed to improve the efficiency of identifying and detecting oscillation modes.

[0090] By calculating the betweenness centrality, it is possible to identify important lines in the power grid that may cause chain reactions or fault propagation. By analyzing the power fluctuation value, the changes in the actual transmission power on the line can be reflected. Especially when the load changes or the power grid fails, the power fluctuation tends to rise sharply. Through sliding window analysis, short-term random fluctuations can be smoothed and the trend of power changes in a long period of time can be captured. The frequency deviation reflects the difference between the node and the system reference frequency, and the phase angle deviation reflects the phase difference between adjacent nodes in the power grid. The key line candidates are determined by the three dimensions of betweenness centrality, power fluctuation, frequency deviation and phase angle deviation, and then the accuracy of identification can be significantly improved through intersection screening. Only lines that meet multiple conditions are identified as key lines, avoiding the misjudgment caused by a single indicator. Through the identification of key lines after multiple screening, key monitoring and protection measures can be implemented on these lines to prevent the spread of faults or oscillations in advance. The centralized monitoring and optimization of key lines can significantly improve the overall safety and stability of the power grid.

[0091] S300: Perform empirical mode decomposition and Hilbert transform on the selected key lines to construct time-frequency domain distribution features;

[0092] S301: Empirical mode decomposition includes: extracting intrinsic mode functions of power fluctuation and frequency deviation signals;

[0093] Calculate the mean of the upper and lower envelopes as the local trend; iteratively remove the local trend until the IMF definition is met.

[0094] Preferably, the key circuit is frequency decomposed using the empirical mode decomposition method, and the signal is analyzed by Hilbert transform to construct a time-frequency domain distribution diagram, including:

[0095] For each selected key line, input the collected power fluctuation value ΔP ij (t) and frequency deviation Δω i (t) signal, and use empirical mode decomposition (EMD) to analyze the power fluctuation value ΔP ij (t) and frequency deviation Δω i (t) frequency decomposition of the signal;

[0096] Find the signal peak points and valley points of the power fluctuation signal and the frequency deviation signal, use all local maximum values ​​as data points, connect the local maximum points by interpolation, and obtain the upper envelope of the original signal. Use all local minimum points to construct the lower envelope of the original signal by interpolation, and calculate the average values ​​of the upper envelope and the lower envelope as the local trend of the power fluctuation.

[0097] Subtract the local trend from the original signal to get the residual signal. If the residual signal value meets the definition of IMF (intrinsic mode function) (in the entire signal, the difference between the number of zero-crossing points and the number of extreme points is at most 1 and at any time, the local mean of the signal should be close to zero), it is extracted as IMF. If the residual signal does not meet the definition of IMF, it is iterated until the IMF that meets the conditions is obtained, which is expressed as:

[0098]

[0099] Where x(t) represents the input power fluctuation value ΔP ij (t) and frequency deviation Δω i (t) Signal, IMF k (t) represents the kth eigenmode function, r n (t) represents the final residual signal, n represents the total number of intrinsic mode functions;

[0100] Perform Hilbert transform on each IMF to obtain the analytical signal, which is expressed as:

[0101] z k (t) = IMF k (t)+jH(IMF k (t));

[0102] Where j represents the imaginary unit, z k (t) represents the complex form analytical signal of the kth intrinsic mode function, which contains the instantaneous frequency and amplitude information of the IMF.k (t)) represents the Hilbert transform of the kth intrinsic mode function;

[0103] Based on the analytical signal z k (t) Extract the instantaneous frequency and amplitude, expressed as:

[0104]

[0105] Among them A k (t) represents the instantaneous amplitude of the kth eigenmode function at time t, θ k (t) represents the phase angle of the kth eigenmode function at time t, f k (t) represents the instantaneous frequency of the kth eigenmode function at time t, θ k (t+1) represents the phase angle of k eigenmode functions at time t+1, and Δt represents the time interval;

[0106] The instantaneous frequency f is calculated k (t) and instantaneous amplitude A k (t) Construct the IMF time-frequency domain distribution diagram, where the horizontal axis is time and the vertical axis is instantaneous frequency.

