Transient overvoltage identification method and device and nonvolatile storage medium

Abstract

The invention discloses a transient overvoltage identification method and device and a nonvolatile storage medium. The method comprises the following steps: acquiring a real-time transient overvoltage signal in a power system; based on a preset first extraction model, a first feature is extracted from the real-time transient overvoltage signal, the first extraction model is a convolutional neural network model, and the first feature is a time-frequency feature; based on a preset second extraction model, a second feature is extracted from the real-time transient overvoltage signal, the second extraction model is a deep long-short term memory network model, and the second feature is a time feature; fusing the first feature and the second feature to obtain a fused feature; and based on the fusion feature, determining an identification result of the real-time transient overvoltage signal, the identification result including a transient overvoltage type. According to the method, the technical problem that the traditional transient overvoltage identification technology is insufficient in the aspects of feature extraction comprehensiveness, model robustness and real-time performance is solved.

Description

Technical Field

[0001] This invention relates to the field of power system safety, and more specifically, to a transient overvoltage identification method, apparatus, and non-volatile storage medium. Background Technology

[0002] Transient overvoltages are common anomalies in power system operation, caused by sudden factors such as lightning strikes, switching operations, or short-circuit faults. They pose a serious challenge to the insulation performance of electrical equipment and may trigger relay protection malfunctions, threatening the safety and stability of the power grid. With the increasing complexity of power grid structures, especially the large-scale integration of new energy sources and the widespread application of DC transmission systems, transient processes have become more variable, highlighting the limitations of traditional overvoltage identification technologies. Most rely on threshold comparisons or single-dimensional (time domain or frequency domain) feature analysis. For example, while lightning overvoltages and switching overvoltages may have similar amplitudes, their harmonic distributions and time-series decay patterns are fundamentally different. However, single-feature-domain analysis struggles to capture these subtle differences, especially when the signal is subjected to strong noise interference or exhibits non-stationary characteristics, significantly reducing classification accuracy. Furthermore, this method cannot achieve the required accuracy when dealing with overvoltages with similar waveforms but different causes.

[0003] Current technology proposes a deep learning-based overvoltage identification technique, which improves the accuracy of feature extraction and the robustness of the classification model by introducing deep neural networks. However, it fails to fully utilize the complex nonlinear combinatorial capabilities of the deep learning framework, potentially leading to redundancy or loss of feature information. Furthermore, despite employing a deep learning model, the aforementioned model does not adequately consider computational efficiency and resource consumption, especially in real-time monitoring scenarios where model inference time may be excessively long, making it difficult to meet the millisecond-level response requirements of smart grids.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a transient overvoltage identification method, apparatus, and non-volatile storage medium to at least address the technical problems of traditional transient overvoltage identification technology in terms of feature extraction comprehensiveness, model robustness, and real-time performance.

[0006] According to one aspect of the present invention, a transient overvoltage identification method is provided, comprising: acquiring a real-time transient overvoltage signal in a power system; extracting a first feature from the real-time transient overvoltage signal based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; extracting a second feature from the real-time transient overvoltage signal based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; fusing the first feature and the second feature to obtain a fused feature; and determining the identification result of the real-time transient overvoltage signal based on the fused feature, wherein the identification result includes the transient overvoltage type.

[0007] Optionally, based on a preset first extraction model, a first feature is extracted from the real-time transient overvoltage signal, including: performing a time-frequency transformation on the real-time transient overvoltage signal through a short-time Fourier transform to obtain a time-frequency map of the real-time transient overvoltage signal; extracting initial features from the time-frequency map based on a convolutional layer of the first extraction model; performing dimensionality reduction processing on the initial features based on a pooling layer of the first extraction model to obtain a multi-dimensional feature map; and flattening the multi-dimensional feature map into a one-dimensional vector based on a fully connected layer of the first extraction model as the first feature.

[0008] Optionally, based on a preset second extraction model, a second feature is extracted from the real-time transient overvoltage signal, including: converting the real-time transient overvoltage signal into normalized time-series data, wherein the time-series data includes transient overvoltage data within multiple time intervals; within any time interval, based on the second extraction model, calculating the forget gate activation value, input gate activation value, and new memory unit candidate value of the transient overvoltage data; updating the memory units of the second extraction model based on the forget gate activation value, input gate activation value, and new memory unit candidate value, and calculating the output gate activation value; updating the hidden layer state of the second extraction model based on the output gate activation value; repeating the above steps until all time intervals are traversed to obtain the final hidden layer state, which is used as the second feature.

[0009] Optionally, the first feature and the second feature are fused to obtain a fused feature, including: concatenating the second feature after the first feature to obtain the fused feature.

[0010] Optionally, based on the fusion features, the identification result of the real-time transient overvoltage signal is determined, including: inputting the fusion features into a preset target classification model to obtain the identification result, wherein the target classification model is trained on an initial classification model using a training dataset, the initial classification model being a deep convolutional neural network model, and the training dataset including simulation data and historical data.

