A circuit breaker mechanical state monitoring data cleaning method

Abstract

The application discloses a kind of circuit breaker mechanical state monitoring data cleaning methods, it is related to data cleaning technical field.A kind of circuit breaker mechanical state monitoring data cleaning method, comprising the following steps: step one: using EMD algorithm to establish data cleaning model;Step two: obtaining real-time monitoring data and processing;Step three: all monitoring data are decomposed one by one;Step four: to all monitoring data decomposition results are clustered;Step five: remove the noise point data outside data set.The application solves the problem that noise and missing in the monitoring data obtained by the prior art, leading to inaccurate fault diagnosis of circuit breaker, the application makes the data after cleaning cleaner and neater, better extracts the features of real-time monitoring circuit breaker state data, effectively improves the data quality, ensures that the data retained after data cleaning better reflects the operating state of circuit breaker, improves the accuracy of circuit breaker fault diagnosis.

Description

Technical Field

[0001] This invention relates to the field of data cleaning technology, specifically a method for cleaning mechanical condition monitoring data of circuit breakers. Background Technology

[0002] Circuit breakers are an important component of the power grid and have a significant impact on regional power supply. If a circuit breaker malfunctions, it may lead to insufficient power supply or power outages in most areas, and in more serious cases, it may cause great losses to society and the economy. In order to prevent circuit breaker malfunctions, it is necessary to monitor the status of the circuit breaker.

[0003] Chinese Patent Publication No. CN109061463A discloses a method for mechanical condition monitoring and fault diagnosis of high-voltage circuit breakers, including the following steps: 1) Based on the statistical results of historical fault data of high-voltage circuit breakers, artificial fault simulation experiments are conducted on some common high-voltage circuit breakers to obtain fault data; 2) The fault current data is analyzed, and the current data feature quantities of each mechanism fault are extracted as one of the criteria for condition diagnosis classification; 3) The circuit breaker condition data of real-time vibration monitoring is analyzed, and the high and low frequency components of the circuit breaker condition data of real-time vibration monitoring are processed by wavelet packet decomposition and sample entropy respectively to obtain the corresponding wavelet packet relative energy and sample entropy, which are used as the feature quantities of the circuit breaker condition data of real-time vibration monitoring; 4) Principal component analysis is used to reduce the dimensionality of the feature quantities of the circuit breaker condition data of real-time vibration monitoring as one of the criteria for condition diagnosis classification, and support vector machine is used to perform condition diagnosis of the circuit breaker, effectively utilizing the multidimensional information of the fault state and promoting the development of multi-parameter multidimensional mapping fault diagnosis of circuit breakers.

[0004] In actual use, the aforementioned patents are prone to noise and missing data in the monitoring data due to the operating environment of the circuit breaker and the loss of communication lines, resulting in inaccurate fault diagnosis of the circuit breaker. Therefore, they do not meet the existing requirements. To address this, we propose a method for cleaning mechanical condition monitoring data of circuit breakers. Summary of the Invention

[0005] The purpose of this invention is to provide a method for cleaning circuit breaker mechanical condition monitoring data, which makes the cleaned data cleaner and neater, saves a lot of unnecessary work, thereby saving time and resources, better extracts the characteristics of the real-time monitored circuit breaker condition data, provides more detailed analysis, effectively improves data quality, ensures that the data retained after cleaning better reflects the operating status of the circuit breaker, improves the accuracy of circuit breaker fault diagnosis, and solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for cleaning circuit breaker mechanical condition monitoring data, comprising the following steps:

[0007] Step 1: Obtain historical circuit breaker mechanical condition monitoring data, and establish a data cleaning model based on the obtained historical circuit breaker mechanical condition monitoring data using the EMD algorithm;

[0008] Step 2: Acquire real-time monitoring data of the mechanical status of the circuit breaker, process the acquired monitoring data, and store the monitoring data after processing to form a monitoring data column;

[0009] Step 3: Input the monitoring data into the data cleaning model, and use the data cleaning model to decompose all the monitoring data one by one;

[0010] Step 4: Save all monitoring data decomposition results and cluster all monitoring data decomposition results to obtain multiple clustered datasets;

[0011] Step 5: Remove noise data points outside the clustered datasets to obtain cleaned circuit breaker mechanical condition monitoring data.

[0012] Preferably, the circuit breaker mechanical condition monitoring data includes multi-dimensional characteristic data of the circuit breaker's mechanical condition, transmission mechanism data, and energy storage mechanism data. The multi-dimensional characteristic data of the circuit breaker's mechanical condition mainly includes opening and closing data, travel and time characteristics, circuit breaker status data from real-time vibration monitoring, coil current and speed. The acquisition of opening and closing data specifically includes: using a timer to acquire the opening and closing time of the circuit breaker, using a photoelectric travel sensor to acquire the opening and closing displacement of the circuit breaker, and using an audio acquisition device to acquire the opening and closing sound data of the circuit breaker.