[0107] By decomposing the power fluctuation value and frequency deviation signal through EMD, the complex signal can be decomposed into several IMFs (intrinsic mode functions). Each IMF represents an independent oscillation mode, providing a fine-grained signal analysis method. Through EMD, different frequency components in the signal can be separated to help identify different types of oscillation phenomena in key lines. By iteratively calculating and extracting IMFs, different frequency components in the signal can be decomposed layer by layer, which is helpful for analyzing the source and propagation path of various oscillation phenomena in the power grid. By performing Hilbert transform on IMF, the signal can be converted into an analytical signal in complex form, from which the instantaneous frequency and amplitude can be extracted. The instantaneous frequency can reflect the specific changes of the oscillation mode over time, while the instantaneous amplitude reflects the intensity of the oscillation. The extraction of the instantaneous frequency can accurately capture the dynamic changes of the oscillation frequency, helping to identify how the oscillation occurs on different lines. The occurrence, evolution and propagation of oscillations on the road, combined with the instantaneous amplitude analysis of the oscillation intensity, can further identify high-risk areas of oscillation, help power grid operators quickly locate the source of faults or oscillations, and by constructing the time-frequency domain distribution diagram of IMF, the changes in the frequency and amplitude of the oscillation mode over time can be intuitively presented. Compared with traditional signal processing methods such as Fourier transform and wavelet transform, the IMF decomposed by EMD and the Hilbert transform are combined to extract the instantaneous frequency and amplitude, thereby providing the oscillation changes of the signal over time, which is particularly suitable for analyzing the rapidly changing oscillation modes in the power grid, such as short-term frequency fluctuations or transient oscillation phenomena. By extracting the instantaneous frequency and amplitude, the power grid operators can monitor the intensity and frequency of oscillations on each key line in real time. Compared with the static analysis method, this dynamic analysis method combining time and frequency domains can identify the occurrence, evolution and disappearance of oscillation phenomena.

[0108] S400: Uses LSTM deep learning model to identify oscillation patterns, calculate line oscillation energy, and locate the source of maximum disturbance.

[0109] S401: The LSTM deep learning model includes: an input layer receiving time-frequency domain feature data;

[0110] A two-layer LSTM network processes time series features; the fully connected layer outputs oscillation frequency and amplitude prediction values; and the mean square error is used as the loss function to optimize the model parameters.

[0111] Furthermore, the LSTM deep learning model is used to identify and detect oscillation patterns, including,

[0112] Build an LSTM model, including input layer, LSTM layer, fully connected layer and output layer;

[0113] The input layer inputs the time-frequency graph data including instantaneous frequency and instantaneous amplitude data, the LSTM layer processes the time series data, the fully connected layer converts the output of the LSTM into the predicted values ​​of the oscillation frequency and amplitude, and the output layer outputs the model prediction results;

[0114] Use the training set to determine the oscillation pattern and calculate the model loss using mean squared error (MSE) as the loss function;

[0115] The parameters of the LSTM model are updated through the adam optimizer and gradient descent optimization. If the model loss no longer decreases significantly during the continuous iteration, the iteration is stopped and the model parameters are output to update the model.

[0116] After training is completed, the LSTM model is used to predict future time-frequency graph data and output the oscillation pattern at future moments.

[0117] By building and using the LSTM model, it is possible to perform dynamic analysis based on time-frequency graph information, identify the frequency change trend over time and the change in oscillation intensity, effectively process time series data, capture the dynamic changes of frequency and amplitude, identify oscillation patterns, and convert high-dimensional time series features into specific oscillation prediction results, helping operators to intuitively grasp the future oscillation patterns of the power grid. Real-time prediction of oscillation patterns can identify potential risks in the power grid in advance, provide protection for the stability and safety of power grid operation, and realize the detection of forced oscillations.

[0118] S402: locating the maximum disturbance source includes: calculating the oscillation energy of the key circuit; comparing the oscillation energy to determine the maximum disturbance circuit; calculating the node frequency deviation change rate; and determining the node with the largest frequency change rate as the oscillation source point.

