Composite power quality disturbance identification method, apparatus and device, and storage medium

Abstract

The invention belongs to the technical field of electric energy quality monitoring, and particularly discloses a composite electric energy quality disturbance identification method and device, equipment and a storage medium. According to the invention, mode conversion is carried out on the current electric energy signal according to the target track visual conversion strategy; determining a local key feature map according to the two-dimensional track image and a target efficient network module; determining a global key feature map according to the local key feature map and a target conversion module; and determining a composite power quality disturbance identification result according to a conversion result of the global key feature map. Through the mode, the current one-dimensional electric energy signal is converted into the two-dimensional track image with obvious shape features, sufficient information is provided for subsequent feature extraction, then the target efficient network module is used for accurately extracting local fine features, the target conversion module is used for deeply mining global features, effective fusion and complementation of the features are achieved, and the accuracy of the feature extraction is improved. Therefore, the accuracy of composite power quality disturbance identification can be effectively improved, and the method has relatively high noise immunity.

Description

Technical Field

[0001] This application belongs to the field of power quality monitoring technology, and more specifically, relates to a method, device, equipment and storage medium for identifying complex power quality disturbances. Background Technology

[0002] Driven by the continuous deepening of smart grid construction and the large-scale grid connection of new energy sources, the power system is undergoing unprecedented structural changes and operational transformation. Against this backdrop, the accurate identification of power quality disturbances (PQDs) has become a core challenge in ensuring the safe and stable operation of the power grid. In actual power system operation, electrical energy is affected by various factors, causing the electrical energy reaching users to deviate from a sinusoidal waveform and become distorted, resulting in various power quality disturbance signals and subsequently triggering various power quality problems. Therefore, the identification of complex power quality disturbances is particularly important.

[0003] The foundation of identifying complex power quality disturbances lies in feature extraction. Currently, common methods for feature extraction rely on Fourier transform, which typically extracts single-dimensional features such as those in the frequency or time domains, resulting in incomplete feature extraction. Furthermore, Fourier transform deals with the power signal itself, leading to less obvious extracted features. Therefore, the accuracy of the above methods for identifying complex power quality disturbances is relatively low. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a method, apparatus, device and storage medium for identifying composite power quality disturbances, which aims to solve the problem that the accuracy of identifying composite power quality disturbances is low due to incomplete and unclear feature extraction in the prior art.

[0005] To achieve the above objectives, in a first aspect, this application provides a method for identifying composite power quality disturbances, comprising: The current power signal is acquired, and modal conversion is performed on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image; Local key feature maps are determined based on the two-dimensional trajectory image and the target efficient network module; The global key feature map is determined based on the local key feature map and the target transformation module; The global key feature map is subjected to a nonlinear transformation, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result.

[0006] In one embodiment, the step of acquiring the current power signal and performing modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image includes: Acquire the current power signal and filter the current power signal; The filtered electrical energy signal is fitted, and multi-dimensional signal parameters are determined based on the fitted electrical energy signal. Calculate trajectory parameters based on the multi-dimensional signal parameters; Based on the trajectory parameters and the target trajectory visual conversion strategy, the current power signal is modally converted to obtain a two-dimensional trajectory image.

[0007] In one embodiment, the step of determining local key feature maps based on the two-dimensional trajectory image and the target efficient network module includes: The two-dimensional trajectory image is dilated and convolved according to the first convolutional layer in the target efficient network module, and the low-level feature map of the dilated convolution is standardized. The processed low-level feature map is enhanced using the activation function in the target high-efficiency network module to obtain the current feature map. The current feature map is subjected to depth convolution based on the second convolutional layer in the target efficient network module, and the rich visual features of the depth convolution are standardized. The processed rich feature map is enhanced according to the activation function to obtain the target feature map; The target feature map is weighted by channel attention based on the feature weighting module in the target efficient network module to obtain a weighted feature map. Channel fusion is performed on the weighted feature map based on the first convolutional layer, and the fused feature map is standardized to obtain a local key feature map.

