Method and device for diagnosing blocking degree of opening and closing coil iron core of circuit breaker based on feature fusion

Abstract

The invention discloses a feature fusion-based circuit breaker opening and closing coil iron core blocking degree diagnosis method and device, and aims to solve the problems of low diagnosis precision, poor real-time performance, lack of multi-dimensional data fusion and the like in the prior art. According to the invention, the accuracy and the real-time performance of fault diagnosis of the circuit breaker can be improved, and the safety and the stability of a power system are enhanced. The diagnosis method comprises the following steps: (1) simulating iron core blocking faults of different degrees to form a plurality of fault category samples; (2) collecting an iron core motion image and a current signal to ensure the accuracy of data collection; (3) preprocessing the acquired image data; (4) extracting the dynamic characteristics of the movement of the iron core; (5) analyzing the synchronously acquired current signals, and extracting time domain features of the current signals; (6) carrying out model training on the fused multi-dimensional feature vectors by adopting a random forest algorithm; and (7) fault diagnosis.

Description

Technical Field

[0001] The invention relates to a power equipment fault diagnosis technology, in particular to a method for diagnosing the blocking degree of an opening and closing coil core of a high-voltage circuit breaker. Background Art

[0002] As a key protection device in the power system, high-voltage circuit breakers undertake the important task of ensuring the safe and stable operation of the power grid. It protects power equipment from overload, short circuit and other faults by quickly switching and controlling the current, thereby ensuring the reliability of power supply. However, with the continuous increase in power load and the increasing complexity of the power grid structure, the operating environment of circuit breakers has become more and more demanding. In this context, the failure rate of circuit breakers has also increased, among which the blocking problem of the opening and closing coil core has become one of the main factors affecting its normal operation.

[0003] The electromagnet of the opening and closing coil is the core component that drives the circuit breaker. Its iron core should be able to move quickly and flexibly under the action of current. However, in actual operation, the iron core may be blocked due to insufficient lubrication, pollution, mechanical wear or corrosion. This phenomenon will not only prolong the response time of the circuit breaker, causing it to fail to cut off the fault current in time, but may also cause serious damage to the equipment and even cause power outages in the power system. Therefore, accurately identifying the degree of blockage of the iron core is crucial to improving the reliability of the circuit breaker and the safety of the power system.

[0004] At present, the monitoring and diagnosis methods for circuit breaker faults mainly focus on the extraction and analysis of current characteristics. These methods use the waveform characteristics of the coil current to judge the status of the equipment. Although they are effective to a certain extent, their limitations are gradually emerging. First, the current characteristics often cannot reflect the actual movement state of the iron core, and when a fault occurs, the change in the current waveform may be relatively weak, resulting in insufficient accuracy and timeliness of fault identification. In addition, most traditional monitoring methods rely on static data analysis, lack real-time capture of the dynamic behavior of the iron core, and cannot respond to sudden faults in a timely manner.

[0005] In this context, the rapid development of computer vision and deep learning technology has provided new possibilities for fault monitoring of power equipment. Through high-speed camera technology, the movement state of the iron core can be captured in real time during the operation of the circuit breaker, and the dynamic characteristics such as its movement trajectory, speed and acceleration can be extracted using the optical flow method. This method can overcome the limitations of traditional technology and more comprehensively reflect the real movement of the iron core under the action of current. Combined with the analysis of the change of current signals, it can provide a deeper understanding of the relationship between current characteristics and iron core movement, providing richer information for fault diagnosis. Summary of the invention

[0006] The purpose of the present invention is to provide a method and device for diagnosing the degree of blockage of the opening and closing coil core of a circuit breaker based on feature fusion, so as to solve the problems of low diagnostic accuracy, poor real-time performance and lack of multi-dimensional data fusion in the prior art. Through the present invention, the accuracy and real-time performance of circuit breaker fault diagnosis can be improved, and the safety and stability of the power system can be enhanced.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for diagnosing the degree of iron core blockage of circuit breaker opening and closing coils based on feature fusion.

