Power quality evaluation method and device

By decomposing and transforming voltage signals using a fully linear KAN network model based on multi-scale feature perception, the type and extent of disturbances in the power system are identified, solving the accuracy and real-time problems of power quality assessment in existing technologies and achieving more efficient power quality assessment.

CN118982280APending Publication Date: 2024-11-19STATE GRID INFORMATION & TELECOMM BRANCH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411053080.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Existing power quality assessment methods have shortcomings in identifying voltage signal disturbances, such as missed detections or misjudgments. They are unable to fully reflect the complex characteristics of voltage signals at different time scales and frequencies, and their computational complexity is high, making it difficult to meet the needs of real-time assessment.

Method used

A fully linear KAN network model based on multi-scale feature perception is used to perform multi-scale decomposition and linear transformation on the voltage signal. The type and degree of disturbance are identified by the target classifier, including voltage drop, voltage rise, voltage interruption, etc., and a disturbance classification model is constructed to improve the accuracy of assessment.

Benefits of technology

It improves the comprehensiveness and accuracy of power quality assessment, can identify the type and degree of voltage signal disturbance in real time, simplifies the calculation complexity, and meets the real-time assessment needs of large-scale power systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982280A_ABST
    Figure CN118982280A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a power quality evaluation method and device, and relates to the technical field of power systems. The method comprises the following steps: acquiring a voltage signal of a power system; a voltage signal is input into a disturbance classification model, a disturbance type and a disturbance degree corresponding to the voltage signal are obtained, and the disturbance classification model adopts a full-linear KAN network model based on multi-scale feature perception; and performing power quality evaluation on the voltage signal according to the disturbance type and the disturbance degree. Therefore, the voltage signal of the power system is input into the disturbance classification model constructed through the KAN network model, the characteristics of the voltage signal under different time scales and different frequencies can be captured, and the disturbance type and disturbance degree of the voltage signal can be identified, so that the comprehensiveness and accuracy of power quality evaluation are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method and device for evaluating power quality. Background Art

[0002] As an indispensable form of energy in modern society, the quality of electric energy is directly related to the production efficiency of all walks of life and the quality of people's daily life. With the rapid development of power technology, the scale of power systems continues to expand, and power loads are becoming more and more complex and diverse. This complexity has brought about a significant increase in disturbance phenomena in voltage signals, such as voltage fluctuations, harmonics, transient pulses, etc. These disturbances will not only affect the stable operation of the power system, but may also pose a threat to the safety of power grid equipment, user equipment and even the entire power system.

[0003] In the related art, the power quality assessment methods of power systems mainly include harmonic analysis based on Fourier transform and time-frequency analysis based on wavelet transform. Both methods rely on manually designed feature extractors to analyze and process voltage signals.

[0004] However, they can only capture some of the features in the signal and cannot fully reflect the complex characteristics of the voltage signal at different time scales and frequencies. Therefore, these methods often miss or misjudge the disturbances in the voltage signal, which results in inaccurate and incomplete identification of the disturbances. Summary of the invention

[0005] Based on the above problems, the present application provides a power quality assessment method and device, which can improve the accuracy and comprehensiveness of disturbance identification.

[0006] The embodiments of the present application disclose the following technical solutions:

[0007] In a first aspect, the present application discloses a method for evaluating power quality, the method comprising:

[0008] Obtain voltage signals of power systems;

[0009] The disturbance type and disturbance degree corresponding to the voltage signal are obtained by inputting the voltage signal into a disturbance classification model, wherein the disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception, and the disturbance type includes at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation;

[0010] Performing power quality assessment on the voltage signal according to the disturbance type and the disturbance degree.

[0011] Optionally, obtaining a disturbance type and a disturbance degree corresponding to the voltage signal includes:

[0012] Through the multi-scale feature perception mechanism of the disturbance classification model, the voltage signal is multi-scale decomposed using N sampling rates to obtain the original features of the voltage signal on N time scales, where N is a natural number;

[0013] Through the full linear network structure of the disturbance classification model, the original features on the N time scales are respectively subjected to linear transformation processing to obtain deep features on the N time scales;

[0014] Concatenate all the deep features on the N time scales to obtain a high-dimensional feature vector;

[0015] The high-dimensional feature vector is subjected to disturbance classification through a target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal. The target classifier is a classifier including a linear layer, an activation function and an exit Dropout mechanism.

