Circuit breaker data optimization method and system based on process feedback

Abstract

The invention discloses a circuit breaker data optimization method and system based on process feedback, and belongs to the technical field of circuit breaker data processing, and the method comprises the steps: carrying out the wavelet transformation of obtained circuit breaker data, and extracting a first error feature of the circuit breaker data; taking the optimal noise reduction effect of the circuit breaker data as a target, carrying out the iterative adjustment of a noise reduction threshold value, and obtaining an optimal noise reduction threshold value and a second error feature corresponding to the optimal noise reduction threshold value; dimensionality reduction and fusion are carried out on the second error feature to obtain a comprehensive feature vector; and introducing the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization to obtain circuit breaker optimization data, so that by implementing the method, the problems that the noise reduction process and the error feature extraction process are independently carried out, information cannot be fed back mutually and the noise reduction efficiency is high in the prior art can be solved. The noise reduction effect of the circuit breaker data is not good and the error cannot be accurately corrected.

Description

Technical Field

[0001] The present application belongs to the technical field of circuit breaker data processing, and specifically relates to a circuit breaker data optimization method and system based on process feedback. Background Art

[0002] Circuit breakers play a vital role in power systems. Their primary mission is to control and protect circuits, ensuring the safe and stable operation of the power grid. As power systems continue to evolve toward intelligent and automated systems, monitoring and analyzing circuit breaker operating status has become increasingly critical. Accurately capturing circuit breaker operating data and effectively processing and analyzing it is fundamental to ensuring power grid reliability and stability. This involves multiple technical areas, including data processing, signal analysis, and machine learning, all aimed at better managing and maintaining circuit breaker operations through technological means.

[0003] Existing optimization of circuit breaker data involves data denoising and error feature extraction, which are often performed independently. Data is decomposed using wavelet transforms, and then the decomposed data is denoised by calculating a separate denoising threshold. Error feature extraction is then performed on the denoised result to correct the error points. Due to insufficient attention to the interaction and feedback between the error feature extraction and denoising processes, it is impossible to flexibly adjust the noise reduction requirements based on the actual data conditions, resulting in unsatisfactory noise reduction effects. The resulting optimized circuit breaker data is of poor quality, making it difficult to accurately determine the operating status of the circuit breaker, and thus limiting performance during fault diagnosis and anomaly detection. Furthermore, the independent denoising and error feature extraction processes result in low data processing efficiency. It is difficult to fully utilize the information from the denoising process during error feature extraction, which can easily lead to incomplete or inaccurate error feature extraction, making it impossible to accurately correct the error points in the circuit breaker data. Summary of the Invention

[0004] This application proposes a circuit breaker data optimization method and system based on process feedback, which can solve the problem in the prior art that the noise reduction process and error feature extraction process are carried out independently and cannot feedback information to each other, resulting in poor noise reduction effect of circuit breaker data and inability to accurately correct errors.

[0005] A first aspect of the present application provides a circuit breaker data optimization method based on process feedback, the method comprising:

[0006] Performing a wavelet transform on the obtained circuit breaker data, and extracting a first error feature of the circuit breaker data during the wavelet transform process;

[0007] With the goal of achieving optimal noise reduction effect on the circuit breaker data, the noise reduction threshold is iteratively adjusted using the first error feature as a feedback signal to obtain an optimal noise reduction threshold and a second error feature corresponding to the optimal noise reduction threshold;

[0008] Performing dimensionality reduction and fusion on the second error features in sequence to obtain a comprehensive feature vector;

[0009] The comprehensive feature vector and the optimal noise reduction threshold are introduced into a preset data optimization model for model optimization. The circuit breaker data is subjected to noise reduction and error correction by the optimized data optimization model to obtain circuit breaker optimized data.

[0010] The above scheme first performs a wavelet transform on collected circuit breaker operating data, electrical data, and other data, extracting characteristic information at different frequency bands. Based on this information, the error characteristics of the circuit breaker data are then derived. The error characteristics are then used to dynamically adjust the noise reduction threshold. A new error characteristic is then derived based on the adjusted noise reduction threshold. This allows the results of the noise reduction process and error characteristic extraction to interact, forming a dynamic feedback mechanism. By learning the relationship between the error characteristics and the noise reduction threshold, the optimal noise reduction threshold is determined for optimal noise reduction, thereby optimizing the noise reduction effect. This approach achieves accurate error characteristics while eliminating excess noise and enhancing error characteristic extraction. Because enhanced error characteristic extraction generates a large number of redundant features, which can affect subsequent processing and analysis, feature dimensionality reduction is required to reduce the dimensionality of the feature space. Feature fusion is then performed after dimensionality reduction to improve overall processing efficiency. Finally, the resulting comprehensive feature vector is introduced into a data optimization model to optimize model performance and guide the model towards more stable and reasonable predictions. Ultimately, circuit breaker data with precise error correction and improved noise reduction is obtained.

[0011] In a possible implementation method of the first aspect, wavelet transform is performed on the obtained circuit breaker data, and a first error feature of the circuit breaker data is extracted during the wavelet transform, specifically:

[0012] The circuit breaker data is decomposed into several frequency segments through wavelet transform, and the circuit breaker data in each frequency segment is analyzed respectively to extract the first error feature; wherein the first error feature includes data mean square error, data variance, data skewness and data kurtosis.