[0013] Preferably, the step of acquiring real-time monitoring data of the circuit breaker's mechanical condition and processing the acquired monitoring data specifically includes:

[0014] Real-time acquisition of multi-dimensional characteristic data of the circuit breaker's mechanical state, transmission mechanism data, and energy storage mechanism data;

[0015] Communication interference, white noise interference, and random impulse interference in the collected data are removed. White noise interference is removed by averaging filtering, while communication interference and random impulse interference are removed by median filtering.

[0016] Data augmentation is performed on the noise-removed data to enhance the feature parameters in the noise-removed data.

[0017] Preferably, the step of acquiring real-time monitoring data of the circuit breaker's mechanical condition and processing the acquired monitoring data further includes:

[0018] Extract communication interference, white noise interference, and random impulse interference from the completed data;

[0019] Retrieve the signal-to-noise ratio parameters from the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random pulse interference from the completed data;

[0020] The signal-to-noise ratio (SNR) evaluation coefficient is obtained using the corresponding SNR parameters from the circuit breaker's multi-dimensional mechanical state feature data, transmission mechanism data, and energy storage mechanism data; wherein, the SNR evaluation coefficient is obtained using the following formula:

[0021]

[0022] Where S represents the signal-to-noise ratio evaluation coefficient; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random impulse interference from the data; S hi S represents the signal-to-noise ratio after denoising the i-th data point; xi S represents the signal-to-noise ratio of the i-th data before denoising; xb S represents the standard deviation of the signal-to-noise ratio (SNR) of n data points (including the multi-dimensional feature data of the circuit breaker's mechanical state, transmission mechanism data, and energy storage mechanism data) before denoising processing. hb S represents the standard deviation of the signal-to-noise ratio (SNR) of n data points after denoising, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; hmax and S hmin This represents the maximum and minimum signal-to-noise ratio (SNR) values ​​of n data points after denoising processing, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; S xmax and S xmin The maximum and minimum signal-to-noise ratios (SNRs) of the n data points corresponding to the multi-dimensional feature data of the mechanical state of the circuit breaker, the data of the transmission mechanism, and the data of the energy storage mechanism before denoising processing are represented.

[0023] The signal-to-noise ratio evaluation coefficient is compared with a preset evaluation coefficient threshold.

[0024] When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data after removing communication interference, white noise interference and random pulse interference from the completed data are denoised until the signal-to-noise ratio evaluation coefficient corresponding to the denoised circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data exceeds the preset evaluation coefficient threshold.

[0025] When the signal-to-noise ratio evaluation coefficient is not lower than the preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements.

[0026] Preferably, when the signal-to-noise ratio evaluation coefficient is not lower than a preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements, including:

[0027] When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data is extracted.

[0028] Distortion determination parameters are obtained by combining the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data to complete the noise reduction process.

[0029] The distortion determination parameter is obtained by the following formula:

[0030]

[0031] Where Z represents the distortion determination parameter; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random pulse interference from the data; S hi S represents the signal-to-noise ratio after denoising the i-th data point; xi Z represents the signal-to-noise ratio of the i-th data before denoising; i Z represents the data gain ratio after the i-th data point has undergone denoising processing; c This indicates the preset data gain ratio reference value; S hb Z represents the standard deviation of the signal-to-noise ratio (SNR) of n data points after denoising processing, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; b Z represents the standard deviation of the data gain ratio corresponding to n data points after denoising processing, representing the multidimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; min This represents the minimum data gain ratio of n data points corresponding to the circuit breaker's multi-dimensional mechanical state feature data, transmission mechanism data, and energy storage mechanism data after noise reduction processing; S zx This represents the ratio of the change in signal-to-noise ratio of the data corresponding to the minimum data gain ratio before and after noise removal processing.

[0032] The distortion determination parameter is compared with a preset parameter threshold;

[0033] When the distortion determination parameter exceeds the preset parameter threshold, it is determined that the data distortion is too large, resulting in the data quality being lower than the preset data quality requirements;

[0034] When excessive data distortion causes data quality to fall below the preset data quality requirements, a data quality anomaly alarm will be triggered.

[0035] Preferably, the step of establishing a data cleaning model using the EMD algorithm specifically includes:

[0036] The EMD algorithm is improved, and the processed real-time monitoring data of the mechanical condition of the circuit breaker is received.

[0037] The real-time mechanical condition monitoring data of the circuit breaker is decomposed into multiple intrinsic mode functions using the improved EMD algorithm;

[0038] The intrinsic mode functions are filtered to remove those containing noise and retain those containing useful information.

[0039] The selected intrinsic mode functions are superimposed, and the superimposed intrinsic mode functions are then trained to obtain the data cleaning model.

[0040] Preferably, the improvement of the EMD algorithm specifically includes:

[0041] An improved wavelet denoising method is used to denoise the real-time monitored circuit breaker status data, and the denoised real-time monitored circuit breaker status data is adaptively decomposed into a series of intrinsic mode function components.

[0042] For components of different frequencies, select appropriate prediction models for prediction, and finally superimpose the predicted values ​​of each component to obtain the final predicted value.

[0043] The EMD algorithm is improved by adding Gaussian noise and cubic spline interpolation sampling, and removing the first intrinsic mode function.