[0009] The following steps are involved:

[0010] (1) Wrap different numbers of tapes around the opening and closing coils of the circuit breaker to simulate different degrees of core blocking faults and form multiple fault category samples;

[0011] (2) Use a high-speed camera and a Hall-type open current clamp to synchronously collect the core motion image and current signal to ensure the accuracy of data collection;

[0012] (3) preprocessing the collected image data, including removing noise and enhancing image contrast, and performing low-pass filtering and normalization on the current waveform data;

[0013] (4) The Lucase-Kanade optical flow method is used to track the motion of the core under current control frame by frame and extract the dynamic characteristics of the core motion;

[0014] (5) Analyze the synchronously collected current signal and extract its time domain characteristics, including peak value, valley value and its rate of change, to form a current feature vector;

[0015] (6) The random forest algorithm is used to train the fused multi-dimensional feature vector model, and the cross-validation technology is used to optimize the model parameters to establish a diagnostic model for the circuit breaker coil core blocking process;

[0016] (7) Fault diagnosis;

[0017] The output of the fault diagnosis includes whether there is a core blocking fault and the classification of its severity.

[0018] In one possible implementation,

[0019] The dynamic characteristics of step (4) include trajectory, velocity and acceleration.

[0020] In one possible implementation,

[0021] The fault categories of the core blocking fault in step (1) include no blocking, slight blocking, moderate blocking and severe blocking, and each category contains 20 groups of samples.

[0022] In one possible implementation,

[0023] The acquisition frequency of the high-speed camera is 3000 frames per second, and the sampling frequency of the current signal reaches 25600 Hz.

[0024] In one possible implementation,

[0025] In the step (4), extracting the dynamic characteristics of the core movement includes extracting the maximum velocity v of the core movement max 、Maximum acceleration a max , speed change rate v avg1 , acceleration rate of change J avg1 , velocity and acceleration peak factor PF v PF a and kurtosis K v , K a etc.; select the current double peak value and valley value I max1 ,I max2 ,I max3 and the corresponding time t max1 ,t max2 ,t max3 For the time domain feature, F(t0) is selected as the time integral feature.

[0026] In one possible implementation,

[0027] In the model training of step (6), the training set and the test set are allocated in proportion, and the training set is used to construct multiple decision trees to form a classifier.

[0028] In one possible implementation,

[0029] It also includes step (7): fault diagnosis;

[0030] The output result of the fault diagnosis step in step (7) includes whether there is a blocking fault in the iron core and the classification of its severity.

[0031] The present invention also discloses a circuit breaker opening and closing coil core blocking degree diagnosis device based on feature fusion,

[0032] Including: Fault simulation and data acquisition unit: simulate different degrees of core blocking faults to form multiple fault category samples; synchronously collect core motion images and current signals to ensure the accuracy of data acquisition;

[0033] Data preprocessing unit: preprocess the collected image data;

[0034] Feature extraction and data fusion unit: extract the dynamic characteristics of the core movement; analyze the synchronously collected current signal and extract its time domain characteristics;

[0035] Model training unit: Perform model training on the fused multi-dimensional feature vector, optimize model parameters using cross-validation technology, and establish a diagnostic model for the circuit breaker coil core blocking process.

[0036] Fault diagnosis and result output unit: Outputs diagnostic results including whether there is a blocking fault in the iron core and the classification of its severity.

[0037] As a preferred embodiment of the circuit breaker opening and closing coil core blocking degree diagnosis device based on feature fusion disclosed in the present invention,

[0038] The fault simulation and data acquisition unit performs fault simulation by winding different numbers of turns of tape on the opening and closing coil of the circuit breaker, and collects the core motion image and current signal by a high-speed camera and a Hall open current clamp.

[0039] The beneficial effects of the present invention are:

[0040] The method and device for diagnosing the degree of core blockage of circuit breaker opening and closing coils based on feature fusion disclosed in the present invention achieves accurate diagnosis of circuit breaker faults by integrating information from multiple data sources and using advanced machine learning technology. Compared with traditional methods, the present invention significantly improves the accuracy and real-time performance of diagnosis, and provides a strong technical guarantee for the stable operation of the power system.

