Oscillation source positioning and control method and system based on power distribution network

Abstract

The invention relates to the technical field of power system maintenance, in particular to an oscillation source positioning and control method and system based on a power distribution network, and the method comprises the steps: collecting power grid time domain signals of all branches in the power distribution network, transmitting the power grid time domain signals to a control center, and carrying out the preprocessing of the power grid time domain signals; analyzing a power grid time domain signal of each branch, extracting a main oscillation mode, and calculating and optimizing instantaneous power in combination with wavelet transform; calculating an accumulation value of oscillation energy on each branch in the time period through integration by using a branch potential function, and constructing a feature matrix; and constructing a deep learning model according to the feature matrix to calculate oscillation source probabilities of all branches and determine a main oscillation source. According to the method, the instantaneous power is optimized by combining wavelet transformation and a branch potential function, the main oscillation mode is effectively extracted, a characteristic matrix is constructed, the oscillation source probability is accurately analyzed by using a deep learning model, so that the position of the main oscillation source is determined, and rapid identification and accurate positioning of the oscillation source are realized by dynamically adjusting power grid configuration and load control.

Description

Technical Field

[0001] The present invention relates to the technical field of power system maintenance, and in particular to an oscillation source positioning and control method and system based on a distribution network. Background Art

[0002] With the development of modern power systems, the structure of distribution networks is becoming increasingly complex. Especially after the large-scale access of distributed power sources, renewable energy and smart devices, power oscillations can cause grid frequency fluctuations, equipment damage and even large-scale power outages. How to effectively detect, locate and control oscillation sources has become an important issue for the stable operation of power systems. Traditional oscillation source detection methods mainly rely on the operating status of the power grid and the system dynamic model, and obtain the operating status of the system by real-time monitoring of time domain signals such as voltage and current. However, such methods have certain limitations in practical applications, such as difficulty in dealing with multi-source and multi-mode oscillation problems. At the same time, due to the wide distribution range of distribution networks, obtaining high-precision synchronous data is also a technical challenge. In recent years, with the continuous development of phasor measurement unit (PMU) technology and data-driven technology, oscillation source detection methods based on wide-area measurement systems have gradually become a research hotspot. PMU technology can provide high-precision, time-synchronized voltage and current phasor data, providing data support for accurate oscillation source positioning. However, in existing technical solutions, the data processing process is relatively simple, and it is difficult to accurately identify multiple oscillation modes in complex power grid environments. The positioning accuracy and response speed of oscillation sources need to be improved urgently.

[0003] The existing technologies have the following shortcomings when dealing with the location and control of oscillation sources in distribution networks. The traditional detection method based on dynamic models relies on accurate modeling of system parameters. However, it is difficult to accurately model the power grid system in a changeable operating environment, and the model is frequently updated and the maintenance cost is high. Although PMU technology can provide accurate time domain signal data, the data processing method mostly uses linear analysis and simple spectrum analysis, which is difficult to deal with nonlinear and time-varying oscillation modes. Especially in the distribution network, due to the large number of nodes and complex distribution, the propagation path of the oscillation energy is difficult to directly determine, resulting in inaccurate positioning results. In recent years, some studies have also attempted to apply signal processing technologies such as wavelet transform to oscillation analysis. By performing frequency decomposition and optimization processing on time domain signals, the detection accuracy can be improved to a certain extent. These methods still have problems of incomplete analysis and long positioning delay when dealing with complex and multi-modal oscillations. After the oscillation source is identified, the traditional method has a single control strategy and it is difficult to dynamically adjust the load according to the actual situation, resulting in the failure to completely solve the oscillation problem. How to combine the existing real-time monitoring technology of the power grid with advanced data analysis methods to improve the accuracy of oscillation source detection and the real-time control has become an urgent problem to be solved in the development of technology. 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 the traditional detection method based on dynamic models relies on accurate modeling of system parameters, while the power grid system is difficult to accurately model in a changeable operating environment, and the model is frequently updated and the maintenance cost is high. Although PMU technology can provide accurate time domain signal data, the data processing method mostly uses linear analysis and simple spectrum analysis, which is difficult to cope with nonlinear and time-varying oscillation modes. Especially in the distribution network, due to the large number of nodes and complex distribution, the propagation path of the oscillation energy is difficult to directly determine, resulting in inaccurate positioning results. In recent years, some studies have also attempted to apply signal processing technologies such as wavelet transform to oscillation analysis. By performing frequency decomposition and optimization processing on the time domain signal, the detection accuracy can be improved to a certain extent. These methods still have problems of incomplete analysis and long positioning delay when dealing with complex and multi-modal oscillations.