[0008] In one embodiment, the step of determining the global key feature map based on the local key feature map and the target transformation module includes: The local key feature map is morphologically transformed according to the input rules of the target transformation module to obtain the input tensor; The input tensor is projected onto the target transformation module using multiple projection matrices, and an attention score matrix is ​​generated based on the projection results. Multi-head attention results are generated based on the projection results and the attention score matrix; The multi-head attention results are fused based on the linear layer in the target transformation module to obtain the output tensor; The output tensor is reshaped to obtain a global key feature map.

[0009] In one embodiment, the step of performing a nonlinear transformation on the global key feature map and determining the composite power quality disturbance identification result of the current power signal based on the transformation result includes: Flatten the global key feature map to obtain a one-dimensional global key feature vector; The probability weight vector is obtained by calculating the one-dimensional global key feature vector element by element according to the standard normal distribution function. The target feature vector is calculated based on the smooth differentiable function, the probability weight vector, and the one-dimensional global key feature vector; The transformation result is determined based on the target feature vector, weight matrix, and bias vector, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result.

[0010] In one embodiment, the step of determining the composite power quality disturbance identification result of the current power signal based on the transformation result includes: Obtain characteristic information of the transformation result; Based on the aforementioned characteristic information, a target disturbance recognition model is determined using a multi-dimensional, efficient scaling method. The transformation results are aggregated using the target perturbation identification model, and the aggregated transformation results are then classified and mapped. The composite power quality disturbance identification result of the current power signal is determined based on the mapping result.

[0011] Secondly, this application provides a composite power quality disturbance identification device, comprising: The acquisition module is used to acquire the current power signal and perform modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image; The determination module is used to determine local key feature maps based on the two-dimensional trajectory image and the target efficient network module; The determining module is further configured to determine a global key feature map based on the local key feature map and the target transformation module; The transformation module is used to perform nonlinear transformation on the global key feature map and determine the composite power quality disturbance identification result of the current power signal based on the transformation result.

[0012] Thirdly, this application provides an electronic device, comprising: at least one memory for storing a program; and at least one processor for executing the program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to execute the method described in the first aspect or any possible implementation thereof.

[0013] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0014] Fifthly, this application provides a computer program product that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0015] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0016] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: (1) After acquiring the current power signal, this application will also perform modal conversion on the current power signal according to the target trajectory visual conversion strategy, that is, convert the current power signal into a two-dimensional trajectory image with obvious shape features, so as to provide sufficient information for subsequent feature extraction.

[0017] (2) In the feature extraction process, this application also uses the EfficientNet-Transformer hybrid architecture, which includes a target efficient network module and a target transformation module. The efficient network module is used to accurately extract local subtle features, i.e., local key feature maps, while the target transformation module is used to deeply mine global features, i.e., global key feature maps, thereby achieving effective feature fusion and complementarity, which can effectively improve the accuracy of composite power quality disturbance identification. In addition, the EfficientNet-Transformer hybrid architecture has strong resistance to noise, i.e., it has both high accuracy and noise resistance.

[0018] In summary, this application acquires the current power signal and performs modal conversion on the current power signal according to a target trajectory visual conversion strategy to obtain a two-dimensional trajectory image. Based on the two-dimensional trajectory image and a target efficient network module, local key feature maps are determined. Based on the local key feature maps and a target conversion module, a global key feature map is determined. A nonlinear transformation is performed on the global key feature map, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result. Through the above method, the one-dimensional current power signal is converted into a two-dimensional trajectory image with obvious shape features, providing sufficient information for subsequent feature extraction. Then, the target efficient network module accurately extracts local subtle features, and the target conversion module deeply mines global features, achieving effective feature fusion and complementarity. Finally, the composite power quality disturbance identification result of the current power signal is determined based on the transformation result, thereby effectively improving the accuracy of composite power quality disturbance identification and providing strong resistance to signal noise. Attached Figure Description