[0044] Preferably, the step of using a data cleaning model to decompose all monitoring data one by one specifically includes:

[0045] Column name renaming: If there are two columns with the same name or the same meaning in the monitored data column, rename one of the data columns.

[0046] Remove duplicate values: If duplicate data values ​​exist in the monitored data column, delete the duplicate data values, keeping only the first record of the duplicate data.

[0047] Missing value handling: If there are missing values ​​in the monitored data column, statistical methods are used to fill in the missing values ​​in the data;

[0048] Consistency processing: If there are inconsistent data or naming rules in a data column, use the split column function to split the data values ​​in the inconsistent data column;

[0049] Data sorting and processing: Filtering and sorting the data in the data column, and identifying and handling outliers in the data.

[0050] Preferably, the clustering of all monitoring data decomposition results specifically includes:

[0051] Multi-scale entropy is used to extract features from the multi-dimensional feature data of the mechanical state of circuit breakers in the monitoring data series, and features useful for cluster analysis are extracted.

[0052] Transform features of different dimensions or orders of magnitude in the multidimensional feature data of the mechanical state of circuit breakers to the same scale;

[0053] The K-Means clustering algorithm is used to cluster the feature information in the multidimensional feature data of the mechanical state of the circuit breaker after feature transformation.

[0054] Preferably, the step of extracting features from the multi-dimensional feature data of the circuit breaker mechanical state in the monitoring data column using multi-scale entropy specifically includes:

[0055] The multidimensional feature data of the mechanical state of the circuit breaker are input into the empirical mode decomposition algorithm, and the parameters of the empirical mode decomposition algorithm are initialized to obtain multiple target decomposition parameters;

[0056] Modal decomposition is performed on the multidimensional feature data of the mechanical state of the circuit breaker based on multiple target decomposition parameters to obtain multiple intrinsic mode component data.

[0057] The multi-scale entropy algorithm is used to analyze the features of multiple intrinsic mode components data to obtain the multi-scale entropy value features of each intrinsic mode component data.

[0058] The multi-scale entropy features of each intrinsic mode component data are classified using a support vector machine model to obtain the feature classification results for each intrinsic mode component data.

[0059] Based on the feature classification results of each intrinsic mode component data, multiple intrinsic mode functions of the multidimensional feature data of the circuit breaker's mechanical state are determined; the multiple intrinsic mode functions are iteratively analyzed to obtain the components of the multiple multidimensional feature data of the circuit breaker's mechanical state.

[0060] State features are extracted from the components of the multidimensional feature data of the mechanical state of multiple circuit breakers to obtain the state features of each multidimensional feature data of the mechanical state of multiple circuit breakers.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] This invention removes interference from real-time collected circuit breaker condition monitoring data, resulting in cleaner and more refined data. This eliminates unnecessary work, saving time and resources. By using an improved EMD algorithm to establish a data cleaning model, the features of the real-time monitored circuit breaker condition data can be better extracted, providing more detailed analysis. Through multi-scale entropy analysis, potential problems with the circuit breaker can be detected earlier, allowing for proactive maintenance measures to avoid equipment failure, extend equipment lifespan, reduce maintenance costs, and effectively improve data quality. This ensures that the data retained after cleaning better reflects the operating status of the circuit breaker, thus improving the accuracy of circuit breaker fault diagnosis. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the circuit breaker mechanical condition monitoring data cleaning method of the present invention;

[0064] Figure 2 This is a schematic diagram of feature extraction for multi-dimensional feature data of the mechanical state of the circuit breaker according to the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] To address the issue that existing patents, in practical use, often suffer from noise and missing data in the acquired monitoring data due to factors such as the operating environment of the circuit breaker and losses in communication lines, leading to inaccurate fault diagnosis of the circuit breaker, please refer to [link to relevant documentation]. Figures 1-2 This embodiment provides the following technical solution:

[0067] A method for cleaning circuit breaker mechanical condition monitoring data includes the following steps:

[0068] Step 1: Obtain historical circuit breaker mechanical condition monitoring data, and establish a data cleaning model based on the obtained historical circuit breaker mechanical condition monitoring data using the EMD algorithm;

[0069] Step 2: Acquire real-time monitoring data of the mechanical status of the circuit breaker, process the acquired monitoring data, and store the monitoring data after processing to form a monitoring data column;

[0070] Step 3: Input the monitoring data into the data cleaning model, and use the data cleaning model to decompose all the monitoring data one by one;

[0071] Step 4: Save all monitoring data decomposition results and cluster all monitoring data decomposition results to obtain multiple clustered datasets;

[0072] Step 5: Remove noise data points outside the clustered datasets to obtain cleaned circuit breaker mechanical condition monitoring data.

[0073] Acquire real-time monitoring data of the mechanical condition of the circuit breaker and process the acquired monitoring data, specifically including:

[0074] Real-time acquisition of multi-dimensional characteristic data of circuit breaker mechanical status, transmission mechanism data and energy storage mechanism data. The multi-dimensional characteristic data of circuit breaker mechanical status mainly includes opening and closing data, stroke and time characteristics, circuit breaker status data of real-time vibration monitoring, coil current and speed.