[0041] The present invention aims to improve the real-time monitoring and fault identification capabilities of the core motion state of the circuit breaker during operation by combining computer vision, machine learning and current signal analysis. By using a high-speed camera and optical flow method to capture the dynamic characteristics of the core and integrating the current characteristics, the mechanical performance and safety of the circuit breaker can be more comprehensively evaluated, thereby providing a guarantee for the stable operation of the power system. This technology has broad application prospects in the fields of power industry, automation control, equipment maintenance and management, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Blocking fault identification flow chart;

[0043] Figure 2 This is a comparison chart before and after current treatment;

[0044] Figure 3 is a schematic diagram of image tracking;

[0045] Figure 4 It is the stroke-current joint analysis diagram;

[0046] Figure 5 It is the optimization result of fault feature recognition. DETAILED DESCRIPTION

[0047] The specific implementation of the present invention is described below in conjunction with the accompanying drawings and embodiments:

[0048] In order to better illustrate the technical solution of the present invention, the present invention is further described in detail below in conjunction with specific embodiments.

[0049] Overall,

[0050] The present invention provides a method for diagnosing a circuit breaker opening and closing coil core blocking process based on image current multi-data fusion, comprising the following steps:

[0051] First, by wrapping different numbers of tapes around the circuit breaker coil, we simulated different degrees of core blocking faults, including no blocking, mild blocking, moderate blocking, and 20 groups of each, to complete the classification of samples. During the opening and closing process of the circuit breaker simulating the fault, the image data and current waveform data of the coil current were collected, and the moving image of the core was recorded. This step ensured the diversity and representativeness of the sample data through physical simulation, providing a reliable training data foundation for the subsequent diagnostic model.

[0052] Next, the collected image data is preprocessed, including denoising and contrast enhancement; the current waveform data is filtered and normalized to remove noise and standardize the data. The purpose of the preprocessing step is to improve the quality of the data, reduce errors caused by noise and other interference factors, and ensure the accuracy and stability of subsequent feature extraction.

[0053] Then, key features such as stroke, motion speed and acceleration are extracted from the preprocessed image data; features such as peaks, troughs, rise time and fall time are extracted from the current waveform data; and motion trajectory and speed features are extracted from the core motion image. These image features, current waveform features and core motion features are fused to form a multi-dimensional feature vector. By fusing multiple features, the dynamic changes of the circuit breaker during the opening and closing process can be fully reflected, providing a richer information basis for fault diagnosis.

[0054] After data fusion is completed, the random forest algorithm is used for model training to establish a diagnostic model for the circuit breaker coil core blocking process. The random forest algorithm can effectively improve the classification performance and noise resistance of the model by constructing multiple decision trees and combining their output results. During the model training process, cross-validation and grid search techniques are used to optimize the model parameters to ensure the generalization ability and stability of the model.

[0055] Finally, the trained random forest model is used to diagnose the newly collected data, determine whether there is a blockage in the circuit breaker coil core, and output the diagnosis results. The diagnosis results can not only identify whether there is a fault, but also classify the degree of fault according to the combination of different features, providing a detailed reference for maintenance and repair. At the same time, by analyzing historical data and diagnostic results, decision support can be provided for the operation and maintenance of the power system, and the operation efficiency and safety of the power system can be improved.

[0056] The above-mentioned method for diagnosing the degree of core blockage of circuit breaker opening and closing coil based on image current multi-data fusion, the method simulates different degrees of core blockage faults, including no blockage, mild blockage, moderate blockage and severe blockage, by winding different numbers of tapes on the circuit breaker coil, and completes the classification calibration of samples. In the process of circuit breaker opening and closing with simulated fault, the image data and current waveform data of the coil current are collected, and the motion image of the core is recorded at the same time. Then, the collected image data is preprocessed, including denoising and contrast enhancement, and the current waveform data is filtered and normalized. After data preprocessing, key features such as stroke, motion speed and acceleration are extracted from the image data, features such as peaks, troughs, rise time and fall time are extracted from the current waveform data, and motion trajectory and speed features are extracted from the core motion image. These image features, current waveform features and core motion features are fused to form a multidimensional feature vector. Using the fused feature vector, the random forest algorithm is used for model training to establish a diagnostic model for the process of core blockage of circuit breaker coil. Finally, the trained random forest model is used to diagnose the newly collected data, determine whether there is a blockage in the circuit breaker coil core, and output the diagnosis results. This method improves the accuracy and real-time performance of circuit breaker fault diagnosis through the application of multi-data fusion and machine learning technology, and provides effective technical support for the stable operation of the power system.