[0019] Figure 1This is one of the flowcharts illustrating the composite power quality disturbance identification method provided in the embodiments of this application; Figure 2 This is a schematic diagram of power quality disturbance signal changes provided in an embodiment of this application; Figure 3 This is a second schematic flowchart of the composite power quality disturbance identification method provided in the embodiments of this application; Figure 4 This is a schematic diagram of the module structure of the composite power quality disturbance identification device provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0022] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0023] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0024] Based on this, embodiments of this application provide a method for identifying composite power quality disturbances, referring to... Figure 1 , Figure 1 This is one of the flowcharts illustrating the composite power quality disturbance identification method provided in this application embodiment. In this embodiment, the composite power quality disturbance identification method includes steps S10 to S40: Step S10: Obtain the current power signal and perform mode conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image.

[0025] It should be noted that the current power signal refers to the signal used for identifying composite power quality disturbances at the current moment. This current power signal can be a three-phase voltage or current signal. Power quality includes, but is not limited to, voltage quality, current quality, supply quality, and consumption quality. In an ideal scenario, the power energy is a standard power frequency sinusoidal voltage maintaining its rated amplitude and frequency. In a three-phase AC power system, it is also required that the voltage and current of each phase be equal in amplitude, symmetrical in phase, and 120° out of phase. In actual operation of the power system, power energy is affected by various factors, causing the power reaching users to deviate from the sinusoidal waveform and become distorted, resulting in various power quality disturbance signals and thus triggering various power quality problems. For example, the large-scale distributed integration of new energy sources such as wind and solar power into the power grid is increasing grid instability; the operation of numerous nonlinear, impulsive, and unbalanced loads in the power system is further deteriorating the grid environment; and the extensive use of power electronic equipment in railway electrification, electrolytic plants in the electrochemical industry, and electric arc furnaces in industrial steelmaking not only causes insulation aging and reduced lifespan of power equipment but also pollutes the power grid to some extent, causing issues such as voltage deviations, voltage fluctuations, and harmonics. Therefore, it is necessary to accurately identify power quality disturbances.

[0026] It is understandable that power quality disturbance signals include, but are not limited to, voltage surges, voltage drops, voltage interruptions, harmonics, voltage oscillation transients, and voltage flicker, etc. (Refer to...) Figure 2 , Figure 2This diagram illustrates the changes in power quality disturbance signals, specifically: 0-0.1 seconds represent a voltage surge, 0.1-0.2 seconds represent a voltage drop, 0.2-0.3 seconds represent a voltage interruption, 0.3-0.4 seconds represent a voltage spike, 0.4-0.5 seconds represent transient oscillations, 0.5-0.6 seconds represent voltage dips, 0.6-0.7 seconds represent voltage flicker, and 0.7-0.8 seconds represent harmonic distortion. A voltage surge, also known as a voltage spike or voltage spurt, refers to an electromagnetic disturbance phenomenon where the root mean square value of the power frequency voltage rises to 1.1-1.8 (pu) for a duration of 0.5 cycles to 60 seconds, while the system frequency remains at its nominal value. The main causes of voltage surges include: recovery from short-circuit faults in the circuit system, shutdown or disconnection of heavy industrial loads, and switching of capacitor banks. A voltage sag, also known as a voltage dip or voltage trough, is the opposite of a voltage rise. It is an electromagnetic disturbance where the root mean square (RMS) value of the power frequency voltage drops to 0.1-0.9 (pu) for a duration of 0.5 cycles to 60 seconds, while the system frequency remains at its nominal value. The main causes of voltage sags include: short circuits in the circuit system, heavy-load startup in industrial applications, or power switching. A voltage interruption refers to a power quality disturbance where the RMS value of the power frequency voltage falls below 0.1 (pu) for a duration of 0.5 cycles to 60 seconds, while the system frequency remains at its nominal value. The main causes of voltage interruptions include: power system faults, mechanical equipment failures, and control equipment malfunctions.