[0011] Optionally, the steps for obtaining simulation data are as follows: construct a simulation model of the power system, which includes transmission lines, transformers, and circuit breakers; set multiple operating conditions in the simulation model to obtain multiple simulation results, which include lightning overvoltage, overvoltage of disconnected unloaded lines, overvoltage of closed unloaded lines, and overvoltage of disconnected unloaded transformers; and use the transient overvoltage data from the multiple simulation results as simulation data.

[0012] According to another aspect of the present invention, a transient overvoltage identification device is also provided, comprising: an acquisition module for acquiring a real-time transient overvoltage signal in a power system; a first extraction module for extracting a first feature from the real-time transient overvoltage signal based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; a second extraction module for extracting a second feature from the real-time transient overvoltage signal based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; a fusion module for fusing the first feature and the second feature to obtain a fused feature; and a determination module for determining the identification result of the real-time transient overvoltage signal based on the fused feature, wherein the identification result includes the transient overvoltage type.

[0013] According to another aspect of the present invention, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored program, wherein, when the program is running, it controls the device where the non-volatile storage medium is located to execute any of the transient overvoltage identification methods described above.

[0014] According to another aspect of the present invention, a computer device is also provided, the computer device including a processor, the processor being configured to run a program, wherein the program executes any of the transient overvoltage identification methods described above during runtime.

[0015] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements any of the transient overvoltage identification methods described above.

[0016] In this embodiment of the invention, a transient overvoltage identification method is employed. This method acquires real-time transient overvoltage signals from the power system. Based on a preset first extraction model, a first feature is extracted from the real-time transient overvoltage signal. The first extraction model is a convolutional neural network model, and the first feature is a time-frequency feature. Based on a preset second extraction model, a second feature is extracted from the real-time transient overvoltage signal. The second extraction model is a deep long short-term memory network model, and the second feature is a time feature. The first feature and the second feature are fused to obtain a fused feature. Based on the fused feature, the identification result of the real-time transient overvoltage signal is determined. The identification result includes the transient overvoltage type, thereby improving the accuracy and anti-interference capability of transient overvoltage identification. This achieves the technical effect of real-time and accurate differentiation of various transient overvoltages, improving the real-time monitoring capability of the smart grid, and solving the technical problems of insufficient feature extraction comprehensiveness, model robustness, and real-time performance in traditional transient overvoltage identification technologies. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0018] Figure 1 A hardware block diagram of a computer terminal for implementing a transient overvoltage identification method is shown.

[0019] Figure 2 This is a flowchart illustrating the transient overvoltage identification method provided according to an embodiment of the present invention;

[0020] Figure 3 This is a flowchart of the feature extraction process of the first extraction model provided by an optional embodiment of the present invention;

[0021] Figure 4 This is a flowchart of the feature extraction process of the second extraction model provided by an optional embodiment of the present invention;

[0022] Figure 5 This is a flowchart of transient overvoltage mode recognition provided by an optional embodiment of the present invention;

[0023] Figure 6 This is a structural block diagram of a transient overvoltage identification device provided according to an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] According to an embodiment of the present invention, a transient overvoltage identification method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0027] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a transient overvoltage identification method is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1The different configurations shown.

[0028] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0029] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the transient overvoltage identification method in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the transient overvoltage identification method of the aforementioned application. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0030] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0031] Figure 2 This is a flowchart illustrating the transient overvoltage identification method provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the method includes the following steps:

[0032] Step S201: Obtain the real-time transient overvoltage signal in the power system.

[0033] In this step, sensors (such as voltage transformers and capacitive voltage dividers) installed at key nodes of the power system can be used to collect voltage signals and capture instantaneous voltage anomalies caused by lightning strikes, switching operations, or short-circuit faults during power system operation in real time. These devices can continuously monitor the voltage status of the power system at a high sampling rate. Once an abnormal change in the voltage signal is detected—that is, a voltage rise exceeding a preset threshold or exhibiting non-stationary characteristics—the transient overvoltage signal acquisition process is triggered. The acquisition process includes continuous recording of the signal, with the recorded data covering the entire transient event, from the initial rise of the overvoltage to its decay to a stable state. To ensure signal quality, the acquired signals need to undergo preprocessing, including noise reduction, signal enhancement, and normalization, to reduce the impact of environmental interference and equipment errors on subsequent feature extraction.

[0034] Step S202: Based on the preset first extraction model, extract the first feature from the real-time transient overvoltage signal, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature.