[0013] This approach partially integrates the error feature extraction process within the wavelet transform process, directly extracting error features using the resulting wavelet transform coefficients. By decomposing the circuit breaker data into multiple frequency bands through the wavelet transform, it facilitates detailed local analysis. By analyzing the details of the data at different scales and locations, potential error features can be identified and accurately extracted.

[0014] In a possible implementation method of the first aspect, with the goal of achieving an optimal noise reduction effect on the circuit breaker data, the first error feature is used as a feedback signal to iteratively adjust the noise reduction threshold to obtain an optimal noise reduction threshold and a second error feature corresponding to the optimal noise reduction threshold, specifically:

[0015] In the noise reduction module of the data optimization model, noise reduction is performed on a preset training set using the current noise reduction threshold, and then the noise reduction effect is evaluated using the first error feature; wherein the training set includes historical data of circuit breakers with different noise levels, different signal types, and different actual operating conditions;

[0016] If the noise reduction effect satisfies the preset optimal noise reduction effect, outputting the current noise reduction threshold as the optimal noise reduction threshold and the second error feature;

[0017] If the noise reduction effect does not meet the preset optimal noise reduction effect, the noise reduction threshold is adjusted according to the noise reduction effect, and the first error feature is updated by the adjusted noise reduction threshold, and then the training set is denoised and evaluated again using the adjusted noise reduction threshold and the updated first error feature.

[0018] The above scheme dynamically adjusts the noise reduction threshold using the first error feature, and then uses the adjusted threshold to obtain a more accurate error feature. This allows the results of the noise reduction process and error feature extraction to feed back into each other, thereby optimizing the noise reduction effect and obtaining the optimal noise reduction threshold for the best noise reduction effect, thereby eliminating as much excess noise as possible. Furthermore, the noise reduction effect corresponding to the noise reduction threshold is evaluated using the error feature, providing a basis for adjusting the noise reduction threshold. The optimized noise reduction threshold can extract more accurate and effective error features, thus eliminating excess noise during the error feature extraction process.

[0019] In a possible implementation method of the first aspect, the second error feature is specifically:

[0020] After obtaining the optimal noise reduction threshold, the first error feature is updated using the optimal noise reduction threshold to obtain the second error feature.

[0021] In a possible implementation method of the first aspect, dimension reduction and fusion are performed on the second error features in sequence to obtain a comprehensive feature vector, specifically:

[0022] Performing dimensionality reduction on the second error feature through linear transformation to obtain several representative features;

[0023] The representative features are weighted averaged and concatenated to obtain a comprehensive feature vector.

[0024] The aforementioned solution generates a large number of redundant features during the enhanced error feature extraction process. These redundant features can affect subsequent processing and analysis, resulting in lower efficiency in subsequent data analysis and processing. Therefore, feature dimensionality reduction reduces redundant information while retaining representative features, improving overall data processing efficiency. Furthermore, fusing features from different sources provides more information for subsequent error analysis and fault diagnosis.

[0025] In a possible implementation method of the first aspect, the comprehensive feature vector and the optimal noise reduction threshold are introduced into a preset data optimization model for model optimization. Noise reduction and error correction are performed on the circuit breaker data using the optimized data optimization model to obtain circuit breaker optimized data, specifically:

[0026] Introducing the comprehensive feature vector into the loss function of the data optimization model, and introducing the optimal noise reduction threshold into the noise reduction module of the data optimization model to obtain the optimized data optimization model;

[0027] The circuit breaker data is denoised using the optimized data optimization model, and errors are identified and corrected on the denoising result according to the loss function containing the comprehensive feature vector to obtain circuit breaker optimization data.

[0028] The above solution optimizes model performance by introducing the comprehensive feature vector into the loss function of the data optimization model, prevents overfitting of the model prediction results, makes the data output by the model more stable and reasonable, and obtains more accurate and error-corrected circuit breaker data.

[0029] In a possible implementation method of the first aspect, the comprehensive feature vector is introduced into the loss function of the data optimization model, specifically:

[0030] The data mean square error of the comprehensive feature vector is introduced into the loss function as a regularization term to obtain the expanded loss function.

[0031] In a possible implementation method of the first aspect, the expanded loss function is specifically expressed as follows:

[0032]

[0033] Where L(θ) is the expanded loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter, ο is the threshold of tolerance error, n is the total amount of data, is the regularization term, λ is the adjustment parameter, and MSE is the mean square error of the data.

[0034] A second aspect of the present application provides a circuit breaker data optimization system based on process feedback, the system comprising: an equivalent magnetic permeability calculation module, a time domain equation construction module, a compensation current calculation module, and an error correction module;

[0035] The equivalent magnetic permeability calculation module is used to calculate the equivalent magnetic permeability of the iron core of the current transformer after the air gap is opened according to the hardware information of the iron core of the current transformer;

[0036] The time domain equation construction module is used to construct a time domain equation of the current transformer with an air gap according to the equivalent magnetic permeability of the iron core of the current transformer after the air gap is opened and the operating data of the secondary circuit of the current transformer with an air gap;

[0037] The compensation current calculation module is used to integrate the time domain equation of the gapped current transformer according to a preset Cotes formula to obtain data points of the primary current of the gapped current transformer within one cycle, and obtain the compensation current based on the data points; wherein the compensation current is the value of the primary current of the gapped current transformer after error correction;

[0038] The error correction module is used to correct the error generated by the current transformer with an air gap according to the compensation current.