[0027] It should also be noted that harmonics refer to the components whose frequencies are integer multiples of the fundamental frequency, obtained by Fourier series decomposition of a periodic non-sinusoidal alternating current. The main causes of harmonics include: nonlinear load injection, transformer no-load or over-excitation, and inrush current during capacitor switching in the power grid. Voltage oscillation transients refer to non-power frequency voltage or current pulses that occur during steady-state operation. They often exist as unipolar pulses or spikes of either polarity. The main causes of transient oscillations include: lightning, poor grounding, and inductive load switching. Voltage flicker refers to a subjective visual phenomenon caused by the periodic variation of the root mean square value of the power frequency voltage within the range of 0.95-1.05 (pu), resulting in fluctuations in the perceived illuminance of incandescent lamps. The main causes of voltage flicker include: fluctuating load injection, system short-circuit faults, and automatic switching of backup power supplies.

[0028] Further, step S10 includes: acquiring the current power signal and filtering the current power signal; fitting the filtered power signal and determining multi-dimensional signal parameters based on the fitted power signal; calculating trajectory parameters based on the multi-dimensional signal parameters; and performing modal conversion on the current power signal based on the trajectory parameters and the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image.

[0029] It should be understood that, in order to provide sufficient information for subsequent feature extraction, this embodiment will also perform mode conversion on the current power signal, that is, convert the current power signal into a two-dimensional trajectory image with obvious shape features. To effectively improve the accuracy of mode conversion, the current power signal needs to be filtered to remove noise and irrelevant interference. Then, the filtered power signal is fitted to extract quantifiable multi-dimensional signal parameters, providing structured data for trajectory image generation. These multi-dimensional signal parameters include, but are not limited to, phase difference, amplitude, and angular frequency. At this point, trajectory parameters can be calculated based on the multi-dimensional signal parameters, specifically:

[0030] in, Indicates in Trajectory parameters in the axial direction, and Both represent amplitude, which determines the aspect ratio of the trajectory in the two-dimensional trajectory image. , It represents angular frequency and determines the basic structure of the trajectory in a two-dimensional trajectory image. This represents the phase difference, which affects the specific shape of the trajectory in a two-dimensional trajectory image. Indicates time.

[0031] Understandably, after obtaining the trajectory parameters, the current electrical signal is converted into a two-dimensional trajectory image according to the target trajectory visual transformation strategy. This two-dimensional trajectory image contains the geometric shape of the current electrical signal, such as an ellipse or a rotated shape, for input to the target high-efficiency network module. Furthermore, this target trajectory visual transformation strategy can be a Lissajous trajectory color mapping strategy.

[0032] Step S20: Determine local key feature maps based on the two-dimensional trajectory image and the target efficient network module.

[0033] As can be understood, a local key feature map refers to an image representing the texture, edges, and local geometric features in a two-dimensional trajectory image. After obtaining the two-dimensional trajectory image, it can be input into a target high-efficiency network module, which will then output a local key feature map. This high-efficiency network module can be an EfficientNet module, used to accurately extract subtle local features, i.e., the local key feature map.

[0034] Step S30: Determine the global key feature map based on the local key feature map and the target conversion module.

[0035] It should be understood that the global key feature map refers to the feature map that captures the temporal and spatial pattern dependencies between perturbations. After obtaining the local key feature map output by the target efficient network module, this local key feature map can be input into the target transformation module, which will then output the global key feature map. This target transformation module can be a Transformer encoder, used to deeply mine global features, i.e., the global key feature map, to achieve effective feature fusion and complementarity. Furthermore, the target efficient network module and the target transformation module can be combined into an EfficientNet-Transformer hybrid architecture. This hybrid framework can accurately extract key features and achieve accurate recognition, while also possessing strong resistance to noise, thus combining high accuracy and noise resistance.

[0036] Further, step S30 includes: performing morphological transformation on the local key feature map according to the input rules of the target transformation module to obtain an input tensor; projecting the input tensor onto multiple projection matrices in the target transformation module respectively, and generating an attention score matrix based on the projection results; generating a multi-head attention result based on the projection results and the attention score matrix; fusing the multi-head attention result according to the linear layer in the target transformation module to obtain an output tensor; and reshaping the output tensor to obtain a global key feature map.

