Feature extraction method of synchronous phasor data, model training method, medium and equipment

Abstract

The invention discloses a synchronous phasor data feature extraction method, a model training method, a medium and equipment, multi-dimensional features of synchronous phasor data are adaptively captured by using a dynamic convolution kernel, an attention mechanism is introduced to highlight key features, global and local feature information is fused, optimization is carried out in combination with a dynamic weighted feature loss function, and the feature extraction efficiency of the synchronous phasor data is improved. And finally, accurate feature expression is generated. The method has the core advantages that the characteristic difference between different channels can be effectively distinguished, and key fault information is highlighted, so that the accuracy and response speed of fault detection are improved, the limitation that a traditional method is difficult to process multi-dimensional complex characteristics is overcome, and an innovative solution is provided for efficient fault diagnosis of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and fault diagnosis of electric power systems, and in particular to a feature extraction method, a model training method, a medium and a device for synchronous phasor data. Background Art

[0002] In the power system, synchronized phasor data is the data of electrical quantities such as voltage and current in the power system collected by the synchronized phasor measurement unit (PMU). However, when processing multi-dimensional synchronized phasor data, traditional fault feature extraction methods often find it difficult to effectively capture the feature differences between different channels and key fault information, thus affecting the accuracy and response speed of subsequent fault detection. Summary of the invention

[0003] Based on this, it is necessary to provide methods, model training methods, media and equipment to solve the problem that traditional fault feature extraction methods are often difficult to effectively capture the feature differences and key fault information between different channels when processing multi-dimensional synchronous phasor data, thereby affecting the accuracy and response speed of subsequent fault detection.

[0004] A training method for an adaptive feature channel convolutional network, the adaptive feature channel convolutional network comprising a dynamic convolutional layer, a local feature extraction layer, a global feature fusion layer and a fully connected layer connected in sequence, the dynamic convolutional layer being provided with a plurality of convolutional kernels of increasing sizes in sequence, the method comprising:

[0005] Acquire a sample set; wherein the sample set includes a plurality of samples, each sample including a plurality of sample synchronized phasor data within a preset time period;

[0006] The samples are sequentially input into the adaptive feature channel convolution network, the fluctuation amplitude of each sample synchronized phasor data in the samples is calculated through the dynamic convolution layer, and a corresponding convolution kernel is selected according to the fluctuation amplitude for convolution processing to obtain a plurality of convolution feature maps; wherein the fluctuation amplitude is positively correlated with the size of the corresponding selected convolution kernel;

[0007] Extracting key features from the multiple feature maps through the attention mechanism of the local feature extraction layer to obtain weighted local features;

[0008] The global feature fusion layer captures the global feature of the weighted local feature, and fuses the global feature with the weighted local feature to obtain a fused feature;

[0009] The fused features are converted into feature vectors of synchronized phasor data through the fully connected layer, the feature vectors of synchronized phasor data are input into a preset dynamic weighted feature loss function to calculate a loss value, and the adaptive feature channel convolutional network is iteratively adjusted according to the loss value until the adaptive feature channel convolutional network converges.

[0010] In one embodiment, the step of calculating the fluctuation amplitude of the synchronized phasor data of each sample in the sample by the dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude includes:

[0011] Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as:

[0012]

[0013] In the above formula, represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; X i (t) represents the i-th sample synchronized phasor data at time t; represents the mean value of the i-th sample synchronized phasor data within a preset time period;

[0014] The mapping size of each sample synchronized phasor data in the sample is calculated according to the fluctuation amplitude, which is expressed as:

[0015]

[0016] In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max represents the maximum size of all convolution kernels in the dynamic convolution layer, σ 2 max represents the maximum fluctuation amplitude of all sample synchronized phasor data, σ 2 min It represents the minimum value of the fluctuation amplitude of all sample synchronized phasor data;

[0017] Based on the mapping size, a convolution kernel corresponding to each sample synchronized phasor data is selected by an integer allocation method to perform convolution processing.