[0039] A third aspect of the present application provides a terminal device, comprising: a terminal device including a processor and a memory, the memory storing a computer program, and the processor implementing the steps of a circuit breaker data optimization method based on process feedback as described in any one of the embodiments of the present application when executing the computer program. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 This is a specific flow chart of a circuit breaker data optimization method based on process feedback provided in a certain embodiment of the present application;

[0042] Figure 2 This is a specific structural diagram of a circuit breaker data optimization system based on process feedback provided in one embodiment of the present application;

[0043] Figure 3 A structural diagram of a terminal device is provided for a certain embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0045] It should be understood that the step numbers used herein are only for convenience of description and are not intended to limit the order in which the steps are to be executed.

[0046] First embodiment

[0047] Directly collected circuit breaker data requires noise reduction and error correction to be more effectively used for transmission line fault diagnosis and anomaly detection. However, the data noise reduction process and error feature extraction process are generally performed independently, resulting in suboptimal noise reduction results. Furthermore, because the noise reduction process cannot provide positive feedback for error correction, it is difficult to fully utilize the information from the noise reduction process during feature extraction, which can easily lead to incomplete or inaccurate feature extraction. Therefore, the main research direction of this application is to establish a dynamic feedback mechanism to mutually optimize the data noise reduction process and error feature extraction process based on their output results, thereby further improving the noise reduction effect and the accuracy of error features.

[0048] like Figure 1 As shown, in order to solve the problem in the prior art that the noise reduction process and the error feature extraction process are performed independently and cannot feedback information to each other, resulting in poor noise reduction effect of circuit breaker data and inability to accurately correct errors, the first embodiment of the present application provides a specific flow chart of a circuit breaker data optimization method based on process feedback. The circuit breaker data optimization method based on process feedback of this embodiment includes steps S1 to S4, which are detailed as follows:

[0049] Step S1 : performing wavelet transform on the obtained circuit breaker data, and extracting a first error feature of the circuit breaker data during the wavelet transform process.

[0050] In this embodiment, a high-precision sensor with excellent anti-interference capabilities collects electrical data and device status data from the circuit breaker during operation in real time at a high sampling frequency. The electrical data includes current, voltage, and power data for the branch circuit in which the circuit breaker is located, and the device status data includes the circuit breaker's on / off state and actuation time during operation.

[0051] Conventional technology typically performs noise reduction on circuit breaker data before extracting error characteristics from the noise reduction results. However, in this embodiment, in order to dynamically adjust the noise reduction threshold using feedback from the error characteristics, the circuit breaker data is first subjected to a wavelet transform. Error characteristics are then extracted from the wavelet transform results. The noise reduction effect is then evaluated using the error characteristics, providing data support for subsequently determining the optimal noise reduction threshold.

[0052] In an embodiment of the present application, the original operation data of the circuit breaker collected by the high-precision sensor is also preprocessed, including wavelet denoising, outlier removal and time series reconstruction.

[0053] The wavelet transform is a transform analysis method that performs time-frequency analysis and processing of signals by providing a frequency-dependent "time-frequency" window. Using scaling and translation operations, the signal is gradually refined at multiple scales, ultimately achieving time segmentation at high frequencies and frequency segmentation at low frequencies. It automatically adapts to the requirements of time-frequency signal analysis, allowing it to focus on any detail in the signal and highlight its characteristics. The number of decomposition layers used in the wavelet transform is generally determined by the complexity and volatility of the data, typically ranging from three to five. The greater the number of layers, the finer the decomposed frequency bands, enabling it to capture more detailed signal features.

[0054] First, the collected circuit breaker data is decomposed into multiple scales using wavelet transform. This is done to extract characteristic information of the circuit breaker data at different frequency bands, facilitating more detailed local analysis. This means that error features are extracted during the wavelet transform process. By analyzing the details of the circuit breaker data at different scales and locations, potential error features are identified and extracted. The specific formula is:

[0055]

[0056] Where W ψ (s, t) is the wavelet coefficient obtained after wavelet transform, which represents the local characteristics of the data at scale s and position t; ψ is the wavelet basis function, f(x) is the circuit breaker data, s is the scale parameter, and t is the translation parameter.

[0057] Based on the wavelet coefficients obtained across all frequency bands after wavelet transformation, relevant error features such as data mean square error, data variance, data skewness, and data kurtosis are directly extracted. These error features can provide feedback signals for the data denoising process and provide a basis for dynamic adjustment of the denoising threshold.

[0058] The specific expression of the data mean square error is:

[0059]

[0060] In the formula, MSE is the mean square error of the data, xi is the i-th data point of the circuit breaker data, is the i-th data point of the circuit breaker data after wavelet transformation, and N is the total number of data points.