[0035] In this step, a first extraction model, namely a Convolutional Neural Network (CNN) model, is used. This model includes convolutional layers, pooling layers, and fully connected layers. The CNN model is used to extract the first feature from the acquired real-time transient overvoltage signal; this feature is a time-frequency feature. Specifically, firstly, the real-time transient overvoltage signal can be subjected to a Short-Time Fourier Transform (STFT) to generate a time-frequency map. The CNN model, through a multi-scale convolutional kernel group, can extract local features from the signal's time-frequency map, capturing key frequency domain details such as harmonic components and waveform distortion, generating multi-dimensional time-frequency features. Secondly, the pooling operation of the CNN model further compresses the feature dimension, improving the stability of the features and the model's noise resistance. Finally, the multi-dimensional feature map is flattened into a one-dimensional vector in the fully connected layer, outputting the first feature vector.

[0036] Step S203: Based on the preset second extraction model, extract the second feature from the real-time transient overvoltage signal, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature.

[0037] In this step, based on the preset second extraction model, namely the Deep Long Short-Term Memory (DLSTM) network model, a second feature is extracted from the real-time transient overvoltage signal, which captures the temporal features of the dynamic characteristics of the transient overvoltage signal. The second extraction model, through stacked Long Short-Term Memory (LSTM) units, can refine and capture the dynamic temporal patterns of the voltage signal. Feature extraction first requires the real-time transient overvoltage signal as a sequence input to the Depth-LSTM. Through the effective operation of the gating units, the temporal changes of the signal are analyzed layer by layer, capturing long-term dependencies, and finally outputting a feature vector that reflects the dynamic characteristics of the transient overvoltage.

[0038] Specifically, the feature vectors output by different LSTM layers can be viewed using visualization tools. Heatmaps or line graphs can be used to analyze whether the model can capture the key temporal features of the transient overvoltage signal, namely the rising edge, duration, and decay trend. The rising edge represents the change of the feature vector during the voltage surge phase, used to determine whether the model can accurately identify the steepness and location of the rising edge; the duration represents the stability of the feature vector during the voltage hold phase, used to determine whether the model can accurately capture the duration of the voltage; and the decay trend represents the change trend of the feature vector during the voltage decay phase, used to determine whether the model can accurately track the voltage decay process. The performance of LSTM models of different depths in feature extraction is compared using quantitative indicators (such as the variance of the feature vector and its correlation with the original signal) and qualitative analysis (such as visualization results), evaluating the impact of different model structures on feature extraction performance. The Depth-LSTM model structure that best captures the key temporal features is selected as the final secondary extraction model.

[0039] Step S204: The first feature and the second feature are fused to obtain the fused feature.

[0040] In this step, the first feature extracted by the first extraction model, namely the time-frequency feature rich in local time-frequency patterns, is fused with the second feature extracted by the second extraction model, namely the time feature that captures dynamic time-series patterns, to obtain a fused feature. By concatenating the first feature after the second feature, the advantages of the two features are effectively integrated, forming a comprehensive, multi-dimensional fused feature that includes both the frequency domain details of the transient overvoltage signal and information on its dynamic evolution in the time domain.

[0041] Step S205: Based on the fusion features, determine the identification result of the real-time transient overvoltage signal, wherein the identification result includes the transient overvoltage type.

[0042] In this step, the identification of real-time transient overvoltage signals is achieved through a constructed Deep Convolutional Network (Deep CNN) model. This optional embodiment pre-constructs an initial classification model, trains it using a training dataset, and obtains a target classification model, which is a Deep CNN classification model. This model takes fused features as input and, through the computation and learning of multiple layers of neurons, ultimately generates a probability distribution of transient overvoltage types at the output layer. By comparing the probability distributions, the category with the highest probability is selected as the identification result, thereby accurately distinguishing different transient overvoltage modes such as lightning overvoltage and switching overvoltage.

[0043] Through the above steps, the accuracy and anti-interference capability of transient overvoltage identification are improved, thereby achieving the technical effect of real-time and accurate differentiation of various transient overvoltages and improving the real-time monitoring capability of smart grids. This solves the technical problems of traditional transient overvoltage identification technology in terms of feature extraction comprehensiveness, model robustness and real-time performance.

[0044] As an optional embodiment, based on a preset first extraction model, a first feature is extracted from the real-time transient overvoltage signal, including: performing a time-frequency transformation on the real-time transient overvoltage signal through a short-time Fourier transform to obtain a time-frequency map of the real-time transient overvoltage signal; extracting initial features from the time-frequency map based on a convolutional layer of the first extraction model; performing dimensionality reduction processing on the initial features based on a pooling layer of the first extraction model to obtain a multi-dimensional feature map; and flattening the multi-dimensional feature map into a one-dimensional vector based on a fully connected layer of the first extraction model as the first feature.