[0119] Preferably, the oscillation energy of the key line is calculated, the key line with the maximum disturbance is determined, and the node where the oscillation frequency changes first occurs is located by the frequency deviation change rate of the line node, including:

[0120] Based on each line of the critical line, the power fluctuation value ΔP ij (t) and frequency deviation Δω i (t) The product of the signals calculates the oscillation energy of the circuit;

[0121] Compare the energies of all critical lines and identify the line with the highest oscillation energy as the critical line with the maximum disturbance;

[0122] Calculate and compare the frequency deviation change rate of each node on the line, and determine the location with the largest frequency change rate as the node where the oscillation frequency changes first, expressed as:

[0123]

[0124] Where Δt represents the time detection, Δω i (t+1) represents the frequency deviation of node i at time t+1, Denotes Δω i (t) The rate of change with time t.

[0125] By organically combining the power fluctuation and frequency deviation on the line, the oscillation energy carried by each line in the power grid is reflected. The larger the oscillation energy, the greater the impact of the line on the stability of the power grid. Therefore, this method can more accurately measure the oscillation intensity of each line. By comparing the oscillation energy of all key lines, the line with the highest oscillation energy, that is, the key line with the maximum disturbance, can be quickly identified. When the line with the highest oscillation energy is determined as the source of disturbance, the operator can make more targeted decisions and operations, such as strengthening monitoring or adjusting the line. By calculating the frequency deviation change rate of the nodes on each line, the node where the oscillation frequency changes first can be accurately identified. The operator can dynamically monitor the propagation path of the oscillation phenomenon in the power grid, and give early warning and control the impact of the oscillation source. Using power fluctuation or frequency deviation alone may not fully reflect the full picture of the oscillation phenomenon. By combining the analysis of the two indicators, it can ensure that both the oscillation energy and the frequency deviation are fully considered, thereby providing more reliable analysis results.

[0126] Furthermore, this embodiment also provides a forced oscillation detection and positioning system for a power distribution network, comprising:

[0127] The data acquisition module obtains the real-time operation data of the distribution network and uses the spectral graph theory to divide the power grid into cut sets;

[0128] The key line identification module analyzes and screens key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set;

[0129] The oscillation mode detection module performs empirical mode decomposition and Hilbert transform on the selected key lines to construct the time-frequency domain distribution characteristics;

[0130] The oscillation source location module uses the LSTM deep learning model to identify the oscillation pattern, calculate the line oscillation energy, and locate the maximum disturbance source.

[0131] Example 2

[0132] Reference Figure 1 - Figure 2 , which is the second embodiment of the present invention, and this embodiment provides a forced oscillation detection and positioning method for a distribution network. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0133] This test selected the eastern area of ​​a provincial power grid as the verification object. The area includes 2 500kV substations, 8 220kV substations, 26 110kV substations, and multiple distributed new energy access points, a total of 68 key nodes, and 126 transmission lines. The test period is from July 1, 2023 to September 30, 2023, covering the peak period of summer electricity consumption.

[0134] Complete the following preparations before the test:

[0135] Deploy 86 high-precision PMU devices with a sampling frequency of 200Hz;

[0136] Build a real-time data transmission system based on optical fiber backbone network with a delay of <5ms;

[0137] Deploy an LSTM deep learning model, which includes an input layer, two LSTM layers (128 nodes / layer), and a fully connected layer;

[0138] Establish an oscillation pattern feature library based on historical data, including more than 1,000 groups of typical oscillation cases.

[0139] In order to compare and verify the effect of the present invention, the test area is divided into two areas, A and B:

[0140] Area A: Apply the oscillation detection and positioning method of the present invention;

[0141] Area B: Use the traditional detection method based on spectrum analysis.

[0142] Key monitoring indicators include:

[0143] Cut set partitioning efficiency and accuracy; key line identification rate;

[0144] Oscillation source positioning accuracy; system response time.

[0145] When dividing the power grid into cut sets based on spectral graph theory, the following parameters are set:

[0146] Energy transfer threshold: 5MW;

[0147] The upper limit of binary split iterations: 8 times;

[0148] Fiedler vector calculation accuracy: 1e-6.

[0149] The following criteria are used in the key line screening process:

[0150] Betweenness centrality threshold: top 20%;

[0151] Power fluctuation threshold: top 25%;

[0152] Frequency deviation threshold: ±0.1Hz;

[0153] Phase angle deviation threshold: ±5 degrees.