[0016] Optionally, the performing multi-scale decomposition on the voltage signal using N sampling rates to obtain original features of the voltage signal on N time scales includes:

[0017] Performing multi-scale decomposition on the voltage signal using N sampling rates to obtain signal sequences on N time scales;

[0018] The signal sequence at each time scale is divided into equal-length and non-overlapping signal windows;

[0019] A linear layer is used to perform feature extraction on each of the signal windows to obtain original features of the voltage signal on the N time scales.

[0020] Optionally, performing linear transformation processing on the original features on the N time scales to obtain deep features on the N time scales includes:

[0021] The linear layer is used to perform linear transformation processing and piecewise polynomial interpolation processing on the original features on the N time scales to obtain deep features on the N time scales.

[0022] Optionally, the method further comprises:

[0023] If the evaluation result of the power quality evaluation is lower than the quality threshold, a warning message is triggered.

[0024] In a second aspect, an embodiment of the present application discloses a power quality assessment device, the device comprising: a signal acquisition module, a disturbance acquisition module and a quality assessment module;

[0025] The signal acquisition module is used to acquire the voltage signal of the power system;

[0026] The disturbance acquisition module is used to obtain the disturbance type and disturbance degree corresponding to the voltage signal by inputting the voltage signal into a disturbance classification model, wherein the disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception, and the disturbance type includes at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation;

[0027] The quality assessment module is used to perform power quality assessment on the voltage signal according to the disturbance type and the disturbance degree.

[0028] Optionally, the disturbance acquisition module includes: a decomposition submodule, a transformation submodule, a splicing submodule and a classification submodule;

[0029] The decomposition submodule is used to perform multi-scale decomposition on the voltage signal using N sampling rates through the multi-scale feature perception mechanism of the disturbance classification model to obtain original features of the voltage signal on N time scales, where N is a natural number;

[0030] The transformation submodule is used to perform linear transformation processing on the original features on the N time scales respectively through the full linear network structure of the disturbance classification model to obtain deep features on the N time scales;

[0031] The splicing submodule is used to splice the deep features on all the N time scales to obtain a high-dimensional feature vector;

[0032] The classification submodule is used to perform disturbance classification on the high-dimensional feature vector through a target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal. The target classifier is a classifier including a linear layer, an activation function and an exit Dropout mechanism.

[0033] Optionally, the decomposition submodule includes: a first decomposition submodule, a second decomposition submodule and a third decomposition submodule;

[0034] The first decomposition submodule is used to perform multi-scale decomposition on the voltage signal using N sampling rates to obtain signal sequences on N time scales;

[0035] The second decomposition submodule is used to divide the signal sequence on each time scale into equal-length and non-overlapping signal windows;

[0036] The third decomposition submodule is used to extract features from each of the signal windows using a linear layer to obtain original features of the voltage signal on the N time scales.

[0037] Optionally, the transformation submodule is specifically used to: use a linear layer to perform linear transformation processing and piecewise polynomial interpolation processing on the original features on the N time scales, so as to obtain deep features on the N time scales.

[0038] Optionally, the device further comprises: a warning triggering module;

[0039] The warning triggering module is used to trigger a warning message if the evaluation result of the power quality evaluation is lower than a quality threshold.

[0040] Compared with the prior art, this application has the following beneficial effects:

[0041] An embodiment of the present application provides a power quality assessment method and device. The method inputs the voltage signal of the power system into a disturbance classification model constructed by a KAN network model, which can capture the characteristics of the voltage signal at different time scales and frequencies, and identify the disturbance type and degree of the voltage signal, thereby improving the comprehensiveness and accuracy of the power quality assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 A flow chart of a power quality assessment method provided in an embodiment of the present application;

[0044] Figure 2 A schematic diagram of the structure of a disturbance classification model provided in an embodiment of the present application;

[0045] Figure 3 A schematic diagram of a KANScanner module provided in an embodiment of the present application;

[0046] Figure 4 A schematic diagram of a KANBlock module provided in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of a power quality assessment device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] First, the technical terms involved in this application are explained:

[0049] Voltage sag refers to the drop of effective voltage value to between 10% and 90% of the rated value. It can be further divided according to the duration: between 0.5 cycles and 30 cycles is called instantaneous sag; between 30 cycles and 3 seconds is called temporary sag; between 3 seconds and 1 minute is called short-term sag.