[0018] In one embodiment, the step of calculating the fluctuation amplitude of the synchronized phasor data of each sample in the sample by the dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude includes:

[0019] Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as:

[0020] A i =max(X i (t))-min(X i (t)), t∈[1,T]

[0021] In the above formula, A i represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; max(X i (t)) represents the maximum value of the ith sample synchronized phasor data within the preset time period; min(X i (t) represents the minimum value of the ith sample synchronized phasor data within a preset time period;

[0022] The mapping size of each sample synchronized phasor data in the sample is calculated according to the fluctuation amplitude, which is expressed as:

[0023]

[0024] In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max represents the maximum size of all convolution kernels in the dynamic convolution layer, A min Indicates the minimum fluctuation amplitude of all sample synchronized phasor data, A max It represents the maximum value of the fluctuation amplitude of all sample synchronized phasor data;

[0025] Based on the mapping size, a convolution kernel corresponding to each sample synchronized phasor data is selected by an integer allocation method to perform convolution processing.

[0026] In one embodiment, the step of selecting a convolution kernel corresponding to each sample synchronized phasor data by an integer allocation method based on the mapping size for convolution processing includes:

[0027] The mapping size is k i The synchronized phasor data is distributed to the floor(k i ) is used for convolution processing;

[0028] If there are at least two synchronized phasor data assigned to the same convolution kernel, the synchronized phasor data whose size difference is not the smallest is assigned to the idle convolution kernel with the smallest corresponding size difference for convolution processing; wherein the size difference is the absolute value of the difference between the mapping size and the convolution kernel size.

[0029] In one embodiment, the key features of the multiple feature maps are extracted through the attention mechanism of the local feature extraction layer to obtain weighted local features, which are expressed as:

[0030]

[0031]

[0032] In the above formula, Z Pw (t) represents the weighted local feature, I represents the total number of sample synchronized phasor data in each sample, represents the fluctuation amplitude of the ith sample synchronized phasor data, Z i Characteristic diagram representing the ith sample synchronized phasor data.

[0033] In one embodiment, the global feature of the weighted local feature is captured by the global feature fusion layer, and the global feature is fused with the weighted local feature to obtain a fused feature, which is expressed as:

[0034] Z Pf (t) = Z Pw (t)+W global ·Z Pg

[0035] Z Pg =GlobalAvgPool(Z Pw )

[0036] In the above formula, Z Pf (t) represents the fusion feature, W global Represents the weight matrix of the global feature, Z Pg Represents the global feature, and GlobalAvgPool() represents the global average pooling operation.

[0037] In one embodiment, the dynamic weighted feature loss function is expressed as:

[0038]

[0039] In the above formula, L DWFL Represents the loss value; F′ Pn represents the expected eigenvector of the synchronized phasor data in the nth eigenchannel; F Pn represents the feature vector extracted from the synchronized phasor data in the nth feature channel; w n represents the weight of the nth feature channel; λ represents the regularization coefficient; W conv Represents the parameter matrix of the convolution kernel.

[0040] A method for extracting features from synchronized phasor data, the method comprising:

[0041] Acquire multiple measured synchronized phasor data within a preset time period;

[0042] The multiple measured synchronized phasor data are input into an adaptive feature channel convolutional network to obtain feature vectors of the multiple measured synchronized phasor data; wherein the adaptive feature channel convolutional network is trained by the above-mentioned adaptive feature channel convolutional network training method.

[0043] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the training method of the above-mentioned adaptive feature channel convolutional network and the feature extraction method of synchronized phasor data.

[0044] A terminal device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above-mentioned training method of the adaptive feature channel convolutional network and the feature extraction method of synchronized phasor data.