[0045] Optionally, Figure 3 This is a flowchart of the feature extraction process for the first extraction model provided by an optional embodiment of the present invention. Figure 3As shown, firstly, the real-time transient overvoltage signal is subjected to time-frequency transformation using Short-Time Fourier Transform (STFT) to generate a time-frequency graph of the signal. STFT can capture the frequency component changes of the signal within a short time window, enabling the first extraction model to identify the time-varying frequency components in the signal. Secondly, a CNN-based convolutional layer extracts initial features from the time-frequency graph. The convolutional layer uses convolutional kernels of different sizes to slide on the time-frequency graph, capturing local features of the signal, such as high-frequency details and low-frequency trends, thus forming a preliminary description of the time-frequency characteristics of the transient overvoltage signal. The initial features are then dimensionality-reduced using a pooling layer to obtain a multi-dimensional feature map. Pooling helps to reduce the dimensionality of the feature vector while retaining key information, ensuring that the extracted features are both stable and have strong anti-interference capabilities. Finally, a CNN-based fully connected layer flattens the multi-dimensional feature map into a one-dimensional vector, outputting a time-frequency feature vector c as the first feature of the real-time transient overvoltage signal. In addition, visualization tools can be used to view the feature maps output by the convolutional layers, analyze the feature extraction effect, and optimize the model's feature extraction performance by trying different convolutional kernel sizes, pooling methods, and the number of filters.

[0046] As an optional embodiment, a second feature is extracted from a real-time transient overvoltage signal based on a preset second extraction model, including: converting the real-time transient overvoltage signal into normalized time-series data, wherein the time-series data includes transient overvoltage data within multiple time intervals; within any time interval, calculating the forget gate activation value, input gate activation value, and new memory unit candidate value of the transient overvoltage data based on the second extraction model; updating the memory units of the second extraction model and calculating the output gate activation value based on the forget gate activation value, input gate activation value, and new memory unit candidate value; updating the hidden layer state of the second extraction model based on the output gate activation value; repeating the above steps until all time intervals are traversed to obtain the final hidden layer state, which is used as the second feature.

[0047] Optionally, the time series of the input data is first defined as follows: ,in Let be the length of the time series. The normalization formula is as follows:

[0048]

[0049] in, This represents the normalized data; Representing time series The minimum value; Representing time series The maximum value. The normalized time series is It can reflect the time-domain characteristics of transient overvoltage signals, such as the rising edge, duration, and decay trend of the overvoltage.

[0050] Figure 4 This is a flowchart of the feature extraction process for the second extraction model provided by an optional embodiment of the present invention. Figure 4 As shown, a Depth-LSTM model is constructed, assuming a two-layer LSTM structure. The number of units in each LSTM layer is set to 100. First, the normalized time series data of transient overvoltage is input. Then, initialize the LSTM model parameters (weight matrix and bias terms). Next, for each time interval... Calculate the activation value of the forget gate , It can determine the cell state at the previous moment. How much information should be retained? The input gate activation value for each time interval also needs to be calculated. and new memory cell candidate values This updates the cell state. Furthermore, based on the calculated forget gate activation value, input gate activation value, and new memory unit candidate value, the memory units of the second retrieval model are updated, and the output gate activation value is calculated. The output gate activation value determines the current cell state. How much information can be output to the hidden state? The hidden state is then updated, and the updated hidden state is used for the calculation of the next time interval. Finally, the above steps are repeated until the data for all time intervals has been traversed, resulting in the final hidden layer state. The second feature provides a deep temporal dimension representation for subsequent feature fusion and classification.

[0051] As an optional embodiment, the first feature and the second feature are fused to obtain a fused feature, including: concatenating the second feature after the first feature to obtain the fused feature.

[0052] Optionally, the time-frequency feature map extracted by the first extraction model reflects the local features of the signal in the time-frequency domain, such as energy distribution and frequency change rate. The time feature vector extracted by the second extraction model can capture the time-domain characteristics of the transient overvoltage signal, such as the rise time, duration, and decay trend of the overvoltage. Feature fusion is performed by concatenation, directly concatenating the two feature vectors into a fused feature vector. The second feature vector... The first feature vector c is directly concatenated with the first feature vector c to form a fused feature vector F, that is... The simple concatenation operation described above preserves complete information from both features, enabling subsequent classification models to learn and classify using both temporal and time-frequency features simultaneously.

[0053] As an optional embodiment, determining the identification result of the real-time transient overvoltage signal based on the fused features includes: inputting the fused features into a preset target classification model to obtain the identification result, wherein the target classification model is trained on an initial classification model using a training dataset, the initial classification model being a deep convolutional neural network model, and the training dataset including simulation data and historical data.

[0054] Optionally, the fused features are provided as input to a pre-defined target classification model, which is trained from the initial classification model using a training dataset. The training dataset includes simulation data and historical data. Specifically, the historical data is based on past power grid transient overvoltage data, which may originate from monitoring equipment records, fault reports, and other relevant literature during power grid operation. This data covers transient overvoltage situations occurring in different regions and types of power grids under various operating conditions, including key information such as overvoltage amplitude, waveform, and duration. The historical data is manually classified and identified to establish a power grid transient overvoltage database based on historical data. Simulation data is obtained by constructing simulation models.