[0050] Voltage swell refers to the effective value of the voltage rising to more than 110% of the rated value. Similarly, it can be divided into instantaneous swells between 0.5 cycles and 30 cycles, temporary swells between 30 cycles and 3 seconds, and short-term swells between 3 seconds and 1 minute according to the duration.

[0051] Voltage interruption refers to the voltage effective value dropping below 10% of the rated value. According to the duration, it can be divided into: between 0.5 cycle and 3 seconds, it is called transient interruption; between 3 seconds and 1 minute, it is called transient interruption; and more than 1 minute, it is called continuous interruption.

[0052] Undervoltage means that the effective value of the voltage is between 80% and 90% of the rated value and lasts for more than 1 minute.

[0053] Overvoltage refers to the voltage whose effective value is between 110% and 120% of the rated value and lasts for more than 1 minute.

[0054] Transient pulse refers to the disturbance phenomenon of unipolar transient fluctuation of voltage or current, which usually occurs when rising or decaying and occurs for a very short time.

[0055] Transient oscillation refers to the disturbance phenomenon of voltage or current with reciprocating fluctuations of positive and negative polarity. According to the oscillation frequency, it can be divided into low-frequency oscillation (frequency < 5kHz), medium-frequency oscillation (5kHz < frequency < 500kHz), and high-frequency oscillation (0.5MHz < frequency < 5MHz).

[0056] Harmonics refer to sinusoidal voltages or currents that are integer multiples of the operating frequency of the power system.

[0057] Interharmonics refer to sinusoidal voltages or currents that are non-integer multiples of the operating frequency of the power system.

[0058] Voltage flicker refers to a disturbance phenomenon in which the voltage envelope changes regularly or randomly.

[0059] A voltage notch is a periodic voltage disturbance with a negative polarity and a duration of less than 0.5 cycle. It is usually caused by commutation of power electronic equipment in the power system.

[0060] Frequency deviation refers to the phenomenon that the actual frequency of the power system deviates from the standard frequency (fundamental frequency), and its duration is less than 10 seconds. The frequency deviation allowed in relevant standards in my country is ±0.2Hz. When the system capacity is small, the frequency deviation can be relaxed to ±0.5Hz.

[0061] As described above, the power quality assessment methods of power systems mainly include harmonic analysis based on Fourier transform and time-frequency analysis based on wavelet transform.

[0062] Among them, the harmonic analysis method based on Fourier transform converts the voltage signal from the time domain to the frequency domain, and uses the Fourier series or Fourier transform to identify the harmonic components in the signal. However, the Fourier transform has limitations when processing non-stationary signals or time-varying signals, and it is difficult to accurately capture the transient characteristics and time-varying information in the signal. The time-frequency analysis method based on wavelet transform can decompose the signal at different time scales and frequencies by introducing the concept of multi-scale analysis, thereby extracting the time-frequency features in the signal. Although the wavelet transform performs well in processing non-stationary signals, its feature extraction process still relies on manually designed filter groups and wavelet basis functions, which makes it difficult to fully capture the complex features in the voltage signal. Therefore, the current power quality assessment method often misses or misjudges when identifying disturbances in the voltage signal, resulting in inaccurate and incomplete identification of disturbances.

[0063] Furthermore, since there are many types of disturbances in the power system and the characteristics of different disturbances vary greatly, when using Fourier transform-based harmonic analysis and wavelet transform-based time-frequency analysis to assess power quality, their recognition ability and accuracy will be greatly reduced when faced with completely new or unknown disturbance types.

[0064] In addition, the computational complexity of harmonic analysis based on Fourier transform and time-frequency analysis based on wavelet transform is usually high. In practical applications, this may lead to a long evaluation process and make it difficult to meet the needs of real-time evaluation. Especially for large-scale power systems, this computational complexity problem is more prominent.

[0065] After research, the inventor proposed a method and device for evaluating power quality, which includes: obtaining the voltage signal of the power system; obtaining the disturbance type and disturbance degree corresponding to the voltage signal by inputting the voltage signal into the disturbance classification model, the disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception, and the disturbance type includes at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage gap and frequency deviation; according to the disturbance type and disturbance degree, the voltage signal is evaluated for power quality. Therefore, the voltage signal of the power system is input into the disturbance classification model constructed by the KAN network model, which can capture the characteristics of the voltage signal at different time scales and frequencies, identify the disturbance type and disturbance degree of the voltage signal, and thus improve the comprehensiveness and accuracy of the power quality evaluation.