[0045] The present invention provides a feature extraction method, model training method, medium and equipment for synchronized phasor data, which uses dynamic convolution kernels to adaptively capture the multi-dimensional features of synchronized phasor data, introduces an attention mechanism to highlight key features, and fuses global and local feature information, combined with a dynamic weighted feature loss function for optimization, and finally generates an accurate feature expression. Its core advantage is that it can effectively distinguish feature differences between different channels and highlight key fault information, thereby improving the accuracy and response speed of fault detection, overcoming the limitations of traditional methods that are difficult to handle multi-dimensional complex features, and providing an innovative solution for efficient fault diagnosis of power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.

[0047] in:

[0048] Figure 1 A flowchart of a training method for an adaptive feature channel convolutional network;

[0049] Figure 2 A flowchart for feature extraction of synchronized phasor data;

[0050] Figure 3 is a flow chart of a feature extraction method for synchronized phasor data;

[0051] Figure 4 This is a structural block diagram of the terminal equipment. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] like Figure 1 As shown, Figure 1 The figure is a flow chart of a training method of an adaptive feature channel convolutional network in one embodiment. The adaptive feature channel convolutional network includes a dynamic convolutional layer, a local feature extraction layer, a global feature fusion layer and a fully connected layer connected in sequence, wherein the dynamic convolutional layer is provided with a plurality of convolutional kernels of increasing sizes in sequence. The steps provided by the training method of the adaptive feature channel convolutional network in this embodiment include:

[0056] S101, obtaining a sample set.

[0057] The sample set includes multiple samples, each of which includes multiple sample synchronized phasor data within a preset time period. The sample set is the basis for model training and fault feature extraction, and provides input data to support algorithm analysis.

[0058] In this embodiment, Figure 2 As shown, the sample synchronized phasor data in each sample includes voltage, current, frequency, reactive power, and active power, which are recorded as:

[0059] X=[V(t), I(t), F(t), Q(t), P(t)], t=1, 2, ..., T

[0060] S102, input the samples into the adaptive feature channel convolution network in sequence, calculate the fluctuation amplitude of the synchronized phasor data of each sample in the samples through the dynamic convolution layer, and select the corresponding convolution kernel for convolution processing according to the fluctuation amplitude to obtain multiple convolution feature maps.

[0061] The fluctuation amplitude refers to the fluctuation intensity of the sample data, which can be measured by statistics (such as variance, standard deviation, difference between maximum and minimum values, etc.) In this embodiment, the fluctuation amplitude is set to be positively correlated with the size of the corresponding selected convolution kernel.

[0062] This is because for sample synchronized phasor data with large fluctuations, a larger convolution kernel is needed to capture the range of changes and fully extract global context features. For example, in synchronized phasors, drastic changes in voltage or current within a specific time period may characterize the overall pattern of fault occurrence. Therefore, after the dynamic convolution layer calculates the fluctuation amplitude of the data, it assigns a convolution kernel with a larger convolution kernel size for convolution processing. Correspondingly, for sample synchronized phasor data with small fluctuations, after the dynamic convolution layer calculates the fluctuation amplitude of the data, it assigns a convolution kernel with a smaller convolution kernel size for convolution processing. This allows the data to focus on extracting local details and avoid mixing subtle features with irrelevant context.

[0063] In this way, by calculating the fluctuation amplitude, selecting the convolution kernel and dynamically adjusting the convolution operation, not only the adaptability of the model to multi-scale features is enhanced, but also the accuracy and diversity of feature extraction are significantly improved.

[0064] S103, extract key features from multiple feature maps through the attention mechanism of the local feature extraction layer to obtain weighted local features.

[0065] In the local feature extraction layer, the attention mechanism is used to dynamically assign weights according to the importance of the features, enhance key features by weighted sum, and weaken redundant or noisy features to obtain weighted local features, providing a basis for subsequent global feature fusion.

[0066] Optional, such as Figure 2 As shown, in S103, the key features of multiple feature maps are extracted through the attention mechanism of the local feature extraction layer to obtain weighted local features, which are expressed as:

[0067]

[0068] In the above formula, Z Pw (t) represents the weighted local feature, I represents the total number of sample synchronized phasor data in each sample, represents the fluctuation amplitude of the ith sample synchronized phasor data, Z i Characteristic diagram representing the ith sample synchronized phasor data.