[0037] It is understandable that the input rules refer to the rules that the input data must meet to reach the target transformation module. To ensure that the local key feature map can be input into the target transformation module, the local key feature map needs to undergo morphological transformation according to the input rules of the target transformation module. Then, the input tensor is projected according to multiple projection matrices in the target transformation module. These projection matrices can be query matrices, key matrices, and value matrices. The projection result can then be expressed as:

[0038] in, This indicates that the query is for projection results. Indicates the input tensor. Indicates querying the projection matrix. Indicates the key projection result. Represents the key projection matrix. Indicates the projection result of the value. Represents a value matrix.

[0039] It should be noted that after obtaining the projection result of the input tensor, an attention score matrix can be generated based on the projection result to capture global dependencies, specifically:

[0040] in, Represents the attention score matrix, Indicates transpose. This represents the square root of the attention dimension.

[0041] It should be understood that, in order to enhance the ability to capture different types of global dependencies, after obtaining the attention score matrix, multi-head attention results can be generated by combining the projection results, and the multi-head attention results can be fused according to the linear layer in the target transformation module, specifically as follows:

[0042] in, Indicates the output tensor. Represents the normalization function. This represents a linear layer in the target transformation module.

[0043] Step S40: Perform a nonlinear transformation on the global key feature map, and determine the composite power quality disturbance identification result of the current power signal based on the transformation result.

[0044] Understandably, in order to effectively improve the accuracy of identifying complex power quality disturbances, after obtaining the global key feature map, it is necessary to perform a nonlinear transformation on the global key feature map through a multi-layer perceptron layer (MLP), and then determine the identification result of the complex power quality disturbance of the current power signal based on the transformation result. The identification result of the complex power quality disturbance can be voltage surge, voltage drop, harmonics, and flicker, etc.

[0045] Further, step S40 includes: flattening the global key feature map to obtain a one-dimensional global key feature vector; calculating the probability weight vector element-by-element of the one-dimensional global key feature vector according to the standard normal distribution function; calculating the target feature vector according to the smooth differentiable function, the probability weight vector, and the one-dimensional global key feature vector; determining the transformation result according to the target feature vector, the weight matrix, and the bias vector; and determining the composite power quality disturbance identification result of the current power signal according to the transformation result.

[0046] It should be noted that, in order to integrate all local subtle features and global dependencies, the global key feature map containing spatial structure needs to be flattened into a one-dimensional vector that can be processed by the multilayer perceptron layer, i.e., a one-dimensional global key feature vector. At this point, the one-dimensional global key feature vector can be calculated element-wise according to the standard normal distribution function. This standard normal distribution function can be the cumulative distribution function (CDF) of the standard normal distribution, specifically:

[0047] in, Represents the probability weight vector. This represents a one-dimensional global key feature vector. , This indicates a temporary variable.

[0048] It should be understood that after obtaining the probability weight vector, the target feature vector can be calculated by combining a smooth, differentiable function and a one-dimensional global key feature vector. This smooth, differentiable function can be... The function, specifically:

[0049] in, Represents the target feature vector. Represents the probability weight vector. This represents a one-dimensional global key feature vector.

[0050] It is understandable that, after obtaining the target feature vector, the transformation result can be determined by combining the target feature vector, the weight matrix, and the bias vector, specifically as follows:

[0051] in, Indicates the result of the transformation. Denotes the weight matrix, and , This represents the bias vector. .

[0052] Furthermore, the step of determining the composite power quality disturbance identification result of the current power signal based on the transformation result includes: acquiring characteristic information of the transformation result; determining a target disturbance identification model based on the characteristic information using a multi-dimensional efficient scaling method; aggregating the transformation result using the target disturbance identification model, and classifying and mapping the aggregated transformation result; and determining the composite power quality disturbance identification result of the current power signal based on the mapping result.