[0066] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0067] See also Figure 1 , which is a flow chart of a power quality assessment method provided in an embodiment of the present application. The method includes:

[0068] S101: Acquire a voltage signal of a power system.

[0069] In the initial stage of power quality assessment, the voltage signal of the power system needs to be obtained first. The fluctuation, stability and any abnormal changes of the voltage signal directly reflect the health status or possible problems of the power system. Therefore, accurate and timely acquisition of voltage signals is the basis for subsequent power quality assessment.

[0070] In some specific practical ways, high-precision voltage sensors can be installed at key nodes or lines of the power system. These sensors can sense and convert voltage signals into processable electrical signals or digital signals in real time. The selection of sensors should take into account their measurement range, accuracy, response speed and anti-interference ability to ensure that subtle changes in voltage signals can be accurately captured.

[0071] S102: The disturbance type and degree of disturbance corresponding to the voltage signal are obtained by inputting the voltage signal into the disturbance classification model. The disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception. The disturbance types include at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation.

[0072] According to the IEC61000 standard, the IEEE1159-2009 standard and the relevant power quality standards formulated by my country, the disturbance type may include at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, oscillatory transient, harmonic, interharmonic, voltage flicker, voltage notching and frequency variation. This application does not limit the specific disturbance type.

[0073] The degree of disturbance can be quantified by voltage deviation (the difference or percentage between the actual voltage and the rated voltage, which reflects the stability and accuracy of the voltage), voltage fluctuation rate (the maximum percentage of the voltage deviation from its average value over a period of time, which is used to measure the degree of voltage fluctuation), harmonic content (the proportion of harmonic components in the voltage signal, which will cause waveform distortion and affect the normal operation of power equipment) and disturbance duration (the length of time that the voltage signal deviates from the normal range, and long-term disturbances may cause more serious damage to the power grid and equipment). This application does not limit the specific degree of disturbance.

[0074] After obtaining the voltage signal, the voltage signal can be input into a pre-trained disturbance classification model to obtain the disturbance type and disturbance degree corresponding to the voltage signal output by the disturbance classification model. In the power quality assessment method disclosed in the embodiment of the present application, the model structure of the above-mentioned disturbance classification model can be a fully linear KAN (Fully Linear KAN Network with Multi-Scale Feature Awareness, KAN Mixer) network model based on multi-scale feature perception. The KAN model combines the advantages of multi-scale feature perception and full linear network, and can be used to parse complex power system voltage signals and identify the disturbance type and disturbance degree corresponding to the voltage signal.

[0075] Specifically, first, after the voltage signal is input into the disturbance classification model, the disturbance classification model will use the multi-scale feature perception mechanism to perform multi-scale decomposition of the voltage signal using N sampling rates. Each decomposed scale corresponds to an independent feature extraction path, thereby obtaining the original features of the voltage signal at N time scales. This design allows the disturbance classification model to focus on the detailed trends and global trends of a voltage signal at different time scales at the same time, thereby more comprehensively understanding the characteristics of the voltage signal.

[0076] In addition, while maintaining efficient feature extraction capabilities, the perturbation classification model also uses a fully linear network structure design to simplify computational complexity, thereby improving computational efficiency. Specifically, the original features on N time scales will be linearly transformed through the fully linear network structure of the perturbation classification model. Subsequently, the features after linear transformation will be further extracted to obtain deep features on N time scales.

[0077] Subsequently, the disturbance classification model concatenates all the deep feature rows on N time scales to form a high-dimensional feature vector. This high-dimensional feature vector contains comprehensive information of the input voltage signal on different time scales, providing strong support for subsequent disturbance classification.

[0078] Finally, the disturbance classification model classifies the high-dimensional feature vector through a classifier composed of a linear layer, an activation function, and a dropout mechanism to obtain the disturbance type and degree of disturbance corresponding to the voltage signal. It should be noted that the design of the classifier fully considers the compression of feature dimensions and the effective use of information, ensuring that the disturbance classification model has a certain ability to resist overfitting while maintaining high classification performance.

[0079] It should be noted that before the voltage signal is input into the disturbance classification model, some preprocessing steps such as denoising, normalization, and resampling can be performed to ensure that the disturbance classification model can process clean and consistent data input.