[0154] Table 1 System performance index comparison

[0155] Performance Indicators Traditional methods Method of the present invention Improvement ratio Cut set partitioning time (s) 12.6 3.2 74.6％ Key line identification accuracy (%) 76.5 94.8 23.9％ Oscillation source positioning accuracy (%) 82.3 96.7 17.5％ System response delay (ms) 850 180 78.8％

[0156] Table 2 Comparison of detection capabilities

[0157]

[0158] Table 3 Economic Benefit Analysis

[0159] Index items Traditional methods Method of the present invention Improved results Annual average power outage time (h) 4.8 1.2 -75.0％ Oscillation fault handling cost (10,000 yuan / year) 386 125 -67.6％ System maintenance cost (10,000 yuan / year) 245 168 -31.4％ Economic loss (10,000 yuan / time) 156 42 -73.1％

[0160] Through three months of actual operation comparison, the solution of the present invention has shown significant advantages in multiple key performance indicators:

[0161] Comprehensive improvement of system performance:

[0162] As can be seen from the data in Table 1, the method of the present invention only takes 3.2 seconds to divide the cut set, which is 74.6% less than the traditional method. This is due to the fast binary segmentation algorithm using the Fiedler vector, which greatly improves the computational efficiency. The accuracy of key line identification reaches 94.8%, which is 23.9% higher than the traditional method, which is due to the synergistic effect of the multi-dimensional screening mechanism. In particular, in terms of the oscillation source positioning accuracy, a high accuracy of 96.7% was achieved, providing a reliable basis for the formulation of subsequent control strategies.

[0163] Significantly enhanced detection capabilities:

[0164] Table 2 compares the detection effects of oscillations of different frequencies in detail. In low-frequency oscillation detection, the detection rate of the present invention reaches 95.6%, while the false alarm rate is only 3.2%, which is 17.1 and 12.6 percentage points higher and lower respectively than the traditional method. This advantage is also obvious in medium-frequency and high-frequency oscillation detection. In particular, in terms of average positioning time, the method of the present invention only takes 5.2-6.8 seconds, which is about 75% shorter than the traditional method, which is crucial for quickly handling oscillation faults.

[0165] Outstanding economic benefits:

[0166] Table 3 reflects the economic benefits of the present invention in practical applications. The average annual power outage time was reduced from 4.8 hours to 1.2 hours, a reduction of 75%. The cost of oscillation fault handling was reduced from 3.86 million yuan to 1.25 million yuan, saving 2.61 million yuan per year. These improvements directly reflect the value of the present invention in practical applications. The economic loss caused by each oscillation event was also reduced from 1.56 million yuan to 420,000 yuan, which fully illustrates the economic benefits brought by fast and accurate positioning.

[0167] The system's adaptive capabilities have been improved, with the accuracy of identifying new oscillation modes reaching 92%. The computing resource usage has been reduced by 45%, significantly reducing the need for hardware investment. The work efficiency of operation and maintenance personnel has been improved by 65%, reducing the need for manual inspections. The life expectancy of power grid equipment has been extended by 15%, reducing the frequency of equipment replacement.

[0168] This test fully verifies the innovative value of the invention in practical applications. Compared with traditional methods, the present invention not only achieves breakthroughs in detection accuracy and response speed, but more importantly, it achieves breakthroughs in both economic benefits and technical effects, providing a complete solution for power grid oscillation detection and positioning.

[0169] Example 3

[0170] This embodiment also provides a computer device, which is applicable to a forced oscillation detection and positioning method for a distribution network, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a forced oscillation detection and positioning method for a distribution network as proposed in the above embodiment.

[0171] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0172] If the function is implemented in the form of 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, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0173] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0174] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0175] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0176] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, a forced oscillation detection and positioning method for a distribution network is implemented as proposed in the above embodiment.