[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 an oscillation source positioning and control method based on a distribution network, which includes obtaining distribution network monitoring data, collecting power grid time domain signals and transmitting them to a control center by installing PMU devices at all branch nodes;

[0008] Preprocessing the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes;

[0009] Analyze the processed time domain signal, extract the main oscillation mode and calculate the optimized instantaneous power;

[0010] Construct the characteristic matrix and calculate the oscillation energy accumulation value;

[0011] Based on the feature matrix, deep learning analysis is performed to determine the location of the oscillation source and execute a corresponding control strategy.

[0012] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, wherein: the node includes the intersection of the power station, the substation, the load center and the transmission line;

[0013] The time domain signal includes voltage signal, current signal, frequency and phase angle data.

[0014] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, wherein: the step of extracting the main oscillation mode includes: extracting the oscillation mode information using the Prony analysis method;

[0015] Analyze the damping coefficient of the oscillation mode and mark the mode with damping coefficient ≤ 0 as stable state;

[0016] Set the modal frequency range to 0.1-2 Hz to filter out normal oscillation modes;

[0017] Calculate the oscillation mode contributions and identify the dominant oscillation mode.

[0018] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, wherein: the instantaneous power after the optimization of the calculation includes: analyzing the time domain signal in combination with Morlet wavelet transform; obtaining the transformation coefficient matrix through wavelet scale and shift;

[0019] The maximum transformation coefficient is selected according to the main oscillation mode frequency; the wavelet transformation coefficient, instantaneous power and modal contribution are combined to calculate the optimized power.

[0020] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, the construction of the characteristic matrix includes: using the branch potential function to calculate the accumulated value of the oscillation energy; constructing a characteristic matrix containing the accumulated values ​​of each branch in different time periods; and standardizing the characteristic matrix.

[0021] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, the deep learning analysis includes: using a convolutional neural network to extract features; processing timing features through a long short-term memory network; calculating the oscillation source probability using a Softmax function; and determining the branch with the maximum probability as the main oscillation source.

[0022] As a preferred solution of the oscillation source positioning and control method based on the distribution network described in the present invention, three risk levels are set according to the oscillation energy accumulation value:

[0023] Low risk: When the accumulated value is ≤0, perform regular configuration adjustments;

[0024] Medium risk: when the accumulated value is >0 and ≤ the threshold, reduce the load;

[0025] High risk: When the accumulation value > threshold, the load is switched and the reactive power compensation equipment is started.

[0026] In a second aspect, an embodiment of the present invention provides an oscillation source positioning and control system based on a distribution network, which includes a data acquisition module to obtain distribution network monitoring data, and collects power grid time domain signals and transmits them to a control center by installing PMU devices at all branch nodes;

[0027] A preprocessing module preprocesses the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes;

[0028] An analysis module analyzes the processed time domain signal, extracts the main oscillation mode and calculates the optimized instantaneous power;

[0029] Construct the module, construct the feature matrix, and calculate the oscillation energy accumulation value;

[0030] The execution module performs deep learning analysis based on the feature matrix to determine the location of the oscillation source and execute a corresponding control strategy.

[0031] 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 an oscillation source positioning and control method based on a distribution network as described in the first aspect of the present invention are implemented.

[0032] 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 method for locating and controlling an oscillation source based on a distribution network as described in the first aspect of the present invention are implemented.

[0033] The beneficial effects of the present invention are as follows: by combining wavelet transform and branch potential function to optimize instantaneous power, effectively extract the main oscillation mode, construct a feature matrix, and use a deep learning model to accurately analyze the probability of the oscillation source, thereby determining the location of the main oscillation source, and by dynamically adjusting the grid configuration and load control, the oscillation source can be quickly identified and accurately located. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] 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.

[0035] Figure 1 It is a flow chart of a method and system for locating and controlling an oscillation source based on a distribution network;

[0036] Figure 2 A computer device diagram of an oscillation source location and control method based on a distribution network. DETAILED DESCRIPTION

[0037] 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.

[0038] 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.

[0039] 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.

[0040] Example 1

[0041] Reference Figure 1-2 , which is the first embodiment of the present invention, provides an oscillation source positioning and control method based on a distribution network, comprising:

[0042] S100: Obtain distribution network monitoring data by installing PMU devices at all branch nodes to collect power grid time domain signals and transmit them to the control center;

[0043] S101: Nodes include power stations, substations, load centers, and intersections of transmission lines;

[0044] Time domain signals include voltage signals, current signals, frequency and phase angle data.

[0045] Specifically, real-time collection of power grid time domain signals and transmission to the control center refers to installing PMU devices at nodes of all branches in the distribution network, configuring the sampling frequency of the PMU devices, regularly calibrating the clock of each PMU device, and collecting power grid time domain signals after the calibration is completed;

[0046] PMU equipment achieves accurate data collection across regions through high-precision GPS clock synchronization, ensuring the accuracy and consistency of power grid time domain signals. Phase angle collection can help calculate power fluctuations and accurately locate the propagation path and oscillation source of oscillation energy.