[0061] The data variance is specifically expressed as:

[0062]

[0063] Where σ 2 is the data variance, μ is the mean of the circuit breaker data, x i is the i-th data point of the circuit breaker data;

[0064] The data skewness is specifically expressed as:

[0065]

[0066] Where Skewness is the data skewness, σ is the data standard deviation, and N is the total number of data points;

[0067] The data kurtosis is specifically expressed as:

[0068]

[0069] Wherein, Kurtosis is the kurtosis of the data.

[0070] The above error characteristics can be used as input to the data optimization model to learn the nonlinear relationship between the noise reduction process and the noise reduction threshold and optimize the data reconstruction.

[0071] Step S2: with the goal of achieving the optimal noise reduction effect of the circuit breaker data, the noise reduction threshold is iteratively adjusted using the first error feature as a feedback signal to obtain the optimal noise reduction threshold and the second error feature corresponding to the optimal noise reduction threshold.

[0072] In the traditional circuit breaker data optimization process, data denoising and error feature extraction are usually two independent steps. However, in this application, the error feature extraction and data denoising results interact with each other to form a dynamic feedback mechanism. The data denoising results are evaluated through error features, and new error features are extracted from the denoised data.

[0073] The noise reduction threshold is the core of the data denoising process. This embodiment combines dynamic threshold calculation with a machine learning algorithm to achieve the optimal noise reduction threshold for the best noise reduction effect. Compared to the embodiments of this application, traditional noise reduction threshold calculation methods are based on fixed formulas or manually set parameters. By introducing machine learning to automatically optimize the threshold based on feedback from error characteristics, this significantly improves the noise reduction effect.

[0074] On the basis of step S1, the obtained error features are used as input to train the denoising module of the preset data optimization model, and the denoising module is trained. The data optimization model is generally based on SVR or neural network. SVR stands for Support Vector Regression, abbreviated as support vector regression, which is a branch of support vector machine.

[0075] Through a preset training set, the noise reduction module automatically derives the optimal noise reduction threshold that can achieve the best noise reduction effect by learning the relationship between the error characteristics and the noise reduction threshold.

[0076] As an improvement to the above solution, the embodiment of the present application uses historical data of circuit breakers containing different noise levels, different signal types and different actual working conditions as a training set. The optimal noise reduction threshold obtained through this training set can be applied to different data sets, effectively expanding the scope of application.

[0077] Specifically, the preset training set is denoised using the current denoising threshold, and then the denoising effect is evaluated by calculating the error feature vector. The current denoising threshold is adjusted according to the evaluation result, and the data is denoised again using the adjusted denoising threshold. New error features are extracted from the denoising results to update the error features. The denoising threshold is then adjusted again according to the updated error features until the denoising effect meets the preset optimal denoising effect. The current denoising threshold is then output as the optimal denoising threshold, and the latest error features obtained by the optimal denoising threshold are also output.

[0078] Exemplarily, the mean square error of the data is used to evaluate the noise reduction effect. If the calculated mean square error of the data is high, the noise reduction threshold is adjusted until the mean square error of the data falls into an acceptable normal range, so that effective features can be better extracted from the data.

[0079] The adjustment of the noise reduction threshold is based on the feedback of the error characteristics, so that data noise reduction and error feature extraction can be optimized simultaneously and interact with each other in the same process to form a more efficient processing process, achieving feature extraction enhancement while obtaining the optimal noise reduction threshold.

[0080] Step S3: perform dimensionality reduction and fusion on the second error features in sequence to obtain a comprehensive feature vector.

[0081] In the embodiment of the present application, because feature extraction enhancement will generate a large number of redundant features, the redundant features will not only affect the accuracy of subsequent data processing and analysis, but also greatly reduce the data processing efficiency. Therefore, it is necessary to reduce the dimension of the updated error features. By reducing the dimension of the feature space through feature dimensionality reduction, important and representative feature information is retained.

[0082] Optionally, the embodiment of the present application uses PCA and LDA techniques to achieve feature dimensionality reduction in order to select the most representative error features. PCA technology is Principal Components Analysis, which aims to use the idea of ​​dimensionality reduction to transform multiple indicators into a few comprehensive indicators, thereby achieving data dimensionality reduction, denoising and visualization; LDA technology is Linear Discriminant Analysis, which uses statistics, pattern recognition and machine learning methods to try to find a linear combination of features of two types of objects or events in order to characterize or distinguish them. The resulting combination can be used as a linear classifier, or, more commonly, to perform dimensionality reduction processing for subsequent classification.

[0083] After dimensionality reduction, the representative features obtained after dimensionality reduction are fused with other data through multimodal fusion to improve the overall processing efficiency.

[0084] Specifically, representative features are combined with multimodal data to fuse data features from different sources. For example, multiple data such as current, voltage, power, and frequency are weighted averaged and concatenated to generate a comprehensive feature vector. This comprehensive feature vector can provide more information for subsequent error analysis and fault diagnosis.

[0085] Step S4: introducing the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization, and performing noise reduction and error correction on the circuit breaker data using the optimized data optimization model to obtain circuit breaker optimized data.

[0086] In order to improve the accuracy of data processing, the embodiment of the present application optimizes the data optimization model by introducing the comprehensive feature vector as a regularization term into the loss function of the data optimization model, so that the model not only focuses on the fitting accuracy during the optimization process, but also takes into account the smoothness and interpretability of the error.