[0069] This step performs weighted processing on the feature map through the attention mechanism of the local feature extraction layer, which can effectively extract key features and highlight important information.

[0070] S104, capturing the global features of the weighted local features through a global feature fusion layer, and fusing the global features with the weighted local features to obtain fused features.

[0071] In this step, the global feature fusion layer is used to extract the global information of the input features (i.e., the overall representation of all features) and fuse them with the local features to form a more comprehensive and unified fusion feature, providing more accurate feature input for subsequent tasks (such as classification and detection).

[0072] Optional, such as Figure 2 As shown, in S104, the global feature fusion layer captures the global features of the weighted local features, and fuses the global features with the weighted local features to obtain the fused features, which are expressed as:

[0073] Z Pf (t) = Z Pw (t)+W global ·Z Pg

[0074] Z Pg =GlobalAvgPool(Z Pw )

[0075] In the above formula, Z Pf (t) represents the fusion feature, which contains the overall information and retains the key details. global Represents the weight matrix of global features, Z Pg Represents global features, GlobalAvgPool() represents the global average pooling operation, which compresses the global information of each channel into one value.

[0076] This step extracts global features through the global feature fusion layer and fuses them with local features, which can enhance the model's comprehensive understanding of the input data.

[0077] S105, converting the fused features into feature vectors of synchronized phasor data through a fully connected layer, inputting the feature vectors of synchronized phasor data into a preset dynamic weighted feature loss function to calculate a loss value, and iteratively adjusting the adaptive feature channel convolutional network through the loss value until the adaptive feature channel convolutional network converges.

[0078] In this step, the fully connected layer is used to achieve high-dimensional compression or mapping of features, thereby outputting a fixed-length feature vector. The dynamic weighted feature loss function is used to dynamically assign weights according to the importance of different feature vectors, thereby guiding the model to focus more on the learning of key features.

[0079] Optionally, the dynamic weighted feature loss function is expressed as:

[0080]

[0081] In the above formula, L DWFL represents the loss value; F′ Pn represents the expected eigenvector of the synchronized phasor data in the nth eigenchannel; F Pn represents the feature vector extracted from the synchronized phasor data in the nth feature channel; w n represents the weight of the nth feature channel; λ represents the regularization coefficient; W conv Represents the parameter matrix of the convolution kernel.

[0082] In the above dynamic weighted feature loss function, w is dynamically adjusted n It can enhance the focus on important features while reducing overfitting of minor features. conv L2 regularization can suppress overfitting and ensure the generalization ability of the model. λ controls the strength of regularization. A large value will affect the feature extraction ability, while a small value may lead to overfitting.

[0083] We calculate the loss value to perform gradient derivation on the network parameters (such as convolution kernel parameters, weights, etc.). We use optimization algorithms (such as gradient descent or Adam) to adjust the network parameters according to the gradient to reduce the loss value. By repeatedly performing the above steps and continuously optimizing the network parameters through multiple iterations, the loss function value eventually converges to the minimum value and the network reaches the optimal state.

[0084] The feature extraction method of synchronized phasor data above uses dynamic convolution kernels to adaptively capture the multi-dimensional features of synchronized phasor data, introduces an attention mechanism to highlight key features, and fuses global and local feature information, combined with a dynamic weighted feature loss function for optimization, and finally generates an accurate feature expression. Its core advantage is that it can effectively distinguish feature differences between different channels and highlight key fault information, thereby improving the accuracy and response speed of fault detection, overcoming the limitations of traditional methods that are difficult to handle multi-dimensional complex features, and providing an innovative solution for efficient fault diagnosis of power systems.