[0075] Communication interference, white noise interference, and random impulse interference in the collected data are removed. White noise interference is removed by averaging filtering, while communication interference and random impulse interference are removed by median filtering.

[0076] Data augmentation enhances the feature parameters of the data after noise removal. By removing noise, invalid, duplicate, and erroneous data, as well as correcting errors and filling in blanks, the data becomes more accurate and reliable. This makes the cleaned data more consistent with reality, leading to more accurate and reliable data analysis results.

[0077] The mechanical condition monitoring data of the circuit breaker includes multi-dimensional characteristic data of the mechanical condition of the circuit breaker, data of the transmission mechanism and data of the energy storage mechanism, and data acquisition of opening and closing. Specifically, it includes: using a timer to collect the opening and closing time of the circuit breaker, using a photoelectric travel sensor to collect the opening and closing displacement of the circuit breaker, and using an audio acquisition device to collect the opening and closing sound data of the circuit breaker. During the acquisition of sound data, echo cancellation technology is used to cancel the echo in the sound data and an adaptive filter is used to filter it. Then, a combination of traditional real-time monitoring circuit breaker condition data processing and deep learning technology is used to further suppress residual echo after echo cancellation and adaptive filtering.

[0078] Specifically, the process of acquiring real-time monitoring data of the circuit breaker's mechanical condition and processing the acquired monitoring data further includes:

[0079] Extract communication interference, white noise interference, and random impulse interference from the completed data;

[0080] Retrieve the signal-to-noise ratio parameters from the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random pulse interference from the completed data;

[0081] The signal-to-noise ratio (SNR) evaluation coefficient is obtained using the corresponding SNR parameters from the circuit breaker's multi-dimensional mechanical state feature data, transmission mechanism data, and energy storage mechanism data; wherein, the SNR evaluation coefficient is obtained using the following formula:

[0082]

[0083] Where S represents the signal-to-noise ratio evaluation coefficient; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random impulse interference from the data; S hi S represents the signal-to-noise ratio after denoising the i-th data point; xi S represents the signal-to-noise ratio of the i-th data before denoising; xb S represents the standard deviation of the signal-to-noise ratio (SNR) of n data points (including the multi-dimensional feature data of the circuit breaker's mechanical state, transmission mechanism data, and energy storage mechanism data) before denoising processing. hb S represents the standard deviation of the signal-to-noise ratio (SNR) of n data points after denoising, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; hmax and S hmin This represents the maximum and minimum signal-to-noise ratio (SNR) values ​​of n data points after denoising processing, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; S xmax and S xmin The maximum and minimum signal-to-noise ratios (SNRs) of the n data points corresponding to the multi-dimensional feature data of the mechanical state of the circuit breaker, the data of the transmission mechanism, and the data of the energy storage mechanism before denoising processing are represented.

[0084] The signal-to-noise ratio evaluation coefficient is compared with a preset evaluation coefficient threshold.

[0085] When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data after removing communication interference, white noise interference and random pulse interference from the completed data are denoised until the signal-to-noise ratio evaluation coefficient corresponding to the denoised circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data exceeds the preset evaluation coefficient threshold.

[0086] When the signal-to-noise ratio evaluation coefficient is not lower than the preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements.

[0087] The technical benefits of the above solution are as follows: by extracting and removing communication interference, white noise interference, and random pulse interference, the accuracy of real-time monitoring data of the circuit breaker's mechanical condition can be effectively improved. The data obtained after noise reduction processing is closer to the actual state, providing a more solid foundation for subsequent analysis and processing.

[0088] Meanwhile, the introduction of a signal-to-noise ratio (SNR) evaluation coefficient as a quantitative indicator of data quality effectively assesses the quality of data after denoising. This evaluation coefficient comprehensively considers multiple factors, including the change in SNR before and after denoising, data standard deviation, maximum and minimum values, and can fully reflect the data quality status. Furthermore, by comparing the SNR evaluation coefficient with a preset evaluation coefficient threshold, the system automatically determines whether further denoising processing is needed. If the evaluation coefficient is below the threshold, denoising continues until the quality requirements are met; if the evaluation coefficient is not below the threshold, the next step is initiated. This automated process reduces manual intervention and improves processing efficiency.

[0089] When excessive data distortion leads to data quality falling below preset requirements, a data quality anomaly alarm is triggered. This helps to promptly identify and address data quality issues, preventing erroneous analysis and decisions based on low-quality data. The entire technical solution enhances the reliability and stability of the circuit breaker mechanical condition monitoring system by optimizing the data processing flow and improving data quality. Accurate data helps to more accurately assess the operating status of circuit breakers, promptly detect potential faults, and thus ensure the safe and stable operation of the power system. The aforementioned technical solution, through a series of scientific data processing and analysis methods, significantly improves the accuracy and reliability of real-time monitoring data of circuit breaker mechanical conditions, providing strong support for the safe and stable operation of the power system.

[0090] Specifically, when the signal-to-noise ratio evaluation coefficient is not lower than the preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements, including:

[0091] When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data is extracted.