[0087] In the embodiment of the present application, the data optimization model performs data prediction based on support vector regression, and the loss function of support vector regression is as follows:

[0088]

[0089] Where L(θ) is the loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter used to control the tolerance of the model to errors, ο is the threshold of tolerance error, n is the total amount of data, is a regularization term used to prevent overfitting.

[0090] After the introduction of the comprehensive feature vector, the expanded loss function is specifically expressed as follows:

[0091]

[0092] Where L(θ) is the expanded loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter, ο is the threshold of tolerance error, n is the total amount of data, is the regularization term, λ is the adjustment parameter, and MSE is the mean square error of the data.

[0093] By optimizing the data optimization model through comprehensive feature vectors, not only the fitting ability of the model is optimized, but also the model is guided toward a more stable and reasonable prediction direction through error characteristics.

[0094] In addition, the data optimization model also uses the obtained optimal noise reduction threshold to reduce noise on the input data to achieve the optimal noise reduction effect.

[0095] In summary, by introducing the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization, noise reduction, error identification and correction of circuit breaker data are achieved, and more accurate circuit breaker optimization data is obtained, which helps to improve the accuracy of fault diagnosis and anomaly detection of transmission lines.

[0096] As an improvement to the above solution, the circuit breaker optimization data output by the model is transmitted to the control execution unit through a communication interface. The control execution unit will perform logical judgment and response strategy matching on the circuit breaker optimization data. For example, when the circuit breaker optimization data is identified as "minor abnormality", the control execution unit will activate the local load reduction control logic to temporarily cut off some non-critical loads on the branch where the circuit breaker is located; when the circuit breaker optimization data is identified as "imminent failure", the control execution unit will link the upstream distribution protection unit to issue a warning signal and prepare to cut off the main circuit; when the circuit breaker optimization data is identified as "failed", the control execution unit will immediately issue a trip command and record the circuit breaker behavior parameters for subsequent maintenance analysis.

[0097] The implementation of the embodiments of the present application has the following beneficial effects:

[0098] In this embodiment of the present application, a wavelet transform is first performed on the collected circuit breaker operation data, electrical data, etc., to extract characteristic information of the circuit breaker data in different frequency bands, and based on this information, the error characteristics of the circuit breaker data are further obtained. The error characteristics are then used to dynamically adjust the noise reduction threshold, and a new error characteristic is obtained based on the adjusted noise reduction threshold. The results of the noise reduction process and error feature extraction are mutually influenced, forming a dynamic feedback mechanism. By learning the relationship between the error characteristics and the noise reduction threshold, the optimal noise reduction threshold with the best noise reduction effect is obtained, thereby optimizing the noise reduction effect, achieving accurate error characteristics while eliminating excess noise and enhancing error feature extraction. Because the extraction of enhanced error features generates a large number of redundant features, which affect subsequent processing and analysis, it is necessary to reduce the dimensionality of the feature space through feature dimensionality reduction, and perform feature fusion after dimensionality reduction to improve overall processing efficiency. Finally, the obtained comprehensive feature vector is introduced into the data optimization model to optimize the model's performance, guiding the model towards a more stable and reasonable prediction direction, and finally obtaining circuit breaker data with accurate error correction and better noise reduction effect.

[0099] Second embodiment

[0100] Furthermore, in order to implement the circuit breaker data optimization system based on process feedback corresponding to the above method embodiment to achieve corresponding functions and technical effects, Figure 2 A structural diagram of a circuit breaker data optimization system based on process feedback is provided. For ease of illustration, only the parts related to this embodiment are shown. The circuit breaker data optimization system based on process feedback provided in this embodiment of the application includes:

[0101] The feature extraction module 201 is configured to perform wavelet transformation on the obtained circuit breaker data, and extract a first error feature of the circuit breaker data during the wavelet transformation.

[0102] In this embodiment, a high-precision sensor with excellent anti-interference capabilities collects electrical data and device status data from the circuit breaker during operation in real time at a high sampling frequency. The electrical data includes current, voltage, and power data for the branch circuit in which the circuit breaker is located, and the device status data includes the circuit breaker's on / off state and actuation time during operation.

[0103] Conventional technology typically performs noise reduction on circuit breaker data before extracting error characteristics from the noise reduction results. However, in this embodiment, in order to dynamically adjust the noise reduction threshold using feedback from the error characteristics, the circuit breaker data is first subjected to a wavelet transform. Error characteristics are then extracted from the wavelet transform results. The noise reduction effect is then evaluated using the error characteristics, providing data support for subsequently determining the optimal noise reduction threshold.

[0104] The wavelet transform is a transform analysis method that performs time-frequency analysis and processing of signals by providing a frequency-dependent "time-frequency" window. Using scaling and translation operations, the signal is gradually refined at multiple scales, ultimately achieving time segmentation at high frequencies and frequency segmentation at low frequencies. It automatically adapts to the requirements of time-frequency signal analysis, allowing it to focus on any detail in the signal and highlight its characteristics. The number of decomposition layers used in the wavelet transform is generally determined by the complexity and volatility of the data, typically ranging from three to five. The greater the number of layers, the finer the decomposed frequency bands, enabling it to capture more detailed signal features.