[0047] All branches include transmission lines (long-distance transmission from power stations to substations), distribution lines (medium and low voltage distribution lines from substations to end users) and inter-regional lines (ultra-high voltage or high voltage direct current transmission lines connecting power grids in different regions);

[0048] Nodes include power stations, substations, load centers (such as industrial areas or urban core areas), and intersections of transmission lines (connections of multiple important transmission lines, distribution points for power flow);

[0049] The power grid time domain signals include voltage signals, current signals, frequency and phase angle;

[0050] The collected power grid time domain signals are transmitted to the control center through the wide area measurement system;

[0051] The wide area measurement system is an advanced monitoring and data acquisition system used in power systems. It monitors the dynamic operating status of the power system in real time through PMU devices distributed throughout the distribution network.

[0052] By installing PMU equipment at all key nodes of the distribution network and using high-precision GPS clock synchronization to achieve accurate data collection across regions, it is possible to monitor the voltage, current, frequency and phase angle time domain signals of the power grid in real time. With the help of a wide-area measurement system, these data are transmitted to the control center to ensure the accuracy and consistency of the power grid signal, thereby effectively calculating power fluctuations and accurately analyzing the propagation path and source of oscillation energy. This method helps to grasp the dynamic operating status of the power system in real time, improve the ability to quickly identify and locate oscillation sources, and enhance the safety and stability of the power grid.

[0053] S200: preprocessing the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes;

[0054] Furthermore, the preprocessing of the power grid time domain signal includes using a low-pass filter to remove high-frequency noise and harmonic interference from the power grid time domain signal, and applying a median filter to remove sudden data spikes.

[0055] Preprocessing the power grid time domain signal and removing high-frequency noise and harmonic interference through a low-pass filter can effectively reduce unnecessary high-frequency components and make the signal smoother and more stable. The application of median filtering can remove sudden data spikes, suppress abnormal fluctuations, and avoid interference with subsequent oscillation analysis and model calculations. This preprocessing method can improve the quality and accuracy of the signal and provide a more reliable data basis for subsequent oscillation source analysis.

[0056] S300: analyzing the processed time domain signal, extracting the main oscillation mode and calculating the optimized instantaneous power;

[0057] S301: The step of extracting the main oscillation mode includes: extracting oscillation mode information using the Prony analysis method;

[0058] Analyze the damping coefficient of the oscillation mode and mark the mode with damping coefficient ≤ 0 as stable state;

[0059] Set the modal frequency range to 0.1-2 Hz to filter out normal oscillation modes;

[0060] Calculate the oscillation mode contributions and identify the dominant oscillation mode.

[0061] Specifically, the grid time domain signal of each branch is analyzed and the main oscillation mode is extracted. The instantaneous power index is calculated and optimized by combining wavelet transform. According to the preprocessed grid time domain signal, the instantaneous power of each branch in the grid time domain signal at different time points is calculated. The formula is:

[0062] P i (t) = V i (t)·I i (t)·cos(θ i,a (t)-θ i,b (t))

[0063] Among them, P i (t) represents the instantaneous power of the i-th branch at time t, V i (t) represents the voltage value of the i-th branch at time point t, I i (t) represents the current value of the i-th branch at time point t, θ i,a (t) and θ i,b (t) represents the phase angle between branch a and branch b of the i-th branch at time t. In the power system, the larger the phase angle difference, the stronger the power transmission of the branch;

[0064] Prony analysis is used to process the power grid time domain signal through the least square method to extract the information of all oscillation modes in each branch;

[0065] Oscillation mode refers to the inherent mode that exhibits oscillation behavior in a dynamic power system. The information of each oscillation mode includes oscillation frequency, damping coefficient and modal amplitude;

[0066] The damping coefficients of all oscillation modes in each branch are analyzed. If the damping coefficient of the oscillation mode is less than or equal to 0, it means that the branch is in a stable state and the oscillation energy remains stable. The oscillation mode is subjected to modal frequency analysis. If the damping coefficient of the oscillation mode is greater than 0, it means that the branch is in an unstable oscillation and the oscillation energy is gradually amplified, and it does not participate in further analysis.

[0067] Analyze the modal frequency of the oscillation mode. According to the main oscillation modal frequency in the power system, set the modal frequency range between 0.1Hz and 2Hz. Oscillation modes beyond this range are marked as abnormal and do not participate in further analysis. Otherwise, mark the oscillation mode as normal and perform contribution analysis.