[0080] S103: Perform power quality assessment on the voltage signal according to the disturbance type and disturbance degree.

[0081] In some specific implementations, if the evaluation result of the power quality evaluation of the voltage signal is lower than the preset quality threshold, a warning message is triggered. It should be noted that the quality threshold is set according to factors such as the operation requirements of the power system, the tolerance of the equipment, and the needs of users, and is used to determine whether the current power quality meets the requirements. If the evaluation result is lower than the preset quality threshold, it means that there is a problem with the current power quality, which may affect the stable operation of the power grid or the normal use of the user. At this time, a warning message will be triggered to notify relevant personnel to pay attention and take corresponding measures. After receiving the warning message, relevant personnel need to formulate corresponding countermeasures according to the specific situation. This may include adjusting the operation parameters of the power grid, repairing or replacing faulty equipment, and notifying users.

[0082] It should be noted that in the power quality assessment method provided in the embodiment of the present application, real-time monitoring needs to be implemented so as to timely discover and handle voltage disturbance problems in the power grid. In addition, during the assessment process, voltage signals from different monitoring points should be fully utilized to improve the accuracy and reliability of power quality assessment.

[0083] Next, the above perturbation classification model is explained in more detail:

[0084] See also Figure 2 , which is a schematic diagram of the structure of a disturbance classification model provided in an embodiment of the present application. Figure 2 It can be seen that the disturbance classification model 200 includes a KANScanner module 201 (linear transformation module) and a KANBlock module 202 (multi-layer embedding module), and the output end of the KANScanner module 201 is connected to the input end of the KANBlock module 202. It should be noted that the number of KANScanner modules 201 and KANBlock modules 202 is equal, and the specific number is not limited in this application.

[0085] The KANScanner module 201 is a module designed for sequence data. It uses a multi-scale feature perception mechanism and N different sampling rates to perform multi-scale decomposition on the input sequence data (i.e., voltage signal) to construct original features on N time scales. Each original feature represents the performance of the voltage signal on different time scales to reveal the multi-faceted characteristics of the voltage signal.

[0086] It is understandable that the sampling rate determines the level of detail at which the voltage signal is sampled. A higher sampling rate can capture subtle changes in the voltage signal, while a lower sampling rate can reflect the overall trend of the voltage signal.

[0087] Specifically, see Figure 3, which is a schematic diagram of a KANScanner module provided in an embodiment of the present application. After the KANScanner module 201 decomposes the voltage signal at multiple scales using multiple sampling rates, a signal sequence on N time scales is first obtained. Next, the KANScanner module 201 divides the signal sequence on each time scale into multiple equal-length and non-overlapping signal windows. These windows serve as basic units for subsequent processing and help capture local features of the voltage signal. Subsequently, the KANScanner module 201 applies KANLinear (linear layer) to perform feature extraction on each window to obtain the original features of the voltage signal on each time scale, which includes information about the voltage signal on different time scales.

[0088] In some specific implementations, the formula for obtaining the original features may be as shown in the following formula (1):

[0089] Base(x)=W base ×f(x)(1)

[0090] Among them, Base(x) is the original feature, x is the signal sequence, and W is the piecewise polynomial weight.

[0091] KANBlock module 202 is an innovative neural network module design that focuses on efficient feature extraction and compression through layer-by-layer dimensionality reduction, and is particularly suitable for tasks that require processing multi-scale features. This design combines techniques such as piecewise polynomial interpolation, basic linear transformation, and residual connection to improve the representation ability, training stability, and inference efficiency of perturbation classification models.

[0092] Specifically, see Figure 4 , which is a schematic diagram of a KANBlock module provided in an embodiment of the present application. The KANBlock module 202 performs linear transformation processing through multiple layers of KANLinear to achieve layer-by-layer extraction and dimensionality reduction of original features, which helps to gradually remove redundant information while retaining deep features useful for disturbance classification.

[0093] In some specific implementations, the formula for obtaining the deep features can be shown as the following formula (2):

[0094] y=Base(x)+Spline(x) (2)

[0095] Where y is the generated feature, Base(x) is the original feature, and Spline(x) is the interpolation output. The formula for the interpolation output can be shown as follows:

[0096] Spline(x)=∑B i (x)×w i(3)

[0097] Among them, Spline(x) is the interpolation output, Bi(x) is the B-spline basis function, w is the piecewise polynomial weight, and x is the signal sequence.