[0057] The following are specific embodiments:

[0058] This embodiment mainly describes the implementation steps of the method for diagnosing the core blocking process of the ZN63 circuit breaker opening and closing coil based on image current multi-data fusion.

[0059] 1. Fault simulation and data collection:

[0060] Select a ZN63 circuit breaker and wrap different numbers of tapes around its opening and closing coils to simulate different degrees of core blocking faults. Fault categories include no blocking, mild blocking, moderate blocking, and severe blocking, with 20 groups in each category to complete the category calibration of the samples. Select a high-speed camera (for example, HT-XG51GC / MT) and a Hall open current clamp to ensure the accuracy of data acquisition. Build an experimental platform such as Figure 2, simulate the experimental fault by wrapping different numbers of tape turns, aim the high-speed camera at the core movement area, and install it on the current clamp in the current path of the opening and closing coil to ensure synchronous acquisition. Start the circuit breaker to perform opening and closing operations, use the high-speed camera to capture the image sequence of the core movement process at a frequency of 3000 frames per second, and synchronously record the current signal of the opening and closing coil, ensuring that the sampling frequency of the current signal reaches 25600Hz to capture its subtle changes.

[0061] 2. Data preprocessing:

[0062] The collected image data is preprocessed. First, the median filter algorithm is used to remove image noise, and then the image contrast is enhanced by the histogram equalization method. For the current waveform data, a low-pass filter is used to remove high-frequency noise, and normalization is performed to eliminate the amplitude difference between different data. The preprocessed data has higher quality and provides a good foundation for feature extraction.

[0063] 3. Feature extraction and data fusion:

[0064] The core is stationary at the initial stage of power on. As the coil current increases, the electromagnetic force generated pushes the core to start moving. The core accelerates until the speed drops to zero when it hits the locking device, and then falls back to the initial position. The core movement process is captured frame by frame by a high-speed camera, and the Lucase-Kanade optical flow method is used to accurately track the current-controlled core movement process. Figure 4 Select the tracking area. Figure 5 Point A is the frame where the movement starts, point B is the frame with the maximum displacement, point C is the lowest frame for the first fall, point D is the highest frame for the fall, and point E is the frame where the movement ends. Identify and extract the dynamic features of the core movement, such as trajectory, velocity, and acceleration. These motion features can accurately reflect the actual motion state of the core during operation and provide basic data for subsequent fault analysis. At the same time, analyze the synchronously collected current signal and extract its time domain features, including peak value, valley value, and its rate of change, to form a current feature vector. Repeat the operation 100 times to extract the maximum velocity v of the core motion image information. max 、Maximum acceleration a max , speed change rate v avg1 , acceleration rate of change J avg1 , velocity and acceleration peak factor PF v PF a and kurtosis K v , K a Select the current double peak value and valley value I max1 ,I max2 ,I max3 and the corresponding time t max1 ,t max2 ,t max3It is a time domain feature. In actual tests, it is found that the current and time integrals change significantly under different degrees of core blockage. The integral value F(t) of the time-current from 0 to t is represented by F(t0), and F(t0) is selected as the feature.

[0065] 4. Model training:

[0066] The fused multi-dimensional feature vectors are used to train the model using the random forest algorithm. Specifically, the data set is first divided into a training set and a test set, and then the random forest algorithm is used to build multiple decision trees on the training set. By combining the output results of these decision trees, a powerful classifier is formed. During the model training process, cross-validation technology is used to optimize parameters and select the optimal combination of model parameters to improve the generalization ability and accuracy of the model.

[0067] 5. Fault diagnosis and result output:

[0068] Use the trained random forest model to diagnose the newly collected data. Input the feature vector of the new data, and the model outputs the diagnosis results, including whether there is a core blocking fault and its severity classification. By comparing the model output results with the actual situation, the accuracy and reliability of the model can be verified. The accuracy of current feature fault recognition is 36.66%, the accuracy of image feature recognition is 83.33%, and the accuracy of the combined feature vector is 93.33%. Figure 5 As shown, the recognition accuracy is greatly improved.