[0068] According to the modal amplitude of each oscillation mode, the contribution of each oscillation mode is calculated. The formula is:

[0069]

[0070] Among them, Cx represents the contribution of the xth oscillation mode. The sum of the contributions of all oscillation modes is 1. x represents the amplitude of the xth oscillation mode, and n represents the total number of oscillation modes marked as normal after modal frequency analysis;

[0071] The contribution of the calculated oscillation modes is sorted, and the oscillation mode with the largest contribution is extracted as the main oscillation mode of the corresponding branch according to actual needs;

[0072] The Morlet wavelet is used to perform wavelet transform on the power grid time domain signal. By adjusting the wavelet scale and shift, the wavelet transform coefficient matrix is ​​obtained to capture the wavelet transform coefficients of the voltage signal and current signal in each branch at different time points and different frequencies.

[0073] The wavelet transform coefficient is the projection result of the power grid time domain signal, which quantifies the energy intensity of the power grid time domain signal within a certain time and frequency range;

[0074] According to the center frequency of the Morlet wavelet, the frequency corresponding to each wavelet transform coefficient is calculated. The formula is:

[0075]

[0076] Among them, f represents the frequency corresponding to the wavelet transform coefficient, f M represents the central frequency of the Morlet wavelet, and h represents the wavelet scale of the wavelet transform coefficient;

[0077] According to the oscillation frequency of the main oscillation mode, the wavelet scale is set, and the wavelet transform coefficients with the largest value corresponding to the frequency range of the oscillation frequency of the main oscillation mode at different time points are traversed in the wavelet transform coefficient matrix, and the obtained wavelet transform coefficients are normalized;

[0078] S302: Calculating the optimized instantaneous power includes: analyzing the time domain signal in combination with Morlet wavelet transform; obtaining a transform coefficient matrix through wavelet scaling and shifting;

[0079] The maximum transformation coefficient is selected according to the main oscillation mode frequency; the wavelet transformation coefficient, instantaneous power and modal contribution are combined to calculate the optimized power.

[0080] The optimized instantaneous power is calculated by combining the wavelet transform coefficient, instantaneous power and contribution of the corresponding main oscillation mode of each branch at different time points. The formula is:

[0081] P o,i (t) = C i ·P i (t)·W i (t)

[0082] Among them, P o,i (t) represents the weighted instantaneous power of the i-th branch at time t, C i represents the contribution of the main oscillation mode corresponding to the i-th branch, P i (t) represents the instantaneous power of the i-th branch at time point t, W i (t) represents the wavelet transform coefficient of the ith branch corresponding to the main oscillation mode at time point t;

[0083] By screening out the main oscillation modes and weighting them, the calculation can be simplified and the main oscillation modes can be ensured to make the greatest contribution to the optimization of instantaneous power, making the calculation more efficient. At the same time, the focus is on the modes that have the greatest impact on the location of the oscillation source, thereby improving the accuracy of positioning and better reflecting the oscillation energy flow in the power grid.

[0084] By analyzing the time domain signals of the power grid, extracting and optimizing the oscillation modal information of each branch, it has a significant beneficial effect. By using Prony analysis combined with the least squares method to process the time domain signals, the oscillation modal information in each branch can be accurately extracted, including the oscillation frequency, damping coefficient and modal amplitude, which provides an accurate data basis for identifying the oscillation source in the power system. By calculating the damping coefficient, the stability of each branch can be judged. If the damping coefficient of the oscillation mode is less than or equal to 0, it means that the branch is stable and there will be no energy diffusion. This analysis helps to distinguish between stable and unstable oscillation sources and simplify subsequent processing. By setting a reasonable modal frequency range, the modes that meet the specific oscillation frequency are screened out to ensure the accuracy of the analysis, avoid the processing of abnormal or irrelevant modes, and help improve the stability of the system. Using the modal amplitude of the oscillation mode, the contribution of each oscillation mode is calculated, which can help identify the main oscillation mode that has the greatest impact on the stability of the system. Contribution sorting can quickly identify the main oscillation source mode. This method ensures that in complex power systems, the main oscillation mode can be optimized and processed in a targeted manner, thereby improving the efficiency of oscillation source positioning. Combined with the application of wavelet transform, especially the use of Morlet wavelet transform, it can capture the characteristics of voltage and current signals at different time points and frequencies, and effectively reflect the dynamic changes in the power grid. By adjusting the wavelet scale, the instantaneous power can be refined and optimized, making the projection results of the power grid time domain signal more intuitive and clearly showing the oscillation frequency and energy intensity. By combining the wavelet transform coefficients and instantaneous power of each branch at different time points with the contribution of the main oscillation mode, the optimized instantaneous power is calculated, which can effectively simplify the calculation and ensure that the main oscillation mode plays a leading role in the optimization process, thereby improving the calculation efficiency and accuracy, and can highlight the mode that has the greatest impact on the positioning of the power grid oscillation source, making the oscillation source positioning more accurate and efficient.