[0098] It should be noted that KANLinear of each layer can not only perform baseline linear transformation on the original features, but also perform piecewise polynomial interpolation on the original features. Piecewise polynomial interpolation can capture the nonlinear relationship in the original features more finely, thereby improving the representation ability of the perturbation classification model.

[0099] It should be noted that in deep networks, gradient vanishing easily makes the KAN model difficult to train. In the perturbation classification model provided in the embodiment of the present application, by introducing residual connections, the original input features are allowed to directly skip certain KANBlocks and be directly added to the output of subsequent KANBlocks, which can effectively alleviate the gradient vanishing problem and enhance the training stability of the perturbation classification model.

[0100] After the KANBlock module 202 extracts the deep features on N time scales, it can also perform layer normalization (LayerNorm) processing to standardize the deep features of different scales, so as to balance the distribution of deep features of each scale, further improve the robustness and generalization ability of the disturbance classification model, and then splice all deep features to obtain a high-dimensional feature vector. Subsequently, the high-dimensional feature vector is subjected to disturbance classification through the target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal. The target classifier is a classifier including a linear layer, an activation function, and an exit Dropout mechanism.

[0101] It can be understood that using KANLinear as a linear layer has the following advantages: First, in power quality classification, voltage information often contains highly nonlinear features. KANLinear can fit these nonlinear relationships more flexibly by introducing piecewise polynomial interpolation (such as B-spline basis functions). This method not only improves the expressiveness of the disturbance classification model, but also enables the disturbance classification model to more accurately capture subtle differences in the data, thereby improving the accuracy of classification. Second, power quality disturbances are diverse and unpredictable, such as voltage sags, swells, and transient disturbances. KANLinear's dynamic spline interpolation framework can dynamically adjust the interpolation weights according to the characteristics of the input data, thereby making a more flexible and accurate response to different types of disturbances. This adaptability makes the disturbance classification model more robust when facing complex and changeable power quality data. Third, power quality disturbances may contain both high-frequency and low-frequency components, and different disturbance types may have significant differences in frequency and amplitude. KANLinear's interpolation mechanism can simultaneously process features of different scales, ensuring that the disturbance classification model captures high-frequency details while not ignoring low-frequency trends. This multi-scale adaptability is critical for comprehensive analysis of power quality data. Fourth, machine learning models often face the risk of overfitting when processing complex data sets. B-spline interpolation in KANLinear naturally introduces a regularization effect by smoothing the feature mapping process, which helps reduce the disturbance classification model's dependence on noise and non-representative features in the training data. This regularization capability helps improve the generalization performance of the disturbance classification model, enabling it to perform well in different systems and conditions.

[0102] It should be noted that after constructing the perturbation classification model through the above method, multiple evaluation indicators can also be used, including accuracy, precision, recall and F1 score, so as to measure the performance of the perturbation classification model from different angles.

[0103] Among them, accuracy refers to the proportion of samples predicted correctly by the perturbation classification model to the total number of samples. It is the most intuitive evaluation indicator used to measure the overall correctness of the model. The calculation formula of accuracy can be shown as follows (4):

[0104]

[0105] Where N is the total number of samples, is the predicted label of the perturbation classification model for the i-th sample, y i is the true label of the i-th sample, is an indicator function that takes the value 1 when the predicted label is the same as the true label, otherwise it takes the value 0.

[0106] Precision refers to the proportion of samples predicted as positive by the perturbation classification model that are actually positive. It reflects the proportion of samples predicted as positive by the perturbation classification model that are correctly predicted. The calculation formula for precision can be shown as follows (5):

[0107]

[0108] Among them, C is the total number of categories, that is, the number of different categories in the multi-classification task, TPi is the number of true positive samples in the i-th category, and FPi is the number of false positive samples in the i-th category.

[0109] Recall rate refers to the proportion of samples that are actually positive that are correctly predicted as positive by the perturbation classification model. It reflects the proportion of samples that are correctly predicted by the perturbation classification model among all actual positive samples. The calculation formula of recall rate can be shown as follows (6):

[0110]

[0111] Where C is the total number of categories, TP i is the number of true positive samples of the i-th class, FP i is the number of false positive samples of the i-th class, FN i is the number of false negative samples in the i-th class.