[0092] Distortion determination parameters are obtained by combining the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data to complete the noise reduction process.

[0093] The distortion determination parameter is obtained by the following formula:

[0094]

[0095] Where Z represents the distortion determination parameter; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random pulse interference from the data; S hi S represents the signal-to-noise ratio after denoising the i-th data point; xi Z represents the signal-to-noise ratio of the i-th data before denoising; i Z represents the data gain ratio after the i-th data point has undergone denoising processing; c This indicates the preset data gain ratio reference value; S hb Z represents the standard deviation of the signal-to-noise ratio (SNR) of n data points after denoising processing, representing the multi-dimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; b Z represents the standard deviation of the data gain ratio corresponding to n data points after denoising processing, representing the multidimensional feature data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; min This represents the minimum data gain ratio of n data points corresponding to the circuit breaker's multi-dimensional mechanical state feature data, transmission mechanism data, and energy storage mechanism data after noise reduction processing; S zx This represents the ratio of the change in signal-to-noise ratio of the data corresponding to the minimum data gain ratio before and after noise removal processing.

[0096] The distortion determination parameter is compared with a preset parameter threshold;

[0097] When the distortion determination parameter exceeds the preset parameter threshold, it is determined that the data distortion is too large, resulting in the data quality being lower than the preset data quality requirements;

[0098] When excessive data distortion causes data quality to fall below the preset data quality requirements, a data quality anomaly alarm will be triggered.

[0099] The technical effect of the above solution is as follows: After the signal-to-noise ratio evaluation coefficient meets the basic requirements (i.e., not lower than the preset evaluation coefficient threshold), a distortion judgment parameter is further introduced to evaluate the authenticity and accuracy of the data. This dual evaluation mechanism can more comprehensively reflect the quality of the data and ensure the reliability of subsequent analysis and decision-making. By extracting the data gain ratio of the multi-dimensional characteristic data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data, and combining it with the denoised data, the distortion judgment parameter is calculated. The data gain ratio reflects the degree of change of the data during the denoising process. Incorporating it into the distortion evaluation can more accurately reflect the effect of data processing and the true state of the data.

[0100] The calculation formula for the distortion judgment parameter comprehensively considers multiple factors, including the signal-to-noise ratio before and after denoising, the data gain ratio and its standard deviation, minimum value, and the change rate of the signal-to-noise ratio corresponding to the minimum data gain ratio before and after denoising. This comprehensive evaluation method can more scientifically judge the degree of data distortion and avoid misjudgments that may be caused by a single indicator. When the distortion judgment parameter exceeds the preset parameter threshold, the system can quickly determine that the data distortion is too large, resulting in data quality lower than the preset requirements. At this time, the system can automatically trigger a data quality anomaly alarm, reminding relevant personnel to pay attention to and deal with data quality problems in a timely manner, avoiding incorrect analysis and decisions based on low-quality data. At the same time, by introducing the distortion judgment and data quality anomaly alarm mechanism, this technical solution can significantly improve the overall performance of the circuit breaker mechanical condition monitoring system. It can not only ensure the accuracy and reliability of data, but also promptly detect and deal with data quality problems, providing a more solid guarantee for the safe and stable operation of the power system.

[0101] The data cleaning model is established using the EMD algorithm, specifically including:

[0102] The EMD algorithm is improved, and the processed real-time monitoring data of the mechanical condition of the circuit breaker is received.

[0103] The real-time monitoring data of the circuit breaker's mechanical condition is decomposed into multiple intrinsic mode functions (IMFs) using an improved EMD algorithm. Noisy IMFs are removed, while IMFs containing useful information are retained. The filtered IMFs are then superimposed, and the superimposed IMFs are used for training to obtain a data cleaning model.

[0104] By using the improved EMD algorithm to establish a data cleaning model, the features of real-time monitored circuit breaker status data can be better extracted. The complex time series real-time monitored circuit breaker status data can be decomposed into a series of intrinsic mode functions, each of which corresponds to features at different time scales. This decomposition method helps to capture the local features of real-time monitored circuit breaker status data, especially when dealing with complex and variable real-time monitored circuit breaker status data, and can provide more detailed analysis.

[0105] The EMD algorithm has been improved, specifically including:

[0106] An improved wavelet denoising method is used to denoise the real-time monitored circuit breaker status data, and the denoised real-time monitored circuit breaker status data is adaptively decomposed into a series of intrinsic mode function components.

[0107] For components of different frequencies, select appropriate prediction models for prediction, and finally superimpose the predicted values ​​of each component to obtain the final predicted value.

[0108] The EMD algorithm is improved by adding Gaussian noise and cubic spline interpolation sampling, and removing the first intrinsic mode function.

[0109] By improving the EMD algorithm, the decomposition accuracy of the data cleaning model is enhanced, resulting in more accurate decomposition results when cleaning circuit breaker condition monitoring data. The improved wavelet denoising method is used to denoise the real-time monitored circuit breaker condition data, making the spectral peaks at the fault characteristic frequencies more obvious, thereby improving the accuracy of circuit breaker condition monitoring data analysis.