[0055] The initial classification model is a deep convolutional neural network (Deep CNN) model, which includes multiple convolutional layers, pooling layers, and fully connected layers. Its construction process is as follows:

[0056] The fully connected layer of the initial classification model will fuse the feature vectors. Mapped to category labels:

[0057]

[0058] in, This represents the output class probability distribution vector; This represents the weight matrix of the fully connected layer; This represents the bias term of the fully connected layer; This represents the fused feature vector; This represents the Softmax function, which is used to convert the output into a probability distribution.

[0059] The Softmax function converts the output of a fully connected layer into a probability distribution, and its calculation formula is as follows:

[0060]

[0061] in, Indicates that the input feature belongs to the first... The probability of a class; Indicates the number of categories; This represents the weights and biases of the i-th neuron in the fully connected layer. The cross-entropy loss function is used to measure the difference between the model's predictions and the true labels; the calculation formula is as follows:

[0062]

[0063] in, Indicates the first in the real label The probability of a class (usually one-hot encoded). The model predicts the first The probability of a class.

[0064] The initial classification model is trained on the training dataset, and the Adam optimization algorithm is used to minimize the loss function, updating the model's weights and biases. Specifically, the Adam algorithm calculates the loss on the training data in terms of model predictions and uses backpropagation to compute the gradient of the loss with respect to the model parameters, including the convolutional kernel weights in the CNN, the parameters in the pooling layers, and the weight matrices and biases in the fully connected layers. Subsequently, the Adam algorithm automatically adjusts the learning rate based on these gradients and their historical information to efficiently update the model parameters, thereby gradually reducing the value of the loss function. This process is repeated in each training iteration until the model's prediction performance on the training dataset reaches its optimum.

[0065] Finally, the target classification model is obtained by training the initial classification model. The target classification model fuses the input feature vector. The system categorizes and uses the probability distribution results as the identification results for different types of real-time transient overvoltage signals, and then visualizes them in two-dimensional or three-dimensional space. For example, a scatter plot can be used to show the distribution of overvoltage samples of different types, with different colors or shapes representing different overvoltage types, which can intuitively show the most likely overvoltage type of the input signal.

[0066] As an optional embodiment, the steps for obtaining simulation data are as follows: Construct a simulation model of the power system, wherein the simulation model includes transmission lines, transformers, and circuit breakers; Set multiple operating conditions in the simulation model to obtain multiple simulation results, wherein the multiple operating conditions include lightning overvoltage, overvoltage of disconnected unloaded lines, overvoltage of closed unloaded lines, and overvoltage of disconnected unloaded transformers; Use the transient overvoltage data in the multiple simulation results as simulation data.

[0067] Optionally, a power system simulation model can be constructed using Matlab / Simulink software. First, for the transmission line section, a distributed parameter model is built, with detailed settings for parameters such as the length of overhead and cable lines and their corresponding voltage levels. Second, a transformer module suitable for transient overvoltage characteristic analysis is selected, and parameters such as the transformer's rated capacity and turns ratio are accurately set. Third, when configuring the circuit breaker module, its actual operating behavior is fully simulated, with parameters such as operating time and opening / closing curves carefully set. Finally, the simulation model is run, and the results are used as simulation data in the training dataset. During the simulation, various typical operating conditions are reasonably set according to the different causes of transient overvoltages, including lightning overvoltage, overvoltage during disconnection of unloaded lines, overvoltage during closing of unloaded lines, and overvoltage during disconnection of unloaded transformers, to ensure the comprehensiveness and accuracy of the simulation results.

[0068] Specifically, a lightning overvoltage simulation model is constructed. Lightning overvoltage is usually caused by lightning impulse waves. The "Lightning Impulse Current Source" module in Matlab / Simulink is used, and its amplitude is set to... Wave head time is The wave tail time is Its waveform can be represented by a double exponential function:

[0069]

[0070] in, This represents the voltage value that changes with time during lightning overvoltage. This indicates the maximum voltage amplitude (peak value) of a lightning surge. The wavefront time constant determines the steepness of the lightning shock wave's ascent. This represents the wave tail time constant, which determines the attenuation rate of the lightning shock wave; Indicates time.

[0071] Construct an overvoltage model for interrupting an unloaded line. First, build a model that includes a circuit breaker, an unloaded line, and a voltage source. Set the amplitude of the voltage source in the operational overvoltage model to be [value missing]. At a frequency of 50Hz or 60Hz, record the voltage waveform across the line at the instant the circuit breaker opens. The relevant formulas are as follows:

[0072]

[0073] in, This represents the voltage value that changes over time when the unloaded line is interrupted. This represents the maximum amplitude under the operational overvoltage model; This represents the attenuation coefficient, which is related to the resistance and inductance of the line. This represents the oscillation angular frequency, which is related to the inductance and capacitance of the circuit. Indicates the initial phase angle; Indicates time.