[0085] Optionally, in S102, calculating the fluctuation amplitude of each sample synchronized phasor data in the sample by a dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude specifically includes the following steps:

[0086] A1. Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as:

[0087]

[0088] In the above formula, represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; X i (t) represents the i-th sample synchronized phasor data at time t; represents the mean value of the i-th sample synchronized phasor data within a preset time period;

[0089] A2. Calculate the mapping size of each sample synchronized phasor data in the sample according to the fluctuation amplitude, expressed as:

[0090]

[0091] In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max represents the maximum size of all convolution kernels in the dynamic convolution layer, σ 2 max Represents the maximum fluctuation amplitude of all sample synchronized phasor data, σ 2 min represents the minimum value of the fluctuation amplitude of all sample synchronized phasor data; optionally, let k min Take 3, k max Take 9.

[0092] In this way, the fluctuation amplitude is mapped to the convolution kernel size in a specified range, ensuring that the convolution kernel can adapt to the degree of change of different sample characteristics.

[0093] A3. Based on the mapping size, the convolution kernel corresponding to each sample synchronized phasor data is selected by the integer distribution method for convolution processing.

[0094] The calculation and mapping mechanism of the fluctuation amplitude in this embodiment enables each sample to be assigned an adapted convolution kernel, thus avoiding the information loss caused by the fixed convolution kernel size in the traditional method.

[0095] Optionally, in S102, calculating the fluctuation amplitude of each sample synchronized phasor data in the sample by a dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude specifically includes the following steps:

[0096] B1. Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as:

[0097] A i=max(X i (t))-min(X i (t)), t∈[1,T]

[0098] In the above formula, A i represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; max(X i (t)) represents the maximum value of the ith sample synchronized phasor data within the preset time period; min(X i (t) represents the minimum value of the ith sample synchronized phasor data within the preset time period.

[0099] B2. Calculate the mapping size of each sample synchronized phasor data in the sample according to the fluctuation amplitude, expressed as:

[0100]

[0101] In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max A represents the maximum size of all convolution kernels in the dynamic convolution layer. min Indicates the minimum fluctuation amplitude of all sample synchronized phasor data, A max Indicates the maximum value of the fluctuation amplitude of all sample synchronized phasor data.

[0102] In this step, the fluctuation amplitude A is converted into i Mapped to the convolution kernel size range [k min , k max ], which ensures that the convolution kernel can adapt to the degree of change of different sample characteristics.

[0103] B3. Based on the mapping size, the convolution kernel corresponding to each sample synchronized phasor data is selected by the integer distribution method for convolution processing.

[0104] Similarly, the calculation and mapping mechanism of the fluctuation amplitude in this embodiment can also enable each sample to be assigned an adapted convolution kernel, thereby avoiding the information loss caused by the fixed convolution kernel size in the traditional method.

[0105] Optionally, selecting a convolution kernel corresponding to each sample synchronized phasor data by an integer allocation method based on the mapping size for convolution processing specifically includes the following steps:

[0106] C1. Set the mapping size to k i The synchronized phasor data is distributed to the floor(k i ) is used for convolution processing.

[0107] C2. If there are at least two synchronized phasor data being assigned to the same convolution kernel, the synchronized phasor data whose size difference is not the smallest is assigned to an idle convolution kernel whose corresponding size difference is the smallest for convolution processing.

[0108] The size difference is the absolute value of the difference between the mapping size and the convolution kernel size.

[0109] Exemplarily, assuming that the convolution kernel size range is [3, 9], there are the following mapping sizes of synchronized phasor data:

[0110] k 1 =4.8, k 2 =4.3, k 3 =6.7, k 4 =4.9

[0111] Initial allocation stage:

[0112] k 1 and k 2 Round down and assign it to the convolution kernel size of 4.

[0113] k 3 Round down and assign it to the convolution kernel size of 6.

[0114] k 4 Round down and assign it to the convolution kernel size of 4.

[0115] Repeat the adjustment phase:

[0116] Due to k 1 , k 2 and k 4 Both are assigned to kernel size 4, so adjustments need to be made based on the size difference.