[0112] The F1 score is the harmonic mean of precision and recall, which combines the two indicators. It balances the precision and recall of the perturbation classification model when evaluating its performance. The calculation formula of the F1 score can be shown in the following formula (7):

[0113]

[0114] Among them, C is the total number of categories, Precision i is the precision of the i-th class, Recall i is the recall rate of the i-th class.

[0115] In summary, the present application discloses a method for evaluating power quality, which includes: obtaining a voltage signal of a power system; obtaining the disturbance type and disturbance degree corresponding to the voltage signal by inputting the voltage signal into a disturbance classification model, wherein the disturbance classification model adopts a fully linear KAN network model based on multi-scale feature perception, and the disturbance types include at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage gap and frequency deviation; and evaluating the power quality of the voltage signal according to the disturbance type and disturbance degree. Thus, the voltage signal of the power system is input into the disturbance classification model constructed by the KAN network model, which can capture the characteristics of the voltage signal at different time scales and frequencies, identify the disturbance type and disturbance degree of the voltage signal, and thus improve the comprehensiveness and accuracy of the power quality evaluation.

[0116] See also Figure 5 , which is a schematic diagram of a power quality assessment device provided in an embodiment of the present application. The power quality assessment device 500 includes: a signal acquisition module 501, a disturbance acquisition module 502 and a quality assessment module 503.

[0117] The signal acquisition module 501 is used to acquire a voltage signal of the power system.

[0118] The disturbance acquisition module 502 is used to obtain the disturbance type and disturbance degree corresponding to the voltage signal by inputting the voltage signal into the disturbance classification model. The disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception. The disturbance types include at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation.

[0119] The quality assessment module 503 is used to perform power quality assessment on the voltage signal according to the disturbance type and disturbance degree.

[0120] In some specific implementations, the disturbance acquisition module 502 includes: a decomposition submodule, a transformation submodule, a splicing submodule and a classification submodule; the decomposition submodule is used to perform multi-scale decomposition of the voltage signal using N sampling rates through the multi-scale feature perception mechanism of the disturbance classification model to obtain the original features of the voltage signal on N time scales, where N is a natural number; the transformation submodule is used to perform linear transformation processing on the original features on N time scales through the full linear network structure of the disturbance classification model to obtain deep features on N time scales; the splicing submodule is used to splice the deep features on all N time scales to obtain a high-dimensional feature vector; the classification submodule is used to perform disturbance classification on the high-dimensional feature vector through a target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal, and the target classifier is a classifier including a linear layer, an activation function and an exit Dropout mechanism.

[0121] In some specific implementations, the decomposition submodule includes: a first decomposition submodule, a second decomposition submodule and a third decomposition submodule; the first decomposition submodule is used to perform multi-scale decomposition of the voltage signal using N sampling rates to obtain signal sequences on N time scales; the second decomposition submodule is used to divide the signal sequence on each time scale into equal-length and non-overlapping signal windows; the third decomposition submodule is used to perform feature extraction on each signal window using a linear layer to obtain the original features of the voltage signal on N time scales.

[0122] In some specific implementations, the transformation submodule is specifically used to: use a linear layer to perform linear transformation processing and piecewise polynomial interpolation processing on the original features on N time scales, so as to obtain deep features on N time scales.

[0123] In some specific implementations, the power quality assessment device 500 further includes: a warning triggering module; the warning triggering module is used to trigger a warning message if the assessment result of the power quality assessment is lower than the quality threshold.

[0124] In summary, the present application discloses a power quality assessment device, which inputs the voltage signal of the power system into a disturbance classification model constructed by a KAN network model, can capture the characteristics of the voltage signal at different time scales and frequencies, and identify the disturbance type and degree of the voltage signal, thereby improving the comprehensiveness and accuracy of the power quality assessment.

[0125] The embodiments of the present application also provide corresponding power quality assessment equipment and computer storage media for implementing the solutions provided in the embodiments of the present application.

[0126] The power quality assessment device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute instructions or codes so that the device executes a power quality assessment method of any embodiment of the present application.

[0127] The computer storage medium stores codes, and when the codes are executed, the device executing the codes implements the power quality assessment method described in any embodiment of the present application.

[0128] The "first" and "second" in the names such as "first" and "second" (if any) mentioned in the embodiments of the present application are only used as name identifiers and do not represent the first or second in order.

[0129] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.