[0110] The data cleaning model was used to decompose all monitoring data one by one, specifically including:

[0111] Column name renaming: If there are two columns with the same name or the same meaning in the monitored data column, rename one of the data columns.

[0112] Remove duplicate values: If duplicate data values ​​exist in the monitored data column, delete the duplicate data values, keeping only the first record of the duplicate data.

[0113] Missing value handling: If there are missing values ​​in the monitored data column, statistical methods are used to fill in the missing values. By determining the data type, missing value situation, and data distribution in the monitored data column, the data quality is improved. When handling missing values, different strategies are adopted according to the number of missing values ​​and the importance of the attribute. For example, when the missing rate is low and the attribute importance is low, the mean or median can be used for filling. When the missing rate is high and the attribute importance is high, imputation or modeling methods may be required.

[0114] Consistency processing: If there are inconsistent data or naming rules in a data column, use the split column function to split the data values ​​in the inconsistent data column;

[0115] Data sorting and processing: Filtering and sorting the data in the data column, and identifying and handling outliers in the data to ensure the accuracy and reliability of the data;

[0116] Clustering is performed on all the decomposed monitoring data results, specifically including:

[0117] Multi-scale entropy is used to extract features from the multi-dimensional characteristic data of the mechanical state of circuit breakers in the monitoring data series, and features useful for cluster analysis are extracted. This helps to reduce the dimensionality of the data, while ensuring that the extracted features can best represent the essence of the data.

[0118] By standardizing the multidimensional feature data of the mechanical state of circuit breakers, features of different dimensions or orders of magnitude can be transformed to the same scale, thereby ensuring that clustering algorithms are more fair and effective in processing these features.

[0119] The K-Means clustering algorithm is used to cluster the feature information in the multidimensional feature data of the mechanical state of the circuit breaker after feature transformation. K-Means is a common clustering algorithm that attempts to divide the data into K clusters, so that the data in each cluster are as similar as possible, while the data in different clusters are as different as possible. Through iterative optimization, the K-Means algorithm attempts to minimize the variance within the clusters, thereby achieving the purpose of clustering.

[0120] The advantages of using data cleaning models to cluster the results of data decomposition mainly include improved data quality and accuracy, enhanced data analysis effectiveness and reliability, and savings in time and resources. Data cleaning removes invalid, duplicate, and erroneous data, making the data more accurate and reliable. This not only improves the overall quality of the data but also provides a more reliable foundation for subsequent data analysis, making the cleaned data more consistent with reality. This leads to more accurate and reliable data analysis results, ensuring the reliability and effectiveness of the data analysis. It also enables clustering algorithms to more accurately identify patterns and structures in the data, thereby improving the quality and effectiveness of clustering. Cleaned data is cleaner and more organized, saving a lot of unnecessary work, such as manually checking and correcting erroneous data, thus saving time and resources and making the data analysis process more efficient.

[0121] Feature extraction is performed on the multi-dimensional feature data of the mechanical state of circuit breakers in the monitoring data column using multi-scale entropy, specifically including:

[0122] The multidimensional feature data of the mechanical state of the circuit breaker are input into the empirical mode decomposition algorithm, and the parameters of the empirical mode decomposition algorithm are initialized to obtain multiple target decomposition parameters;

[0123] Modal decomposition is performed on the multidimensional feature data of the mechanical state of the circuit breaker based on multiple target decomposition parameters to obtain multiple intrinsic mode component data.

[0124] The multi-scale entropy algorithm is used to analyze the features of multiple intrinsic mode components data to obtain the multi-scale entropy value features of each intrinsic mode component data.

[0125] The multi-scale entropy features of each intrinsic mode component data are classified using a support vector machine model to obtain the feature classification results for each intrinsic mode component data.

[0126] Based on the feature classification results of each intrinsic mode component data, multiple intrinsic mode functions of the multidimensional feature data of the circuit breaker's mechanical state are determined; the multiple intrinsic mode functions are iteratively analyzed to obtain the components of the multiple multidimensional feature data of the circuit breaker's mechanical state.

[0127] State features are extracted from the components of the multidimensional feature data of the mechanical state of multiple circuit breakers to obtain the state features of each multidimensional feature data of the mechanical state of multiple circuit breakers.

[0128] By utilizing multi-scale entropy to analyze real-time monitored circuit breaker status data across different time scales, it is possible to capture the behavioral characteristics of circuit breakers at various time scales. For example, a circuit breaker may experience frequent operations in a short period of time, while exhibiting stable performance over a longer time scale. Furthermore, since circuit breaker operation may be affected by various factors, including environmental factors and usage frequency, multi-scale entropy can more accurately reflect the impact of these factors on circuit breaker performance, thereby improving the accuracy of fault diagnosis. Through multi-scale entropy analysis, potential problems with circuit breakers can be detected earlier, allowing for proactive maintenance measures to avoid equipment failure, extend equipment lifespan, reduce maintenance costs, and effectively improve data quality. This ensures that the data retained after cleaning better reflects the operating status of the circuit breaker, thus enhancing the accuracy of circuit breaker fault diagnosis.