[0085] S400: construct a characteristic matrix and calculate the oscillation energy accumulation value;

[0086] S401: constructing a characteristic matrix includes: using a branch potential function to calculate the oscillation energy accumulation value; constructing a characteristic matrix containing the accumulation values ​​of each branch in different time periods; and standardizing the characteristic matrix.

[0087] Specifically, based on the optimized instantaneous power, the accumulated value of the oscillation energy on each branch in the time period is calculated by integration using the branch potential function and a characteristic matrix is ​​constructed. The accumulated value of the oscillation energy on each branch in the time period is calculated using the branch potential function, and the formula is:

[0088]

[0089] Among them, R z,i( T) represents the accumulated value of the oscillation energy of the i-th branch in the acquisition time period T, t0 and t1 represent the initial time and the end time of the acquisition time period T respectively, P o,i( t) represents the weighted instantaneous power of the i-th branch at time point t, P w,i represents the steady-state power of the ith branch. The steady-state power of each branch is obtained from the historical power grid time-frequency signal. i( t) represents the angular frequency difference between the two ends of the i-th branch at time point t. The angular frequency difference between the two ends of the branch is obtained by differential calculation of the phase angles at the two ends of the branch. g(t) represents the time attenuation function, and dt represents the small change of the time integral t.

[0090] The expression of the time decay function is: g(t) = e-αt, where α represents the time decay factor, which controls the decay speed of g(t);

[0091] The accumulated value of the oscillation energy of each branch during the acquisition period is used as the input feature to construct the feature matrix R. The expression of the feature matrix R is:

[0092]

[0093] Among them, R z,m( T y) represents the accumulated value of the oscillation energy of the mth branch in the yth time period T, where y represents the number of acquisition time periods;

[0094] The constructed feature matrix R is standardized.

[0095] By integrating and calculating the weighted instantaneous power within the acquisition time period, and combining the angular frequency difference and time decay function at both ends of the branch, the dynamic accumulation state of the oscillation energy in the power grid can be accurately reflected. The combination of weighted instantaneous power and steady-state power enables the calculation process to effectively distinguish normal power transmission from abnormal oscillation behavior, thereby more accurately evaluating the oscillation of each branch. The introduction of the time decay function enhances the processing capability of historical data, so that the influence of historical signals far away from the current moment on the energy accumulation value gradually decreases, thereby ensuring the real-time and effectiveness of the oscillation analysis. The branch potential function is used to calculate the oscillation energy accumulation value of each branch in a specific time period, which can capture the oscillation intensity of each branch in a specific time period, help identify potential oscillation sources, construct a feature matrix and perform standardized processing, which helps to ensure data consistency and stability in subsequent machine learning models. The feature matrix systematically arranges the oscillation energy accumulation values ​​of each branch in different time periods, so that the model can comprehensively consider the oscillation characteristics in different time periods, thereby improving the accuracy and reliability of oscillation source analysis.

[0096] S500: Perform deep learning analysis based on the feature matrix to determine the location of the oscillation source and execute the corresponding control strategy.

[0097] S501: Deep learning analysis includes: extracting features using convolutional neural networks; processing time series features through long short-term memory networks; calculating the probability of oscillation sources using the Softmax function; and determining the branch with the maximum probability as the main oscillation source.

[0098] Specifically, building a deep learning model based on the feature matrix to calculate the oscillation source probabilities of all branches and determine the main oscillation source refers to building a deep learning model by combining a convolutional neural network and a long short-term memory network;

[0099] Using a convolutional neural network, a convolution operation is performed on the feature matrix R through a convolution kernel in the convolution layer. 64 convolution kernels are set. The result of each convolution operation forms a feature map, and 64 feature maps are obtained. The 64 feature maps are combined into a feature map matrix F.

[0100] Through actual verification, it is found that 64 convolution kernels can achieve a good balance between computational efficiency and feature extraction capability, so that the model has high generalization ability while ensuring performance;

[0101] The feature mapping matrix F retains the time series features and spatial features in the feature matrix R, and its size depends on the settings of the convolution kernel and step size;

[0102] In the pooling layer, the maximum pooling is used to perform a pooling operation on the feature mapping matrix F of the convolutional layer;

[0103] After the first convolution and pooling operation, add another convolution layer and pooling layer, set the number of convolution kernels of the second convolution layer to 128, further extract high-level features, and obtain the feature matrix G;

[0104] Using the LSTM network, the feature matrix G is flattened at the input layer to form a vector as the input of the LSTM network;

[0105] Use two stacked LSTM layers, each with 100 LSTM units. Each LSTM unit processes the current input and the previous state, and updates the current state and output.