[0117] Step 1: Adjust k based on size difference 1 , k 2 and k 4 :

[0118] k 1 Size difference: |4.8-4|=0.8

[0119] k 2 Size difference: |4.3-4|=0.3

[0120] k 4 Size difference: |4.9-4|=0.9

[0121] So keep k 2 In convolution kernel 4, continue to adjust other data.

[0122] Step 2: Adjust k based on size difference 1 and k 4To other idle convolution kernels:

[0123] k 1 Size difference: |4.8-5|=0.2

[0124] k 4 Size difference: |4.9-5|=0.1

[0125] So keep k 4 In convolution kernel 5, continue to adjust other data.

[0126] Step 3: Adjust k based on size difference 1 To other idle convolution kernels:

[0127] Convolution kernel sizes 3 and 7 are free, and are adjusted based on the size difference:

[0128] k 1 Size difference 1: |4.8-3|=1.8

[0129] k 1 Size difference 2: |4.8-7|=2.2

[0130] Therefore, keep k 1 The convolution kernel size is 3.

[0131] In this embodiment, the rationality of convolution kernel allocation is ensured by adjusting the repeatedly allocated data, avoiding feature conflict or loss when multiple synchronized phasor data share the same convolution kernel. After the optimization allocation process, the convolution kernel is closer to the ideal size, improving the feature extraction effect, especially when processing multi-dimensional synchronized phasor data, it is easier to capture key fault information.

[0132] like Figure 3 As shown, the present invention also proposes a feature extraction method for synchronized phasor data, and the method comprises the following steps:

[0133] S301, obtaining a plurality of measured synchronized phasor data within a preset time period.

[0134] That is, multiple synchronized phasor data are collected from a preset time period. These data include measured values ​​such as voltage, current, power, and frequency that reflect the operating status of the power system.

[0135] S302, inputting a plurality of measured synchronized phasor data into an adaptive feature channel convolutional network to obtain feature vectors of the plurality of measured synchronized phasor data.

[0136] The adaptive feature channel convolutional network is trained by the above method. The input data is processed by the adaptive network and the feature vectors representing the key characteristics of the synchronized phasor data are output. These feature vectors can be used for subsequent analysis, such as fault detection or classification tasks.

[0137] Figure 4 FIG. 2 shows an internal structure diagram of a terminal device in an embodiment. Figure 4 As shown, the terminal device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the terminal device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor may implement a feature extraction method for synchronized phasor data. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor may implement a feature extraction method for synchronized phasor data. Those skilled in the art will appreciate that Figure 4 The structure shown in the figure is only a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the terminal device to which the scheme of the present application is applied. The specific terminal device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0138] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented: obtaining a sample set; wherein the sample set includes multiple samples, each sample includes multiple sample synchronized phasor data within a preset time period; sequentially inputting the samples into the adaptive feature channel convolution network, calculating the fluctuation amplitude of each sample synchronized phasor data in the sample through the dynamic convolution layer, and selecting a corresponding convolution kernel according to the fluctuation amplitude for convolution processing to obtain multiple convolution feature maps; wherein the fluctuation amplitude is positively correlated with the size of the corresponding selected convolution kernel; Key features of the multiple feature maps are extracted through the attention mechanism of the local feature extraction layer to obtain weighted local features; the global features of the weighted local features are captured through the global feature fusion layer, and the global features are fused with the weighted local features to obtain fused features; the fused features are converted into feature vectors of synchronized phasor data through the fully connected layer, the feature vectors of the synchronized phasor data are input into a preset dynamic weighted feature loss function to calculate the loss value, and the adaptive feature channel convolutional network is iteratively adjusted through the loss value until the adaptive feature channel convolutional network converges.