[0074] Construct an overvoltage model for a closed-circuit unloaded line. First, build a model that includes a circuit breaker, an unloaded line, and a voltage source, and set the system rated voltage under the operational overvoltage model. At a frequency of 50Hz or 60Hz, record the voltage waveform across the line at the instant the circuit breaker opens. The relevant formulas are as follows:

[0075]

[0076] in, This represents the voltage value that changes over time when the line is closed and unloaded. Indicates the system's rated voltage; Indicates the attenuation coefficient; Indicates the angular frequency of oscillation; Indicates the initial phase angle; Indicates time.

[0077] Construct an overvoltage model for disconnecting an unloaded transformer. First, build a model including a circuit breaker, an unloaded line, and a voltage source. Set the amplitude of the voltage source in the operational overvoltage model to be [value missing]. At a frequency of 50Hz or 60Hz, record the voltage waveform across the transformer windings at the instant the circuit breaker is opened.

[0078]

[0079] in, This indicates the voltage value that changes over time when the overvoltage of the unloaded transformer is disconnected. This represents the maximum amplitude under the operational overvoltage model; This represents the attenuation coefficient, which is related to the winding resistance and inductance of the transformer. Indicates the angular frequency of oscillation; Indicates time.

[0080] As an optional embodiment, Figure 5 This is a flowchart of transient overvoltage mode recognition provided according to an optional embodiment of the present invention. Figure 5As shown, transient overvoltage pattern recognition requires data preparation first. This step involves collecting and preprocessing transient overvoltage signals to ensure the data quality and applicability for subsequent analysis. Next, a dual-track feature extraction process is performed. On one hand, the real-time signal undergoes a Short-Time Fourier Transform (STFT) to transform it into a time-frequency graph, which is then fed into the convolutional and pooling layers of a Convolutional Neural Network (CNN) to extract features covering frequency and temporal local details. On the other hand, the normalized time-series signal is input into a Deep Long Short-Term Memory (DLSTM) network, which uses its gating mechanism to capture the dynamic evolution of the signal, producing key feature vectors reflecting time-series dependencies. Subsequently, the feature vectors extracted from the two paths are concatenated to obtain a fused feature vector. This fused feature vector becomes the input to a Deep Convolutional Neural Network (Deep CNN) classification model. Through multiple rounds of training and parameter optimization, this model learns how to map the fused features to probability distributions for different transient overvoltage types. Finally, the trained classification model can receive the real-time fused features and output accurate transient overvoltage pattern recognition results.

[0081] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0082] Through the above description of the embodiments, those skilled in the art can clearly understand that the transient overvoltage identification method according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0083] According to embodiments of the present invention, an apparatus for implementing the above-described transient overvoltage identification method is also provided. Figure 6 This is a structural block diagram of a transient overvoltage identification device provided according to an embodiment of the present invention, such as... Figure 6 As shown, the device includes: an acquisition module 61, a first extraction module 62, a second extraction module 63, a fusion module 64, and a determination module 65. The device will be described below.

[0084] The acquisition module 61 is used to acquire real-time transient overvoltage signals in the power system.

[0085] The first extraction module 62 is connected to the acquisition module 61 and is used to extract a first feature from the real-time transient overvoltage signal based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a first feature.

[0086] The second extraction module 63 is connected to the first extraction module 62 and is used to extract a second feature from the real-time transient overvoltage signal based on a preset second extraction model. The second extraction model is a deep long short-term memory network model and the second feature is a time feature.

[0087] The fusion module 64 is connected to the second extraction module 63 and is used to fuse the first feature and the second feature to obtain the fused feature.

[0088] The determination module 65, connected to the fusion module 64, is used to determine the identification result of the real-time transient overvoltage signal based on the fusion features, wherein the identification result includes the transient overvoltage type.

[0089] It should be noted that the acquisition module 61, the first extraction module 62, the second extraction module 63, the fusion module 64, and the determination module 65 mentioned above correspond to steps S201 to S205 in the embodiments. Multiple modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of the device, can run on the computer terminal 10 provided in the embodiments.

[0090] Embodiments of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one of a plurality of network devices in a computer network. The computer device includes a memory and a processor.

[0091] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the transient overvoltage identification method and device in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned transient overvoltage identification method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0092] The processor can access information and application programs stored in the memory via a transmission device to perform the following steps: acquiring real-time transient overvoltage signals in the power system; extracting a first feature from the real-time transient overvoltage signals based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; extracting a second feature from the real-time transient overvoltage signals based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; fusing the first feature and the second feature to obtain a fused feature; and determining the identification result of the real-time transient overvoltage signal based on the fused feature, wherein the identification result includes the transient overvoltage type.

[0093] Optionally, the processor may also execute program code for the following steps: extracting a first feature from a real-time transient overvoltage signal based on a preset first extraction model, including: performing a time-frequency transformation on the real-time transient overvoltage signal using a short-time Fourier transform to obtain a time-frequency map of the real-time transient overvoltage signal; extracting initial features from the time-frequency map using a convolutional layer based on the first extraction model; performing dimensionality reduction processing on the initial features using a pooling layer based on the first extraction model to obtain a multi-dimensional feature map; and flattening the multi-dimensional feature map into a one-dimensional vector using a fully connected layer based on the first extraction model as the first feature.