[0177] In summary, by using the Fiedler vector for binary segmentation, the natural cut set structure in the power grid can be effectively identified, and the power grid can be divided into two subnets with stronger connectivity between the nodes in each subnet. The key line candidates are determined by the three dimensions of betweenness centrality, power fluctuation, frequency deviation and phase angle deviation, which can significantly improve the accuracy of identification. By combining the IMF decomposed by EMD with the Hilbert transform, the instantaneous frequency and amplitude can be extracted, and the intensity and frequency of oscillations on each key line can be monitored in real time. By constructing and using the LSTM model, dynamic analysis can be performed based on the time-frequency graph information to identify the frequency change trend over time and the change in oscillation intensity. By calculating the frequency deviation change rate of the nodes on each line, the node where the oscillation frequency changes first occurs can be accurately identified.

[0178] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A forced oscillation detection and positioning method for a distribution network, characterized in that: include, Obtain real-time operation data of the distribution network and use spectral graph theory to divide the power grid into cut sets; Analyze and screen key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set; Perform empirical mode decomposition and Hilbert transform on the selected key lines to construct time-frequency domain distribution features; The LSTM deep learning model is used to identify oscillation patterns, calculate line oscillation energy, and locate the maximum disturbance source.

2. A forced oscillation detection and positioning method for a distribution network as claimed in claim 1, characterized in that: The method of dividing the power grid by using the spectrum graph theory includes: Construct a graph theory model of the distribution network, with power stations, substations and load centers as nodes and lines and transformers as edges; Construct an adjacency matrix based on node connection relationships, and calculate the degree matrix and Laplace matrix; The eigenvector corresponding to the second smallest eigenvalue is selected as the Fiedler vector, and the nodes are assigned to two subnetworks according to whether they are greater than or equal to 0 or less than 0.

3. A forced oscillation detection and positioning method for a distribution network as claimed in claim 2, characterized in that: Calculate subnet energy transfer threshold based on historical data; When the subnet energy transfer is greater than or equal to the threshold, the cut set segmentation continues; When the subnet energy transfer is less than the threshold, the segmentation is stopped.

4. A forced oscillation detection and positioning method for a distribution network as claimed in claim 3, characterized in that: The analysis and screening of key lines include: calculating line betweenness centrality, selecting lines ranked 20% as first-category candidate lines; calculating power fluctuation values, selecting lines ranked 25% as second-category candidate lines; calculating frequency and phase angle deviations, selecting lines ranked 20% as third-category candidate lines; and determining the intersection of any two categories of candidate lines as the final key line.

5. A forced oscillation detection and positioning method for a distribution network as claimed in claim 4, characterized in that: The empirical mode decomposition includes: extracting the intrinsic mode functions of power fluctuation and frequency deviation signals; Calculate the mean of the upper and lower envelopes as the local trend; iteratively remove the local trend until the IMF definition is met.

6. A forced oscillation detection and positioning method for a distribution network as claimed in claim 5, characterized in that: The LSTM deep learning model includes: an input layer receiving time-frequency domain feature data; A two-layer LSTM network processes time series features; the fully connected layer outputs oscillation frequency and amplitude prediction values; and the mean square error is used as the loss function to optimize model parameters.

7. A forced oscillation detection and positioning method for a distribution network as claimed in claim 6, characterized in that: The method of locating the maximum disturbance source includes: calculating the oscillation energy of the key circuit; determining the maximum disturbance circuit by comparing the oscillation energy; calculating the change rate of the node frequency deviation; and determining the node with the largest frequency change rate as the oscillation source point.

8. A forced oscillation detection and positioning system for a distribution network, based on a forced oscillation detection and positioning method for a distribution network according to any one of claims 1 to 7, characterized in that: Also includes, The data acquisition module obtains the real-time operation data of the distribution network and uses the spectral graph theory to divide the power grid into cut sets; The key line identification module analyzes and screens key lines based on the betweenness centrality, power fluctuation value and frequency phase angle data of the cut set; The oscillation mode detection module performs empirical mode decomposition and Hilbert transform on the selected key lines to construct the time-frequency domain distribution characteristics; The oscillation source location module uses the LSTM deep learning model to identify the oscillation pattern, calculate the line oscillation energy, and locate the maximum disturbance source.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the forced oscillation detection and positioning method for a distribution network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a forced oscillation detection and positioning method for a distribution network as described in any one of claims 1 to 7 are implemented.

Images