[0130] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0131] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for evaluating power quality, characterized in that: The method comprises: Obtain voltage signals of power systems; The disturbance type and disturbance degree corresponding to the voltage signal are obtained by inputting the voltage signal into a disturbance classification model, wherein the disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception, and the disturbance type includes at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation; Performing power quality assessment on the voltage signal according to the disturbance type and the disturbance degree.

2. The method according to claim 1, characterized in that The obtaining of the disturbance type and disturbance degree corresponding to the voltage signal includes: Through the multi-scale feature perception mechanism of the disturbance classification model, the voltage signal is multi-scale decomposed using N sampling rates to obtain the original features of the voltage signal on N time scales, where N is a natural number; Through the full linear network structure of the disturbance classification model, the original features on the N time scales are respectively subjected to linear transformation processing to obtain deep features on the N time scales; Concatenate all the deep features on the N time scales to obtain a high-dimensional feature vector; The high-dimensional feature vector is subjected to disturbance classification through a target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal. The target classifier is a classifier including a linear layer, an activation function and an exit Dropout mechanism.

3. The method according to claim 2, characterized in that The multi-scale decomposition of the voltage signal using N sampling rates to obtain original features of the voltage signal on N time scales includes: Performing multi-scale decomposition on the voltage signal using N sampling rates to obtain signal sequences on N time scales; The signal sequence at each time scale is divided into equal-length and non-overlapping signal windows; A linear layer is used to perform feature extraction on each of the signal windows to obtain original features of the voltage signal on the N time scales.

4. The method according to claim 2, characterized in that: The linear transformation processing is performed on the original features on the N time scales to obtain the deep features on the N time scales, including: The linear layer is used to perform linear transformation processing and piecewise polynomial interpolation processing on the original features on the N time scales to obtain deep features on the N time scales.

5. The method according to claim 1, characterized in that: The method further comprises: If the evaluation result of the power quality evaluation is lower than the quality threshold, a warning message is triggered.

6. A power quality assessment device, characterized in that: The device comprises: a signal acquisition module, a disturbance acquisition module and a quality assessment module; The signal acquisition module is used to acquire the voltage signal of the power system; The disturbance acquisition module is used to obtain the disturbance type and disturbance degree corresponding to the voltage signal by inputting the voltage signal into a disturbance classification model, wherein the disturbance classification model adopts a full linear KAN network model based on multi-scale feature perception, and the disturbance type includes at least one of voltage sag, voltage swell, voltage interruption, undervoltage, overvoltage, transient pulse, transient oscillation, harmonic, interharmonic, voltage flicker, voltage notch and frequency deviation; The quality assessment module is used to perform power quality assessment on the voltage signal according to the disturbance type and the disturbance degree.

7. The device according to claim 6, characterized in that The disturbance acquisition module includes: a decomposition submodule, a transformation submodule, a splicing submodule and a classification submodule; The decomposition submodule is used to perform multi-scale decomposition on the voltage signal using N sampling rates through the multi-scale feature perception mechanism of the disturbance classification model to obtain original features of the voltage signal on N time scales, where N is a natural number; The transformation submodule is used to perform linear transformation processing on the original features on the N time scales respectively through the full linear network structure of the disturbance classification model to obtain deep features on the N time scales; The splicing submodule is used to splice the deep features on all the N time scales to obtain a high-dimensional feature vector; The classification submodule is used to perform disturbance classification on the high-dimensional feature vector through a target classifier to obtain the disturbance type and disturbance degree corresponding to the voltage signal. The target classifier is a classifier including a linear layer, an activation function and an exit Dropout mechanism.

8. The device according to claim 7, characterized in that The decomposition submodule includes: a first decomposition submodule, a second decomposition submodule and a third decomposition submodule; The first decomposition submodule is used to perform multi-scale decomposition on the voltage signal using N sampling rates to obtain signal sequences on N time scales; The second decomposition submodule is used to divide the signal sequence on each time scale into equal-length and non-overlapping signal windows; The third decomposition submodule is used to extract features from each of the signal windows using a linear layer to obtain original features of the voltage signal on the N time scales.

9. The device according to claim 7, characterized in that The transformation submodule is specifically used to: use a linear layer to perform linear transformation processing and piecewise polynomial interpolation processing on the original features on the N time scales, so as to obtain deep features on the N time scales.

10. The device according to claim 6, characterized in that The device also includes: a warning triggering module; The warning triggering module is used to trigger a warning message if the evaluation result of the power quality evaluation is lower than a quality threshold.