[0106] Pass the output of the last LSTM layer to the fully connected layer to generate the classification score of each branch;

[0107] The classification score of each branch is converted into the probability of the oscillation source of each branch using the Softmax function;

[0108] The cross entropy loss function is used to measure the difference between the prediction results of the deep learning model and the true label. By minimizing the loss function, the prediction accuracy of the deep learning model is optimized.

[0109] The stacking of convolutional layers can help the model capture more complex feature patterns, making it more adaptable to different oscillation modes;

[0110] After calculating the oscillation source probabilities of all branches, the branch with the highest probability is selected as the main oscillation source.

[0111] By combining convolutional neural networks and long short-term memory networks, the accuracy and efficiency of oscillation source positioning are effectively improved. Convolutional neural networks are used to perform convolution operations on the feature matrix. Two convolutional layers with 64 and 128 convolution kernels are used to help the model extract rich time series and spatial features from the feature matrix and achieve a deep understanding of complex oscillation patterns. After the convolution operation, the maximum pooling layer is used to further compress the feature map, remove redundant information, and improve the generalization ability of the model. The extracted high-level features are flattened and input into the stacked LSTM layer. The LSTM layer can capture the temporal dependencies in the features, enabling the model to effectively handle the dynamic changes in the power grid oscillation and accurately identify the oscillation. The behavior pattern of the source is output through the fully connected layer, and the classification score of each branch is output, and the probability of the oscillation source is calculated using the Softmax function. The branch with the largest probability is further selected as the main oscillation source. The entire model is optimized through the cross entropy loss function, which helps to improve the accuracy of the model prediction results. The combination of convolutional layer and LSTM layer can not only capture complex spatial and temporal characteristics, but also accurately locate the oscillation source through multi-layer feature extraction and timing analysis. Compared with traditional methods, this comprehensive solution based on deep learning can more efficiently handle complex power grid oscillation patterns, making the oscillation source positioning more accurate. The model also has strong adaptability and robustness when dealing with various types of oscillations.

[0112] S502: Set three risk levels according to the oscillation energy accumulation value:

[0113] Low risk: When the accumulated value is ≤0, perform regular configuration adjustments;

[0114] Medium risk: when the accumulated value is >0 and ≤ the threshold, reduce the load;

[0115] High risk: When the accumulation value > threshold, the load is switched and the reactive power compensation equipment is started.

[0116] Specifically, dynamically adjusting the grid configuration and load control according to the accumulated oscillation energy value of the main oscillation source means that after determining the branch of the main oscillation source, the accumulated value of the oscillation energy of the branch in all acquisition time periods is analyzed respectively and the oscillation energy threshold is set to judge the oscillation severity and trend of the branch in the acquisition time period;

[0117] If the accumulated value of the oscillation energy is less than or equal to 0, it means that the oscillation energy is decreasing, the distribution network system tends to be stable, and it is judged to be low risk, and the grid configuration is adjusted proactively;

[0118] Record the current status and upload the data to the control center in real time for subsequent analysis;

[0119] If the accumulated value of the oscillation energy is greater than 0 and less than or equal to the oscillation energy threshold, it means that the oscillation energy is slightly increasing and is judged as medium risk. The load in the area where the oscillation source branch is located is reduced;

[0120] If the main oscillation source is associated with a specific generator set, the output power of that generator set is fine-tuned to keep its output within a controllable range;

[0121] If the accumulated value of the oscillation energy is greater than the oscillation energy threshold, it means that the oscillation energy is increasing rapidly and is judged to be high risk. Through the load switching strategy, part of the load in the branch area corresponding to the main oscillation source is switched to the backup line and the reactive power compensation equipment is started immediately;

[0122] After the control strategy is executed, the real-time status and executed control strategy are notified to all operation and maintenance personnel, a confirmation signal is sent to the control center through the wide area monitoring system, and the log is recorded in the central database for subsequent analysis and improvement.

[0123] By real-time monitoring and analysis of the energy accumulation value of the main oscillation source, the current oscillation risk level of the power grid can be accurately determined, so that corresponding control measures can be taken. The entire dynamic adjustment process can not only flexibly respond to oscillation problems at different risk levels, but also ensure the comprehensive and transparent transmission of information through real-time data upload and control strategy execution confirmation, so that operation and maintenance personnel can grasp the system status in a timely manner, facilitate real-time monitoring and post-analysis. This multi-level control scheme improves the stability and reliability of the power grid system, reduces the potential accident risks caused by oscillation, and ensures the efficient and safe operation of the power system.