[0105] First, the collected circuit breaker data is decomposed into multiple scales using wavelet transform to extract the characteristic information of the circuit breaker data at different frequency bands, which facilitates more detailed local analysis. The details of the circuit breaker data are analyzed at different scales and locations to identify and extract potential error features. The specific formula is:

[0106]

[0107] Where W ψ (s, t) is the wavelet coefficient obtained after wavelet transform, which represents the local characteristics of the data at scale s and position t; ψ is the wavelet basis function, f(x) is the circuit breaker data, s is the scale parameter, and t is the translation parameter.

[0108] Based on the wavelet coefficients obtained across all frequency bands after wavelet transformation, relevant error features such as data mean square error, data variance, data skewness, and data kurtosis are directly extracted. These error features can provide feedback signals for the data denoising process and provide a basis for dynamic adjustment of the denoising threshold.

[0109] The specific expression of the data mean square error is:

[0110]

[0111] In the formula, MSE is the mean square error of the data, x i is the i-th data point of the circuit breaker data, is the i-th data point of the circuit breaker data after wavelet transformation, and N is the total number of data points.

[0112] The data variance is specifically expressed as:

[0113]

[0114] Where, σ 2 is the data variance, μ is the mean of the circuit breaker data, x i is the i-th data point of the circuit breaker data;

[0115] The data skewness is specifically expressed as:

[0116]

[0117] Where Skewness is the data skewness, σ is the data standard deviation, and N is the total number of data points;

[0118] The data kurtosis is specifically expressed as:

[0119]

[0120] Wherein, Kurtosis is the kurtosis of the data.

[0121] The above error characteristics can be used as input to the data optimization model to learn the nonlinear relationship between the noise reduction process and the noise reduction threshold and optimize the data reconstruction.

[0122] The threshold dynamic adjustment module 202 is configured to iteratively adjust the noise reduction threshold using the first error feature as a feedback signal with the goal of achieving the optimal noise reduction effect on the circuit breaker data, thereby obtaining an optimal noise reduction threshold and a second error feature corresponding to the optimal noise reduction threshold.

[0123] In an embodiment of the present application, in the traditional circuit breaker data optimization process, data denoising and error feature extraction are usually two independent steps. However, in the present application, the error feature extraction and data denoising results influence each other to form a dynamic feedback mechanism, in which the data denoising results are evaluated through error features, and new error features are extracted from the denoised data.

[0124] The noise reduction threshold is the core of the data denoising process. This embodiment combines dynamic threshold calculation with a machine learning algorithm to achieve the optimal noise reduction threshold for the best noise reduction effect. Compared to the embodiments of this application, traditional noise reduction threshold calculation methods are based on fixed formulas or manually set parameters. By introducing machine learning to automatically optimize the threshold based on feedback from error characteristics, this significantly improves the noise reduction effect.

[0125] On the basis of step S1, the obtained error features are used as input to train the denoising module of the preset data optimization model, and the denoising module is trained. The data optimization model is generally based on SVR or neural network. SVR stands for Support Vector Regression, abbreviated as support vector regression, which is a branch of support vector machine.

[0126] Through a preset training set, the noise reduction module automatically derives the optimal noise reduction threshold that can achieve the best noise reduction effect by learning the relationship between the error characteristics and the noise reduction threshold.

[0127] As an improvement to the above solution, the embodiment of the present application uses historical data of circuit breakers containing different noise levels, different signal types and different actual working conditions as a training set. The optimal noise reduction threshold obtained through this training set can be applied to different data sets, effectively expanding the scope of application.

[0128] Specifically, the preset training set is denoised using the current denoising threshold, and then the denoising effect is evaluated by calculating the error feature vector. The current denoising threshold is adjusted according to the evaluation result, and the data is denoised again using the adjusted denoising threshold. New error features are extracted from the denoising results to update the error features. The denoising threshold is then adjusted again according to the updated error features until the denoising effect meets the preset optimal denoising effect. The current denoising threshold is then output as the optimal denoising threshold, and the latest error features obtained by the optimal denoising threshold are also output.

[0129] Exemplarily, the mean square error of the data is used to evaluate the noise reduction effect. If the calculated mean square error of the data is high, the noise reduction threshold is adjusted until the mean square error of the data falls into an acceptable normal range, so that effective features can be better extracted from the data.

[0130] The adjustment of the noise reduction threshold is based on the feedback of the error characteristics, so that data noise reduction and error feature extraction can be optimized simultaneously and interact with each other in the same process to form a more efficient processing process.

[0131] The feature fusion module 203 is used to perform dimensionality reduction and fusion on the second error features in sequence to obtain a comprehensive feature vector.

[0132] In the embodiment of the present application, because feature extraction enhancement will generate a large number of redundant features, the redundant features will not only affect the accuracy of subsequent data processing and analysis, but also greatly reduce the data processing efficiency. Therefore, it is necessary to reduce the dimension of the updated error features. By reducing the dimension of the feature space through feature dimensionality reduction, important and representative feature information is retained.