[0053] Understandably, in order to effectively improve the accuracy of determining the composite power quality disturbance identification results, this embodiment will also determine the target disturbance identification model through multi-dimensional efficient scaling based on characteristic information. This multi-dimensional efficient scaling method includes, but is not limited to, width scaling, depth scaling, etc. Efficient scaling of the model is achieved by jointly optimizing the three dimensions of network width, depth, and input resolution. Specifically:

[0054] in, Indicates network depth. Indicates network width. Indicates the input resolution. Indicates network depth. This represents a hyperparameter that controls the network width. This represents the hyperparameter that controls the input resolution.

[0055] This embodiment acquires the current power signal and performs modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image. Local key feature maps are determined based on the two-dimensional trajectory image and the target efficient network module. Global key feature maps are determined based on the local key feature maps and the target conversion module. A nonlinear transformation is performed on the global key feature map, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result. Through the above method, the one-dimensional current power signal is converted into a two-dimensional trajectory image with obvious shape features, providing sufficient information for subsequent feature extraction. Then, the target efficient network module accurately extracts local subtle features, and the target conversion module deeply mines global features, achieving effective feature fusion and complementarity. Finally, the composite power quality disturbance identification result of the current power signal is determined based on the transformation result, thereby effectively improving the accuracy of composite power quality disturbance identification and providing strong resistance to signal noise.

[0056] In one specific implementation, this application provides steps for determining local key feature maps. Please refer to... Figure 3 , Figure 3 This is the second schematic flowchart of the composite power quality disturbance identification method provided in the embodiments of this application. Step S20 includes steps S201 to S206: Step S201: Perform dilated convolution on the two-dimensional trajectory image according to the first convolutional layer in the target efficient network module, and standardize the low-level feature map of the dilated convolution.

[0057] It should be noted that the target efficient network module includes, but is not limited to, the first convolutional layer, the second convolutional layer, and the feature weighting module. The kernel size of the first convolutional layer can be... At this point, the two-dimensional trajectory image can be dilated and convolved based on the first convolutional layer in the target efficient network module, and the low-level feature map of the dilated convolution can be standardized. Step S202: Perform feature enhancement on the processed low-level feature map according to the activation function in the target efficient network module to obtain the current feature map.

[0058] It is understandable that after obtaining the low-level feature map after output, feature enhancement can be performed on the processed low-level feature map according to the activation function in the target efficient network module, which can be specifically expressed as:

[0059] in, Indicates the current feature map, Represents a two-dimensional trajectory image. This represents the batch normalization function. This represents the activation function. This represents a filter.

[0060] Step S203: Perform depth convolution on the current feature map according to the second convolutional layer in the target high-efficiency network module, and standardize the rich visual features of the depth convolution.

[0061] It should be understood that the kernel size of the second convolutional layer can be... To expand the receptive field and capture a wider range of local geometric features without increasing the number of parameters, a depthwise convolution can be performed on the current feature map using the second convolutional layer in the target efficient network module. The rich visual features from the depthwise convolution are then standardized. The dilation factor of this second convolutional layer can be... .

[0062] Step S204: Perform feature enhancement on the processed rich feature map according to the activation function to obtain the target feature map.

[0063] Understandably, compared to the current feature map, the target feature map has a larger receptive field and contains richer local information. After obtaining the processed rich feature map, feature enhancement can be performed on the processed rich feature map according to the activation function, which can be specifically expressed as:

[0064] in, Represents the target feature map. Indicates the current feature map, This represents the batch normalization function. This represents the activation function. This represents a filter.

[0065] Step S205: Perform channel attention weighting on the target feature map according to the feature weighting module in the target efficient network module to obtain a weighted feature map.

[0066] It should be understood that, in order to achieve shape invariance and weighted enhancement of channel features, channel attention weighting of the target feature map needs to be applied according to the feature weighting module in the target efficient network module. Specifically, this can be expressed as:

[0067] in, This represents a weighted feature map. This indicates the feature weighting module. Represents the target feature map.

[0068] Step S206: Channel fusion is performed on the weighted feature map according to the first convolutional layer, and the fused feature map is standardized to obtain a local key feature map.