[0124] Furthermore, this embodiment also provides an oscillation source positioning and control system based on a distribution network, comprising:

[0125] The data acquisition module obtains the distribution network monitoring data. By installing PMU devices at all branch nodes, it collects the grid time domain signals and transmits them to the control center;

[0126] A preprocessing module preprocesses the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes;

[0127] An analysis module analyzes the processed time domain signal, extracts the main oscillation mode and calculates the optimized instantaneous power;

[0128] Construct the module, construct the feature matrix, and calculate the oscillation energy accumulation value;

[0129] The execution module performs deep learning analysis based on the feature matrix to determine the location of the oscillation source and execute the corresponding control strategy.

[0130] In summary, the present invention optimizes instantaneous power by combining wavelet transform and branch potential function, effectively extracts the main oscillation mode, constructs a feature matrix, and uses a deep learning model to accurately analyze the probability of the oscillation source, thereby determining the location of the main oscillation source. By dynamically adjusting the grid configuration and load control, the present invention achieves rapid identification and precise positioning of the oscillation source.

[0131] Example 2

[0132] Reference Figure 1 - Figure 2 , which is the second embodiment of the present invention, and this embodiment provides an oscillation source positioning and control method based on 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 distribution network of a megacity as the verification object. The area covers an area of ​​about 280 square kilometers, including 3 500kV substations, 12 220kV substations, 42 110kV substations, and multiple distributed photovoltaic power stations and energy storage facilities. The test time is from September 1, 2023 to November 30, 2023, a total of 91 days.

[0134] In the early stage of the experiment, the following preparations were completed:

[0135] 162 PMU devices were installed at all key nodes, with a sampling frequency of 100Hz and a time synchronization accuracy better than 1μs;

[0136] A data transmission system based on 5G network has been built, with data delay of <10ms;

[0137] A deep learning model consisting of 8-layer CNN and 2-layer LSTM was built and pre-trained using historical data;

[0138] Twenty edge computing servers have been deployed to process PMU data locally.

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

[0140] Area A: using the oscillation source positioning and control method of the present invention;

[0141] Area B: Using the traditional oscillation detection method based on spectrum analysis;

[0142] The following indicators are monitored during the test:

[0143] Oscillation source positioning accuracy and response time; system stability indicators; control strategy execution effect; computing resource usage.

[0144] Table 1 Comparison of oscillation source positioning performance

[0145] index Traditional methods Method of the present invention Improvement Positioning accuracy (%) 78.3 96.5 23.2％ Average response time (ms) 850 180 78.8％ False alarm rate (%) 12.6 3.2 74.6％ Missing rate (%) 15.8 2.8 82.3％

[0146] Table 2 Comparison of system stability indicators

[0147] index Traditional methods Method of the present invention Improvement System damping ratio (%) 3.2 5.8 81.3％ Frequency deviation(Hz) ±0.28 ±0.12 57.1％ Voltage fluctuation rate (%) 2.8 1.3 53.6％ Power fluctuation rate (%) 4.2 1.8 57.1％

[0148] Table 3 Control effect comparison

[0149] Scene Type Average processing time (s) Oscillation suppression rate (%) System recovery time (min) Traditional method - mild vibration 25.3 82.5 8.5 The Inventive Method - Mild Oscillation 5.8 95.8 3.2 Traditional method - moderate oscillation 42.6 75.3 15.8 The Inventive Method - Moderate Oscillation 8.2 92.4 5.5 Traditional method - heavy vibration 68.5 65.8 25.3 The Inventive Method - Severe Oscillation 12.5 88.6 8.8

[0150] Through a 91-day actual operation comparison test, the solution of the present invention showed significant advantages in multiple key performance indicators:

[0151] Positioning performance is greatly improved:

[0152] As can be seen from the data in Table 1, the oscillation source positioning accuracy of the method of the present invention reaches 96.5%, which is 23.2% higher than that of the traditional method. More significantly, the average response time is reduced from 850ms to 180ms, an increase of 78.8%. This is due to the multi-level signal processing method and deep learning model adopted by the present invention. In particular, the false alarm rate and missed alarm rate are reduced by 74.6% and 82.3% respectively, which is of great significance for the safe operation of the power grid.

[0153] System stability has been significantly improved:

[0154] Table 2 shows that after adopting the method of the present invention, the system damping ratio is increased from 3.2% to 5.8%, an increase of 81.3%. The frequency deviation is controlled within the range of ±0.12Hz, which is 57.1% better than the traditional method. The voltage and power fluctuation rates are reduced by more than 50%, which shows that the method of the present invention can effectively suppress system oscillation and improve network stability.