[0139] And implement the following steps: obtain multiple measured synchronized phasor data within a preset time period; input the multiple measured synchronized phasor data into an adaptive feature channel convolutional network to obtain feature vectors of the multiple measured synchronized phasor data; wherein the adaptive feature channel convolutional network is trained by the above-mentioned adaptive feature channel convolutional network training method.

[0140] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: obtaining a sample set; wherein the sample set comprises a plurality of samples, each sample comprising a plurality of sample synchronized phasor data within a preset time period; sequentially inputting the samples into the adaptive feature channel convolution network, calculating the fluctuation amplitude of each sample synchronized phasor data in the sample through the dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude to obtain a plurality of convolution feature maps; wherein the fluctuation amplitude is proportional to the size of the corresponding selected convolution kernel; The invention relates to a method for realizing a method for realizing a multi-layered convolutional network of synchronous phasor data. The method comprises the following steps: extracting key features of the multiple feature maps through the attention mechanism of the local feature extraction layer to obtain weighted local features; capturing the global features of the weighted local features through the global feature fusion layer, and fusing the global features with the weighted local features to obtain fused features; converting the fused features into feature vectors of synchronized phasor data through the fully connected layer, inputting the feature vectors of the synchronized phasor data into a preset dynamic weighted feature loss function to calculate a loss value, and iteratively adjusting the adaptive feature channel convolutional network through the loss value until the adaptive feature channel convolutional network converges.

[0141] And implement the following steps: obtain multiple measured synchronized phasor data within a preset time period; input the multiple measured synchronized phasor data into an adaptive feature channel convolutional network to obtain feature vectors of the multiple measured synchronized phasor data; wherein the adaptive feature channel convolutional network is trained by the above-mentioned adaptive feature channel convolutional network training method.

[0142] It should be noted that the above-mentioned feature extraction method, model training method, device and computer-readable storage medium for synchronized phasor data belong to a general inventive concept, and the contents in the embodiments of the feature extraction method, model training method, device and computer-readable storage medium for synchronized phasor data can be applied to each other.

[0143] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0144] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A training method for an adaptive feature channel convolutional network, characterized in that: The adaptive feature channel convolution network includes a dynamic convolution layer, a local feature extraction layer, a global feature fusion layer and a fully connected layer connected in sequence, the dynamic convolution layer is provided with a plurality of convolution kernels of increasing sizes in sequence, and the method includes: Acquire a sample set; wherein the sample set includes a plurality of samples, each sample including a plurality of sample synchronized phasor data within a preset time period; The samples are sequentially input into the adaptive feature channel convolution network, the fluctuation amplitude of each sample synchronized phasor data in the samples is calculated through the dynamic convolution layer, and a corresponding convolution kernel is selected according to the fluctuation amplitude for convolution processing to obtain a plurality of convolution feature maps; wherein the fluctuation amplitude is positively correlated with the size of the corresponding selected convolution kernel; Extracting key features from the multiple feature maps through the attention mechanism of the local feature extraction layer to obtain weighted local features; The global feature fusion layer is used to capture the global feature of the weighted local feature, and the global feature is fused with the weighted local feature to obtain a fused feature; The fused features are converted into feature vectors of synchronized phasor data through the fully connected layer, the feature vectors of synchronized phasor data are input into a preset dynamic weighted feature loss function to calculate a loss value, and the adaptive feature channel convolutional network is iteratively adjusted according to the loss value until the adaptive feature channel convolutional network converges.

2. The method according to claim 1, characterized in that The step of calculating the fluctuation amplitude of each sample synchronized phasor data in the sample by the dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude includes: Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as: In the above formula, represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; X i (t) represents the i-th sample synchronized phasor data at time t; represents the mean value of the i-th sample synchronized phasor data within a preset time period; The mapping size of each sample synchronized phasor data in the sample is calculated according to the fluctuation amplitude, which is expressed as: In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max represents the maximum size of all convolution kernels in the dynamic convolution layer, σ 2 max Represents the maximum fluctuation amplitude of all sample synchronized phasor data, σ 2 min It represents the minimum value of the fluctuation amplitude of all sample synchronized phasor data; Based on the mapping size, a convolution kernel corresponding to each sample synchronized phasor data is selected by an integer allocation method to perform convolution processing.