[0069] It should be noted that, in order to integrate the attention-weighted multi-channel feature maps, channel fusion of the weighted feature maps is required based on the first convolutional layer; to ensure stable output of local key feature maps, the fused feature maps also need to be standardized, specifically as follows:

[0070] in, Represents a local key feature map. This represents a weighted feature map. This represents the batch normalization function. This represents the activation function. This represents a filter.

[0071] In this embodiment, the two-dimensional trajectory image is subjected to dilated convolution by the first convolutional layer in the target efficient network module, and the low-level feature map of the dilated convolution is standardized. The processed low-level feature map is then enhanced using the activation function in the target efficient network module to obtain the current feature map. Next, the current feature map is subjected to depth convolution by the second convolutional layer in the target efficient network module, and the rich visual features of the depth convolution are standardized. The processed rich feature map is then enhanced using the activation function to obtain the target feature map. The target feature map is then subjected to channel attention weighting by the feature weighting module in the target efficient network module to obtain a weighted feature map. Finally, the weighted feature map is fused using the first convolutional layer, and the fused feature map is standardized to obtain a local key feature map. By using the above method, multiple convolutional layers are used for dilated convolution and depthwise convolution. Then, channel attention weighting is applied to the target feature map according to the feature weighting module to achieve shape invariance and weighted enhancement of channel features. The weighted feature map is then processed into a local key feature map through channel fusion, standardization and other methods. This enables the accurate extraction of local subtle features of two-dimensional trajectory images and effectively improves the accuracy of obtaining local key feature maps.

[0072] The composite power quality disturbance identification device provided in this application is described below. The composite power quality disturbance identification device described below corresponds to the composite power quality disturbance identification method described above. Please refer to... Figure 4 , Figure 4 This is a schematic diagram of the module structure of the composite power quality disturbance identification device provided in the embodiments of this application, including: The acquisition module T10 is used to acquire the current power signal and perform modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image.

[0073] The determination module T20 is used to determine local key feature maps based on the two-dimensional trajectory image and the target efficient network module.

[0074] The determining module T20 is also used to determine the global key feature map based on the local key feature map and the target conversion module.

[0075] The transformation module T30 is used to perform nonlinear transformation on the global key feature map and determine the composite power quality disturbance identification result of the current power signal based on the transformation result.

[0076] This embodiment acquires the current power signal and performs modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image. Local key feature maps are determined based on the two-dimensional trajectory image and the target efficient network module. Global key feature maps are determined based on the local key feature maps and the target conversion module. A nonlinear transformation is performed on the global key feature map, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result. Through the above method, the one-dimensional current power signal is converted into a two-dimensional trajectory image with obvious shape features, providing sufficient information for subsequent feature extraction. Then, the target efficient network module accurately extracts local subtle features, and the target conversion module deeply mines global features, achieving effective feature fusion and complementarity. Finally, the composite power quality disturbance identification result of the current power signal is determined based on the transformation result, thereby effectively improving the accuracy of composite power quality disturbance identification and providing strong resistance to signal noise.

[0077] It is understood that the detailed functional implementation of each of the above modules can be found in the description of the aforementioned method embodiments, and will not be repeated here.

[0078] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0079] Based on the methods in the above embodiments, this application provides an electronic device, please refer to... Figure 5 , Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0080] It should be noted that the system may include: a processor 10, a communications interface 20, a memory 30, and a communication bus 40. The processor 10, communications interface 20, and memory 30 communicate with each other via the communication bus 40. The processor 10 can invoke logical instructions stored in the memory 30 to execute the methods described in the above embodiments.

[0081] Furthermore, the logical instructions in the aforementioned memory 30 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a 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, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0082] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0083] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0084] It is understood that the processor in the embodiments of this application can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, field-programmable gate arrays, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0085] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor.

[0086] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application. Those skilled in the art will readily understand that the above descriptions are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for identifying composite power quality disturbances, characterized in that, include: The current power signal is acquired, and modal conversion is performed on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image; Local key feature maps are determined based on the two-dimensional trajectory image and the target efficient network module; The global key feature map is determined based on the local key feature map and the target transformation module; The global key feature map is subjected to a nonlinear transformation, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result.