[0155] Excellent hierarchical control effect:

[0156] Table 3 compares the control effects under different oscillation degrees in detail. In the case of mild oscillation, the average processing time of the method of the present invention is only 5.8 seconds, the oscillation suppression rate reaches 95.8%, and the system recovery time is shortened to 3.2 minutes. In particular, when dealing with severe oscillation, the method of the present invention can still maintain a high suppression rate of 88.6%, while the traditional method is only 65.8%. This fully proves the effectiveness of the three-level risk control strategy proposed by the present invention.

[0157] The system computing resource usage is reduced by 35%, thanks to the optimized signal processing algorithm; the life expectancy of power grid equipment is extended by 20%, thanks to more precise oscillation control; and the work efficiency of operation and maintenance personnel is improved by 45%, thanks to the accurate early warning mechanism.

[0158] This test fully verifies the innovative value of the invention in practical applications. Compared with traditional methods, the present invention not only achieves breakthrough progress in oscillation source positioning accuracy and response speed, but more importantly, it achieves a qualitative leap in system stability and control effect, providing reliable technical support for the intelligent operation and maintenance of large-scale distribution networks.

[0159] Example 3

[0160] 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.

[0161] 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.

[0162] 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.

[0163] 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.

[0164] 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.

[0165] 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.

[0166] 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.

[0167] 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.

[0168] 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 method for locating and controlling an oscillation source based on a distribution network, characterized in that: include, Obtain distribution network monitoring data by installing PMU devices at all branch nodes, collecting grid time domain signals and transmitting them to the control center; Preprocessing the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes; Analyze the processed time domain signal, extract the main oscillation mode and calculate the optimized instantaneous power; Construct the characteristic matrix and calculate the oscillation energy accumulation value; Based on the feature matrix, deep learning analysis is performed to determine the location of the oscillation source and execute a corresponding control strategy.

2. The method for locating and controlling an oscillation source based on a distribution network according to claim 1, characterized in that: The nodes include power stations, substations, load centers and intersections of transmission lines; The time domain signal includes voltage signal, current signal, frequency and phase angle data.

3. The method for locating and controlling an oscillation source based on a distribution network according to claim 2, characterized in that: The step of extracting the main oscillation mode comprises: extracting oscillation mode information using the Prony analysis method; Analyze the damping coefficient of the oscillation mode and mark the mode with damping coefficient ≤ 0 as stable state; Set the modal frequency range to 0.1-2 Hz to filter out normal oscillation modes; Calculate the oscillation mode contributions and identify the dominant oscillation mode.

4. The method for locating and controlling an oscillation source based on a distribution network according to claim 3, characterized in that: The calculation of the optimized instantaneous power includes: analyzing the time domain signal in combination with Morlet wavelet transform; obtaining the transform coefficient matrix through wavelet scale and shift; The maximum transformation coefficient is selected according to the main oscillation mode frequency; the wavelet transformation coefficient, instantaneous power and modal contribution are combined to calculate the optimized power.

5. The method for locating and controlling an oscillation source based on a distribution network according to claim 4, characterized in that: The construction of the characteristic matrix includes: using the branch potential function to calculate the oscillation energy accumulation value; constructing a characteristic matrix containing the accumulation values ​​of each branch in different time periods; and performing standardization processing on the characteristic matrix.

6. The method for locating and controlling an oscillation source based on a distribution network according to claim 5, characterized in that: The deep learning analysis includes: extracting features using a convolutional neural network; processing time series features through a long short-term memory network; calculating the probability of an oscillation source using a Softmax function; and determining the branch with the maximum probability as the main oscillation source.

7. The method for locating and controlling an oscillation source based on a distribution network according to claim 6, characterized in that: Three risk levels are set according to the oscillation energy accumulation value: Low risk: When the accumulated value is ≤0, perform regular configuration adjustments; Medium risk: when the accumulated value is >0 and ≤ the threshold, reduce the load; High risk: When the accumulation value > threshold, the load is switched and the reactive power compensation equipment is started.

8. An oscillation source positioning and control system based on a distribution network, based on the oscillation source positioning and control method based on a distribution network according to any one of claims 1 to 7, characterized in that: Also includes, The data acquisition module obtains the distribution network monitoring data. By installing PMU devices at all branch nodes, it collects the grid time domain signals and transmits them to the control center; A preprocessing module preprocesses the power grid time domain signal, including using a low-pass filter to remove high-frequency noise and a median filter to remove data spikes; An analysis module analyzes the processed time domain signal, extracts the main oscillation mode and calculates the optimized instantaneous power; Construct the module, construct the feature matrix, and calculate the oscillation energy accumulation value; The execution module performs deep learning analysis based on the feature matrix to determine the location of the oscillation source and execute a corresponding control strategy.

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 method for locating and controlling an oscillation source based on a distribution network according to 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 the method for locating and controlling an oscillation source based on a distribution network according to any one of claims 1 to 7 are implemented.

Images