3. The method according to claim 1, characterized in that The step of calculating the fluctuation amplitude of each sample synchronized phasor data in the sample by the dynamic convolution layer, and selecting a corresponding convolution kernel for convolution processing according to the fluctuation amplitude includes: Calculate the fluctuation amplitude of each sample synchronized phasor data in the sample, expressed as: A i =max(X i (t))-min(X i (t)),t∈[1,T] In the above formula, A i represents the fluctuation amplitude of the ith sample synchronized phasor data; T represents the duration of the preset time period; max(X i (t)) represents the maximum value of the ith sample synchronized phasor data within the preset time period; min(X i (t) represents the minimum value of the ith sample synchronized phasor data within a preset time period; The mapping size of each sample synchronized phasor data in the sample is calculated according to the fluctuation amplitude, which is expressed as: In the above formula, k i represents the mapping size of the i-th sample synchronized phasor data, k min represents the minimum size of all convolution kernels in the dynamic convolution layer, k max represents the maximum size of all convolution kernels in the dynamic convolution layer, A min Indicates the minimum fluctuation amplitude of all sample synchronized phasor data, A max It represents the maximum value of the fluctuation amplitude of all sample synchronized phasor data; Based on the mapping size, a convolution kernel corresponding to each sample synchronized phasor data is selected by an integer allocation method to perform convolution processing.

4. The method according to claim 2 or 3, characterized in that: The step of selecting a convolution kernel corresponding to each sample synchronized phasor data by an integer allocation method based on the mapping size for convolution processing includes: The mapping size is k i The synchronized phasor data is distributed to the floor(k i ) is used for convolution processing; If there are at least two synchronized phasor data assigned to the same convolution kernel, the synchronized phasor data whose size difference is not the smallest is assigned to the idle convolution kernel with the smallest corresponding size difference for convolution processing; wherein the size difference is the absolute value of the difference between the mapping size and the convolution kernel size.

5. The method according to claim 1, characterized in that The key features of the multiple feature maps are extracted through the attention mechanism of the local feature extraction layer to obtain weighted local features, which are expressed as: In the above formula, Z Pw (t) represents the weighted local feature, I represents the total number of sample synchronized phasor data in each sample, represents the fluctuation amplitude of the ith sample synchronized phasor data, Z i Characteristic diagram representing the ith sample synchronized phasor data.

6. The method according to claim 5, characterized in that The global feature fusion layer captures the global feature of the weighted local feature, and fuses the global feature with the weighted local feature to obtain a fused feature, which is expressed as: Z Pf (t)=Z Pw (t)+W global ·Z Pg With Pg =GlobalAvgPool(Z Pw ) In the above formula, Z Pf (t) represents the fusion feature, W global Represents the weight matrix of the global feature, Z Pg Represents the global feature, and GlobalAvgPool() represents the global average pooling operation.

7. The method according to claim 1, characterized in that The dynamic weighted feature loss function is expressed as: In the above formula, L DWFL Represents the loss value; F P ' n represents the expected eigenvector of the synchronized phasor data in the nth eigenchannel; F Pn represents the feature vector extracted from the synchronized phasor data in the nth feature channel; w n Represents the weight of the nth feature channel; λ represents the regularization coefficient; W conv Represents the parameter matrix of the convolution kernel.

8. A method for extracting features from synchronized phasor data, characterized in that: The method comprises: Acquire multiple measured synchronized phasor data within a preset time period; The multiple measured synchronized phasor data are input into an adaptive feature channel convolutional network to obtain feature vectors of the multiple measured synchronized phasor data; wherein the adaptive feature channel convolutional network is trained by the method according to claim 1.

9. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 or 8.

10. A terminal device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 or 8.

Images