[0133] Optionally, the embodiment of the present application uses PCA and LDA techniques to achieve feature dimensionality reduction in order to select the most representative error features. PCA technology is Principal Components Analysis, which aims to use the idea of ​​dimensionality reduction to transform multiple indicators into a few comprehensive indicators, thereby achieving data dimensionality reduction, denoising and visualization; LDA technology is Linear Discriminant Analysis, which uses statistics, pattern recognition and machine learning methods to try to find a linear combination of features of two types of objects or events in order to characterize or distinguish them. The resulting combination can be used as a linear classifier, or, more commonly, to perform dimensionality reduction processing for subsequent classification.

[0134] After dimensionality reduction, the representative features obtained after dimensionality reduction are fused with other data through multimodal fusion to improve the overall processing efficiency.

[0135] Specifically, representative features are combined with multimodal data to fuse data features from different sources. For example, multiple data such as current, voltage, power, and frequency are weighted averaged and concatenated to generate a comprehensive feature vector. This comprehensive feature vector can provide more information for subsequent error analysis and fault diagnosis.

[0136] The data optimization module 204 is configured to introduce the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization, and perform noise reduction and error correction on the circuit breaker data using the optimized data optimization model to obtain circuit breaker optimized data.

[0137] In order to improve the accuracy of data processing, the embodiment of the present application optimizes the data optimization model by introducing the comprehensive feature vector as a regularization term into the loss function of the data optimization model, so that the model not only focuses on the fitting accuracy during the optimization process, but also takes into account the smoothness and interpretability of the error.

[0138] In the embodiment of the present application, the data optimization model performs data prediction based on support vector regression, and the loss function of support vector regression is as follows:

[0139]

[0140] Where L(θ) is the loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter used to control the tolerance of the model to errors, ο is the threshold of tolerance error, n is the total amount of data, is a regularization term used to prevent overfitting.

[0141] After the introduction of the comprehensive feature vector, the expanded loss function is specifically expressed as follows:

[0142]

[0143] Where L(θ) is the expanded loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter, ο is the threshold of tolerance error, n is the total amount of data, is the regularization term, λ is the adjustment parameter, and MSE is the mean square error of the data.

[0144] By optimizing the data optimization model through comprehensive feature vectors, not only the fitting ability of the model is optimized, but also the model is guided toward a more stable and reasonable prediction direction through error characteristics.

[0145] In addition, the data optimization model also uses the obtained optimal noise reduction threshold to reduce noise on the input data to achieve the optimal noise reduction effect.

[0146] In summary, by introducing the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization, noise reduction, error identification and correction of circuit breaker data are achieved, and more accurate circuit breaker optimization data is obtained, which helps to improve the accuracy of fault diagnosis and anomaly detection of transmission lines.

[0147] The implementation of the embodiments of the present application has the following beneficial effects:

[0148] In this embodiment of the present application, a wavelet transform is first performed on the collected circuit breaker operation data, electrical data, etc., to extract characteristic information of the circuit breaker data in different frequency bands, and based on this information, the error characteristics of the circuit breaker data are further obtained. The error characteristics are then used to dynamically adjust the noise reduction threshold, and a new error characteristic is obtained based on the adjusted noise reduction threshold. The results of the noise reduction process and error feature extraction are mutually influenced, forming a dynamic feedback mechanism. By learning the relationship between the error characteristics and the noise reduction threshold, the optimal noise reduction threshold with the best noise reduction effect is obtained, thereby optimizing the noise reduction effect, achieving accurate error characteristics while eliminating excess noise and enhancing error feature extraction. Because the extraction of enhanced error features generates a large number of redundant features, which affect subsequent processing and analysis, it is necessary to reduce the dimensionality of the feature space through feature dimensionality reduction, and perform feature fusion after dimensionality reduction to improve overall processing efficiency. Finally, the obtained comprehensive feature vector is introduced into the data optimization model to optimize the model's performance, guiding the model towards a more stable and reasonable prediction direction, and finally obtaining circuit breaker data with accurate error correction and better noise reduction effect.

[0149] Further, Figure 3 This is a structural diagram of a terminal device provided in one embodiment of the present application. Figure 3 As shown, the terminal device 3 of this embodiment includes: at least one processor 30 (in Figure 3 Only one is shown) and a memory 31 and a computer program 32 stored in the memory 31 and executable on the at least one processor. When the processor 30 executes the computer program 32, the steps of a circuit breaker data optimization method based on process feedback according to any one of the embodiments of the present application can be implemented.

[0150] The terminal device 3 may be a computing device such as a desktop computer, a cloud server, or a laptop computer. The computing device may include but is not limited to a processor 30 and a memory 31 . Figure 3 This is merely an example of the terminal device 3 and does not constitute a limitation on the terminal device 3 , which may include more or fewer components than those shown in the figure.