[0094] Optionally, the processor may also execute program code with the following steps: extracting a second feature from a real-time transient overvoltage signal based on a preset second extraction model, including: converting the real-time transient overvoltage signal into normalized time-series data, wherein the time-series data includes transient overvoltage data within multiple time intervals; within any time interval, calculating the forget gate activation value, input gate activation value, and new memory unit candidate value of the transient overvoltage data based on the second extraction model; updating the memory units of the second extraction model and calculating the output gate activation value based on the forget gate activation value, input gate activation value, and new memory unit candidate value; updating the hidden layer state of the second extraction model based on the output gate activation value; repeating the above steps until all time intervals are traversed to obtain the final hidden layer state as the second feature.

[0095] Optionally, the processor may also execute program code that performs the following steps: fusing the first feature and the second feature to obtain a fused feature, including: concatenating the second feature after the first feature to obtain the fused feature.

[0096] Optionally, the processor may also execute program code for the following steps: determining the identification result of the real-time transient overvoltage signal based on the fused features, including: inputting the fused features into a preset target classification model to obtain the identification result, wherein the target classification model is trained on an initial classification model using a training dataset, the initial classification model being a deep convolutional neural network model, and the training dataset including simulation data and historical data.

[0097] Optionally, the processor may also execute program code with the following steps: The steps for obtaining simulation data are as follows: Construct a simulation model of the power system, wherein the simulation model includes transmission lines, transformers, and circuit breakers; Set multiple operating conditions in the simulation model to obtain multiple simulation results, wherein the multiple operating conditions include lightning overvoltage, overvoltage of disconnected unloaded lines, overvoltage of closed unloaded lines, and overvoltage of disconnected unloaded transformers; Use the transient overvoltage data in the multiple simulation results as simulation data.

[0098] This invention provides a method for identifying transient overvoltages. The method involves acquiring real-time transient overvoltage signals from a power system; extracting a first feature from the real-time transient overvoltage signals based on a preset first extraction model (a convolutional neural network model, where the first feature is a time-frequency feature); extracting a second feature from the real-time transient overvoltage signals based on a preset second extraction model (a deep long short-term memory network model, where the second feature is a time feature); fusing the first and second features to obtain a fused feature; and determining the identification result of the real-time transient overvoltage signal based on the fused feature. The identification result includes the transient overvoltage type, thereby improving the accuracy and anti-interference capability of transient overvoltage identification and solving the technical problems of traditional transient overvoltage identification techniques in terms of feature extraction comprehensiveness, model robustness, and real-time performance.

[0099] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a non-volatile storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0100] Embodiments of the present invention also provide a non-volatile storage medium. Optionally, in this embodiment, the aforementioned non-volatile storage medium can be used to store the program code executed by the transient overvoltage identification method provided in the above embodiments.

[0101] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0102] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: acquiring real-time transient overvoltage signals in the power system; extracting a first feature from the real-time transient overvoltage signals based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; extracting a second feature from the real-time transient overvoltage signals based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; fusing the first feature and the second feature to obtain a fused feature; and determining the identification result of the real-time transient overvoltage signal based on the fused feature, wherein the identification result includes the transient overvoltage type.

[0103] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: extracting a first feature from a real-time transient overvoltage signal based on a preset first extraction model, including: performing a time-frequency transformation on the real-time transient overvoltage signal through a short-time Fourier transform to obtain a time-frequency map of the real-time transient overvoltage signal; extracting initial features from the time-frequency map based on a convolutional layer of the first extraction model; performing dimensionality reduction processing on the initial features based on a pooling layer of the first extraction model to obtain a multi-dimensional feature map; and flattening the multi-dimensional feature map into a one-dimensional vector based on a fully connected layer of the first extraction model as the first feature.

[0104] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: extracting a second feature from a real-time transient overvoltage signal based on a preset second extraction model, including: converting the real-time transient overvoltage signal into normalized time-series data, wherein the time-series data includes transient overvoltage data within multiple time intervals; within any time interval, calculating the forget gate activation value, input gate activation value, and new memory unit candidate value of the transient overvoltage data based on the second extraction model; updating the memory unit of the second extraction model and calculating the output gate activation value based on the forget gate activation value, input gate activation value, and new memory unit candidate value; updating the hidden layer state of the second extraction model based on the output gate activation value; repeating the above steps until all time intervals are traversed to obtain the final hidden layer state as the second feature.

[0105] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: fusing the first feature and the second feature to obtain a fused feature, including: concatenating the second feature after the first feature to obtain the fused feature.

[0106] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining the identification result of the real-time transient overvoltage signal based on the fused features, including: inputting the fused features into a preset target classification model to obtain the identification result, wherein the target classification model is trained on an initial classification model using a training dataset, the initial classification model being a deep convolutional neural network model, and the training dataset including simulation data and historical data.