2. The method as described in claim 1, characterized in that, The step of acquiring the current power signal and performing modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image includes: Acquire the current power signal and filter the current power signal; The filtered electrical energy signal is fitted, and multi-dimensional signal parameters are determined based on the fitted electrical energy signal. Calculate trajectory parameters based on the multi-dimensional signal parameters; Based on the trajectory parameters and the target trajectory visual conversion strategy, the current power signal is modally converted to obtain a two-dimensional trajectory image.

3. The method as described in claim 1, characterized in that, The step of determining local key feature maps based on the two-dimensional trajectory image and the target efficient network module includes: The two-dimensional trajectory image is dilated and convolved according to the first convolutional layer in the target efficient network module, and the low-level feature map of the dilated convolution is standardized. The processed low-level feature map is enhanced using the activation function in the target high-efficiency network module to obtain the current feature map. The current feature map is subjected to depth convolution based on the second convolutional layer in the target efficient network module, and the rich visual features of the depth convolution are standardized. The processed rich feature map is enhanced according to the activation function to obtain the target feature map; The target feature map is weighted by channel attention based on the feature weighting module in the target efficient network module to obtain a weighted feature map. Channel fusion is performed on the weighted feature map based on the first convolutional layer, and the fused feature map is standardized to obtain a local key feature map.

4. The method as described in claim 1, characterized in that, The step of determining the global key feature map based on the local key feature map and the target transformation module includes: The local key feature map is morphologically transformed according to the input rules of the target transformation module to obtain the input tensor; The input tensor is projected onto the target transformation module using multiple projection matrices, and an attention score matrix is ​​generated based on the projection results. Multi-head attention results are generated based on the projection results and the attention score matrix; The multi-head attention results are fused based on the linear layer in the target transformation module to obtain the output tensor; The output tensor is reshaped to obtain a global key feature map.

5. The method according to any one of claims 1 to 4, characterized in that, The step of performing a nonlinear transformation on the global key feature map and determining the composite power quality disturbance identification result of the current power signal based on the transformation result includes: Flatten the global key feature map to obtain a one-dimensional global key feature vector; The probability weight vector is obtained by calculating the one-dimensional global key feature vector element by element according to the standard normal distribution function. The target feature vector is calculated based on the smooth differentiable function, the probability weight vector, and the one-dimensional global key feature vector; The transformation result is determined based on the target feature vector, weight matrix, and bias vector, and the composite power quality disturbance identification result of the current power signal is determined based on the transformation result.

6. The method as described in claim 5, characterized in that, The step of determining the composite power quality disturbance identification result of the current power signal based on the transformation result includes: Obtain characteristic information of the transformation result; Based on the aforementioned characteristic information, a target disturbance recognition model is determined using a multi-dimensional, efficient scaling method. The transformation results are aggregated using the target perturbation identification model, and the aggregated transformation results are then classified and mapped. The composite power quality disturbance identification result of the current power signal is determined based on the mapping result.

7. A composite power quality disturbance identification device, characterized in that, include: The acquisition module is used to acquire the current power signal and perform modal conversion on the current power signal according to the target trajectory visual conversion strategy to obtain a two-dimensional trajectory image; The determination module is used to determine local key feature maps based on the two-dimensional trajectory image and the target efficient network module; The determining module is further configured to determine a global key feature map based on the local key feature map and the target transformation module; The transformation module is used to perform nonlinear transformation on the global key feature map and determine the composite power quality disturbance identification result of the current power signal based on the transformation result.

8. An electronic device, characterized in that, include: At least one memory for storing computer programs; At least one processor is configured to execute a program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is run on the processor, it causes the processor to perform the method as described in any one of claims 1-6.

10. A computer program product, characterized in that, When the computer program product is run on a processor, the processor causes the processor to perform the method as described in any one of claims 1-6.

Images