[0151] The specific embodiments described above further illustrate the purpose, technical solutions, and beneficial effects of this application. It should be understood that the above description is merely a specific embodiment of this application and is not intended to limit the scope of protection of this application. In particular, it should be noted that for those skilled in the art, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims

1. A circuit breaker data optimization method based on process feedback, characterized in that: include: Performing a wavelet transform on the obtained circuit breaker data, and extracting a first error feature of the circuit breaker data during the wavelet transform process; With the goal of achieving optimal noise reduction effect on the circuit breaker data, the noise reduction threshold is iteratively adjusted using the first error feature as a feedback signal to obtain an optimal noise reduction threshold and a second error feature corresponding to the optimal noise reduction threshold; Performing dimensionality reduction and fusion on the second error features in sequence to obtain a comprehensive feature vector; The comprehensive feature vector and the optimal noise reduction threshold are introduced into a preset data optimization model for model optimization. The circuit breaker data is subjected to noise reduction and error correction through the optimized data optimization model to obtain circuit breaker optimized data.

2. The circuit breaker data optimization method based on process feedback according to claim 1, characterized in that: The wavelet transform is performed on the obtained circuit breaker data, and the first error feature of the circuit breaker data is extracted during the wavelet transform process, specifically: The circuit breaker data is decomposed into several frequency segments through wavelet transform, and the circuit breaker data in each frequency segment is analyzed respectively to extract the first error feature; wherein the first error feature includes data mean square error, data variance, data skewness and data kurtosis.

3. The circuit breaker data optimization method based on process feedback according to claim 1, characterized in that: The goal is to achieve the optimal noise reduction effect of the circuit breaker data, and the noise reduction threshold is iteratively adjusted using the first error feature as a feedback signal to obtain the optimal noise reduction threshold and the second error feature corresponding to the optimal noise reduction threshold, specifically: In the noise reduction module of the data optimization model, noise reduction is performed on a preset training set using the current noise reduction threshold, and then the noise reduction effect is evaluated using the first error feature; wherein the training set includes historical data of circuit breakers with different noise levels, different signal types, and different actual operating conditions; If the noise reduction effect satisfies the preset optimal noise reduction effect, outputting the current noise reduction threshold as the optimal noise reduction threshold and the second error feature; If the noise reduction effect does not meet the preset optimal noise reduction effect, the noise reduction threshold is adjusted according to the noise reduction effect, and the first error feature is updated by the adjusted noise reduction threshold, and then the training set is denoised and evaluated again using the adjusted noise reduction threshold and the updated first error feature.

4. The circuit breaker data optimization method based on process feedback according to claim 3, characterized in that: The second error characteristic is specifically: After obtaining the optimal noise reduction threshold, the first error feature is updated using the optimal noise reduction threshold to obtain the second error feature.

5. The circuit breaker data optimization method based on process feedback according to claim 1, characterized in that: The second error features are sequentially subjected to dimensionality reduction and fusion to obtain a comprehensive feature vector, specifically: Performing dimensionality reduction on the second error feature through linear transformation to obtain several representative features; The representative features are weighted averaged and concatenated to obtain a comprehensive feature vector.

6. The circuit breaker data optimization method based on process feedback according to claim 1, characterized in that: The comprehensive feature vector and the optimal noise reduction threshold are introduced into a preset data optimization model for model optimization, and the circuit breaker data is subjected to noise reduction and error correction by the optimized data optimization model to obtain circuit breaker optimization data, specifically: Introducing the comprehensive feature vector into the loss function of the data optimization model, and introducing the optimal noise reduction threshold into the noise reduction module of the data optimization model to obtain the optimized data optimization model; The circuit breaker data is denoised using the optimized data optimization model, and errors are identified and corrected on the denoising result according to the loss function containing the comprehensive feature vector to obtain circuit breaker optimization data.

7. The circuit breaker data optimization method based on process feedback according to claim 6, characterized in that: The step of introducing the comprehensive feature vector into the loss function of the data optimization model is as follows: The data mean square error of the comprehensive feature vector is introduced into the loss function as a regularization term to obtain the expanded loss function.

8. The circuit breaker data optimization method based on process feedback according to claim 7, characterized in that: The loss function after the expansion is specifically expressed as follows: Where L(θ) is the expanded loss function, y i is the actual value of the i-th data, f(x i ) is the predicted value of the i-th data, C is the penalty parameter, ο is the threshold of tolerance error, n is the total amount of data, is the regularization term, λ is the adjustment parameter, and MSE is the mean square error of the data.

9. A circuit breaker data optimization system based on process feedback, characterized in that: include: Feature extraction module, threshold dynamic adjustment module, feature fusion module and data optimization module; The feature extraction module is used to perform wavelet transform on the obtained circuit breaker data, and extract the first error feature of the circuit breaker data during the wavelet transform process; The threshold dynamic adjustment module is configured to iteratively adjust the noise reduction threshold using the first error characteristic as a feedback signal with the goal of achieving an optimal noise reduction effect on the circuit breaker data, thereby obtaining an optimal noise reduction threshold and a second error characteristic corresponding to the optimal noise reduction threshold; The feature fusion module is used to sequentially reduce the dimension and fuse the second error features to obtain a comprehensive feature vector; The data optimization module is used to introduce the comprehensive feature vector and the optimal noise reduction threshold into a preset data optimization model for model optimization, and to perform noise reduction and error correction on the circuit breaker data through the optimized data optimization model to obtain circuit breaker optimized data.

10. A terminal device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the method implements the steps of a circuit breaker data optimization method based on process feedback according to any one of claims 1 to 8.

Images