[0107] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: The steps for obtaining simulation data are as follows: Construct a simulation model of the power system, wherein the simulation model includes transmission lines, transformers, and circuit breakers; Set multiple operating conditions in the simulation model respectively to obtain multiple simulation results, wherein the multiple operating conditions include lightning overvoltage, overvoltage of disconnected unloaded lines, overvoltage of closed unloaded lines, and overvoltage of disconnected unloaded transformers; Use the transient overvoltage data in the multiple simulation results as simulation data.

[0108] Embodiments of the present invention also provide a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it can: acquire a real-time transient overvoltage signal in a power system; extract a first feature from the real-time transient overvoltage signal based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; extract a second feature from the real-time transient overvoltage signal based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; fuse the first feature and the second feature to obtain a fused feature; and determine the identification result of the real-time transient overvoltage signal based on the fused feature, wherein the identification result includes the transient overvoltage type.

[0109] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0110] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0112] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0113] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0115] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying transient overvoltages, characterized in that, include: Acquire real-time transient overvoltage signals in the power system; Based on a preset first extraction model, a first feature is extracted from the real-time transient overvoltage signal, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; Based on a preset second extraction model, a second feature is extracted from the real-time transient overvoltage signal, wherein the second extraction model is a deep long short-term memory network model, and the second feature is a time feature; The first feature and the second feature are fused to obtain a fused feature; Based on the fusion features, the identification result of the real-time transient overvoltage signal is determined, wherein the identification result includes the transient overvoltage type.

2. The method according to claim 1, characterized in that, The extraction of a first feature from the real-time transient overvoltage signal based on a preset first extraction model includes: The real-time transient overvoltage signal is transformed by short-time Fourier transform to obtain the time-frequency diagram of the real-time transient overvoltage signal. Based on the convolutional layer of the first extraction model, initial features are extracted from the time-frequency map; Based on the pooling layer of the first extraction model, the initial features are dimensionality reduced to obtain a multidimensional feature map; Based on the fully connected layer of the first extraction model, the multidimensional feature map is flattened into a one-dimensional vector, which is used as the first feature.

3. The method according to claim 1, characterized in that, The second feature extracted from the real-time transient overvoltage signal based on the preset second extraction model includes: The real-time transient overvoltage signal is converted into normalized time series data, wherein the time series data includes transient overvoltage data within multiple time intervals; Within any of the time intervals, based on the second extraction model, the forget gate activation value, input gate activation value, and new memory cell candidate value of the transient overvoltage data are calculated. Based on the forget gate activation value, the input gate activation value, and the new memory unit candidate value, update the memory units of the second extraction model and calculate the output gate activation value; Based on the output gate activation value, update the hidden layer state of the second extraction model; Repeat the above steps until all the time intervals have been traversed to obtain the final hidden layer state, which is used as the second feature.

4. The method according to claim 1, characterized in that, The step of fusing the first feature and the second feature to obtain the fused feature includes: The second feature is appended to the first feature to obtain the fused feature.

5. The method according to claim 1, characterized in that, The determination of the identification result of the real-time transient overvoltage signal based on the fused features includes: The fused features are input into a preset target classification model to obtain the recognition result. The target classification model is trained on an initial classification model using a training dataset. The initial classification model is a deep convolutional neural network model, and the training dataset includes simulation data and historical data.

6. The method according to claim 5, characterized in that, The steps for obtaining the simulation data are as follows: Construct a simulation model of the power system, wherein the simulation model includes transmission lines, transformers, and circuit breakers; Multiple operating conditions were set in the simulation model to obtain multiple simulation results. The multiple operating conditions include lightning overvoltage, disconnection of unloaded line overvoltage, closing of unloaded line overvoltage, and disconnection of unloaded transformer overvoltage. The transient overvoltage data from the multiple simulation results are used as the simulation data.

7. A transient overvoltage identification device, characterized in that, include: The acquisition module is used to acquire real-time transient overvoltage signals in the power system; The first extraction module is used to extract a first feature from the real-time transient overvoltage signal based on a preset first extraction model, wherein the first extraction model is a convolutional neural network model and the first feature is a time-frequency feature; The second extraction module is used to extract a second feature from the real-time transient overvoltage signal based on a preset second extraction model, wherein the second extraction model is a deep long short-term memory network model and the second feature is a time feature; The fusion module is used to fuse the first feature and the second feature to obtain a fused feature; The determination module is used to determine the identification result of the real-time transient overvoltage signal based on the fused features, wherein the identification result includes the transient overvoltage type.

8. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the non-volatile storage medium to perform the transient overvoltage identification method according to any one of claims 1 to 6.

9. A computer device, characterized in that, include: Memory and processor The memory stores computer programs; The processor is configured to execute a computer program stored in the memory, wherein when the computer program is executed, the processor performs the transient overvoltage identification method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the transient overvoltage identification method according to any one of claims 1 to 6.

Images