[0129] This invention discloses a method for cleaning circuit breaker mechanical condition monitoring data. By removing interference from real-time collected circuit breaker condition monitoring data, the cleaned data becomes cleaner and more concise, saving unnecessary work, thus conserving time and resources. By using an improved EMD algorithm to establish a data cleaning model, the characteristics of the real-time monitored circuit breaker condition data can be better extracted. Complex time-series real-time monitored circuit breaker condition data can be decomposed into a series of intrinsic mode functions (IMFs), each corresponding to characteristics at different time scales. This decomposition method helps capture the local characteristics of the real-time monitored circuit breaker condition data, especially when dealing with complex and variable real-time monitored circuit breaker condition data, providing more detailed analysis. Through multi-scale entropy analysis, potential problems of the circuit breaker can be detected earlier, allowing for proactive maintenance measures, avoiding equipment failure, extending equipment lifespan, reducing maintenance costs, effectively improving data quality, ensuring that the data retained after cleaning better reflects the operating status of the circuit breaker, and improving the accuracy of circuit breaker fault diagnosis.

[0130] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0131] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for cleaning mechanical condition monitoring data of circuit breakers, characterized in that, Includes the following steps: Step 1: Obtain historical circuit breaker mechanical condition monitoring data, and establish a data cleaning model based on the obtained historical circuit breaker mechanical condition monitoring data using the EMD algorithm; Step 2: Acquire real-time monitoring data of the mechanical status of the circuit breaker, process the acquired monitoring data, and store the monitoring data after processing to form a monitoring data column; Step 3: Input the monitoring data into the data cleaning model, and use the data cleaning model to decompose all the monitoring data one by one; Step 4: Save all monitoring data decomposition results and cluster all monitoring data decomposition results to obtain multiple clustered datasets; Step 5: Remove noisy data points outside the clustered datasets to obtain the cleaned circuit breaker mechanical condition monitoring data; The process of acquiring real-time monitoring data of the mechanical status of the circuit breaker and processing the acquired monitoring data also includes: extracting data information after removing communication interference, white noise interference and random pulse interference from the completed data; Retrieve the signal-to-noise ratio parameters from the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random pulse interference from the completed data; The signal-to-noise ratio (SNR) evaluation coefficient is obtained using the corresponding SNR parameters from the circuit breaker's multi-dimensional mechanical state feature data, transmission mechanism data, and energy storage mechanism data; wherein, the SNR evaluation coefficient is obtained using the following formula: ； Where S represents the signal-to-noise ratio (SNR) evaluation coefficient; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random impulse interference; Shi represents the SNR of the i-th data point after denoising; Sxi represents the SNR of the i-th data point before denoising; Sxb represents the standard deviation of the SNR of the n data points corresponding to the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data before denoising; and Shb represents the standard deviation of the SNR of the n data points corresponding to the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after denoising. Shmax and Shmin represent the maximum and minimum signal-to-noise ratios (SNR) of n data points after denoising processing, representing the multi-dimensional characteristic data of the circuit breaker's mechanical state, the transmission mechanism data, and the energy storage mechanism data; Sxmax Sxmin represents the maximum and minimum signal-to-noise ratios (SNRs) of n data points corresponding to the circuit breaker's mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data before denoising processing; the SNR evaluation coefficient is compared with a preset evaluation coefficient threshold. When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data after removing communication interference, white noise interference and random pulse interference from the completed data are denoised until the signal-to-noise ratio evaluation coefficient corresponding to the denoised circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data exceeds the preset evaluation coefficient threshold. When the signal-to-noise ratio evaluation coefficient is not lower than the preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements.

2. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 1, characterized in that: The circuit breaker mechanical condition monitoring data includes multi-dimensional characteristic data of the circuit breaker's mechanical condition, transmission mechanism data, and energy storage mechanism data. The multi-dimensional characteristic data of the circuit breaker's mechanical condition includes opening and closing data, travel and time characteristics, circuit breaker status data from real-time vibration monitoring, coil current and speed. The acquisition of opening and closing data specifically includes: using a timer to collect the opening and closing time of the circuit breaker, using a photoelectric travel sensor to collect the opening and closing displacement of the circuit breaker, and using an audio acquisition device to collect the opening and closing sound data of the circuit breaker.

3. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 2, characterized in that: The process of acquiring real-time monitoring data of the mechanical state of the circuit breaker and processing the acquired monitoring data specifically includes: real-time acquisition of multi-dimensional characteristic data of the circuit breaker's mechanical state, transmission mechanism data, and energy storage mechanism data; removal of communication interference, white noise interference, and random pulse interference from the acquired data; and data enhancement of the data after interference removal to enhance the characteristic parameters in the data after interference removal.

4. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 1, characterized in that: When the signal-to-noise ratio evaluation coefficient is not lower than the preset evaluation coefficient threshold, data distortion is determined, and a data quality anomaly alarm is triggered if the data distortion is too large and the data quality is lower than the preset data quality requirements, including: When the signal-to-noise ratio evaluation coefficient is lower than the preset evaluation coefficient threshold, the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data and energy storage mechanism data is extracted. Distortion determination parameters are obtained by combining the data gain ratio of the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data to complete the noise reduction process. The distortion determination parameter is obtained by the following formula: WITH log . ; Where Z represents the distortion determination parameter; n represents the number of data points contained in the circuit breaker mechanical state multidimensional feature data, transmission mechanism data, and energy storage mechanism data after removing communication interference, white noise interference, and random impulse interference; Shi represents the signal-to-noise ratio of the i-th data after denoising; Sxi represents the signal-to-noise ratio of the i-th data before denoising; Zi represents the data gain ratio of the i-th data after denoising; Zc represents the preset data gain ratio reference value; Shb represents the standard deviation of the signal-to-noise ratio of the n data points after denoising; Zb represents the standard deviation of the data gain ratio of the n data points after denoising; and Zmin represents the standard deviation of the data gain ratio of the n data points after denoising. The minimum data gain ratio corresponding to each data point; Szx represents the change ratio of the signal-to-noise ratio of the data corresponding to the minimum data gain ratio before and after noise removal processing. The distortion determination parameter is compared with a preset parameter threshold; When the distortion determination parameter exceeds the preset parameter threshold, it is determined that the data distortion is too large, resulting in the data quality being lower than the preset data quality requirements; When excessive data distortion causes data quality to fall below the preset data quality requirements, a data quality anomaly alarm will be triggered.

5. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 1, characterized in that: The establishment of a data cleaning model using the EMD algorithm specifically includes: The EMD algorithm is improved, and the processed real-time monitoring data of the mechanical condition of the circuit breaker is received. The real-time mechanical condition monitoring data of the circuit breaker is decomposed into multiple intrinsic mode functions by the improved EMD algorithm for filtering. Intrinsic mode functions containing noise are removed, and intrinsic mode functions containing useful information are retained. The selected intrinsic mode functions are superimposed, and the superimposed intrinsic mode functions are then trained to obtain the data cleaning model.

6. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 5, characterized in that: The improvements to the EMD algorithm specifically include: An improved wavelet denoising method is used to denoise the real-time monitored circuit breaker status data, and the denoised real-time monitored circuit breaker status data is adaptively decomposed into a series of intrinsic mode function components. For components of different frequencies, select appropriate prediction models for prediction, and finally superimpose the predicted values ​​of each component to obtain the final predicted value. The EMD algorithm is improved by adding Gaussian noise and cubic spline interpolation sampling, and removing the first intrinsic mode function.

7. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 1, characterized in that: The data cleaning model decomposes all monitoring data one by one, specifically including: Column name renaming: If there are two columns with the same name or the same meaning in the monitored data column, rename one of the data columns. Remove duplicate values: If duplicate data values ​​exist in the monitored data column, delete the duplicate data values, keeping only the first record of the duplicate data. Missing value handling: If there are missing values ​​in the monitored data column, statistical methods are used to fill in the missing values ​​in the data; Consistency processing: If there are inconsistent data or naming rules in a data column, use the split column function to split the data values ​​in the inconsistent data column; Data sorting and processing: Filtering and sorting the data in the data column, and identifying and handling outliers in the data.

8. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 1, characterized in that: The clustering of all monitoring data decomposition results specifically includes: Multi-scale entropy is used to extract features from the multi-dimensional feature data of the mechanical state of circuit breakers in the monitoring data series, and features useful for cluster analysis are extracted. Transform features of different dimensions or orders of magnitude in the multidimensional feature data of the mechanical state of circuit breakers to the same scale; The K-Means clustering algorithm is used to cluster the feature information in the multidimensional feature data of the mechanical state of the circuit breaker after feature transformation.

9. The method for cleaning circuit breaker mechanical condition monitoring data according to claim 8, characterized in that: The feature extraction of multi-dimensional feature data of circuit breaker mechanical status from the monitoring data column using multi-scale entropy specifically includes: The multidimensional feature data of the mechanical state of the circuit breaker are input into the empirical mode decomposition algorithm, and the parameters of the empirical mode decomposition algorithm are initialized to obtain multiple target decomposition parameters; Modal decomposition is performed on the multidimensional feature data of the mechanical state of the circuit breaker based on multiple target decomposition parameters to obtain multiple intrinsic mode component data. The multi-scale entropy algorithm is used to analyze the features of multiple intrinsic mode components data to obtain the multi-scale entropy value features of each intrinsic mode component data. The multi-scale entropy features of each intrinsic mode component data are classified using a support vector machine model to obtain the feature classification results for each intrinsic mode component data. Based on the feature classification results of each intrinsic mode component data, multiple intrinsic mode functions of the multidimensional feature data of the circuit breaker's mechanical state are determined; the multiple intrinsic mode functions are iteratively analyzed to obtain the components of the multiple multidimensional feature data of the circuit breaker's mechanical state. State features are extracted from the components of the multidimensional feature data of the mechanical state of multiple circuit breakers to obtain the state features of each multidimensional feature data of the mechanical state of multiple circuit breakers.

Images