[0069] Through the above steps, the method of the present invention realizes accurate diagnosis and classification of the ZN63 circuit breaker opening and closing coil core blocking fault. Compared with the traditional method, the present invention significantly improves the accuracy and real-time performance of fault diagnosis, and provides a strong guarantee for the safe operation of the power system.

[0070] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in the field without departing from the purpose of the present invention. These changes involve related technologies well known to those skilled in the art, which all fall within the scope of protection of the patent of the present invention.

[0071] Many other changes and modifications may be made without departing from the concept and scope of the present invention.It should be understood that the present invention is not limited to the specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. A method for diagnosing the degree of iron core blockage of circuit breaker opening and closing coils based on feature fusion. It is characterized in that The following steps are involved: (1) Wrap different numbers of tapes around the opening and closing coils of the circuit breaker to simulate different degrees of core blocking faults and form multiple fault category samples; (2) Use a high-speed camera and a Hall-type open current clamp to synchronously collect the core motion image and current signal to ensure the accuracy of data collection; (3) preprocessing the collected image data, including removing noise and enhancing image contrast, and performing low-pass filtering and normalization on the current waveform data; (4) The Lucase-Kanade optical flow method is used to track the motion of the core under current control frame by frame and extract the dynamic characteristics of the core motion; (5) Analyze the synchronously collected current signal and extract its time domain characteristics, including peak value, valley value and its rate of change, to form a current feature vector; (6) The random forest algorithm is used to train the fused multi-dimensional feature vector model, and the cross-validation technology is used to optimize the model parameters to establish a diagnostic model for the circuit breaker coil core blocking process; (7) Fault diagnosis; The output of the fault diagnosis includes whether there is a core blocking fault and the classification of its severity.

2. The method for diagnosing the degree of blockage of the circuit breaker opening and closing coil core based on feature fusion according to claim 1, Features: The dynamic characteristics of step (4) include trajectory, velocity and acceleration.

3. The method for diagnosing the degree of blockage of the circuit breaker opening and closing coil core based on feature fusion according to claim 1, It is characterized in that The fault categories of the core blocking fault in step (1) include no blocking, slight blocking, moderate blocking and severe blocking, and each category contains 20 groups of samples.

4. The method for diagnosing the degree of blockage of the circuit breaker opening and closing coil core based on feature fusion according to claim 1, Features: The acquisition frequency of the high-speed camera is 3000 frames per second, and the sampling frequency of the current signal reaches 25600 Hz.

5. The method for diagnosing the degree of blockage of the circuit breaker opening and closing coil core based on feature fusion according to claim 1, Features: In the step (4), extracting the dynamic characteristics of the core movement includes extracting the maximum velocity v of the core movement max 、Maximum acceleration a max , velocity change rate v avg1 , acceleration rate of change J avg1 , velocity and acceleration peak factor PF v PF a and kurtosis K v , K a etc.; select the current double peak value and valley value I max1 ,I max2 ,I max3 and the corresponding time t max1 ,t max2 ,t max3 For the time domain feature, F(t0) is selected as the time integral feature.

6. The method for diagnosing the degree of blockage of the circuit breaker opening and closing coil core based on feature fusion according to claim 1, Features: In the model training of step (6), the training set and the test set are allocated in proportion, and the training set is used to construct multiple decision trees to form a classifier.

7. Circuit breaker opening and closing coil core blocking degree diagnosis device based on feature fusion, Features: include: Fault simulation and data acquisition unit: simulates core blocking faults of different degrees to form multiple fault category samples; Synchronously collect core motion images and current signals to ensure the accuracy of data collection; Data preprocessing unit: preprocess the collected image data; Feature extraction and data fusion unit: extract the dynamic characteristics of the core movement; Analyze the synchronously collected current signal and extract its time domain characteristics; Model training unit: Perform model training on the fused multi-dimensional feature vector, optimize model parameters using cross-validation technology, and establish a diagnostic model for the circuit breaker coil core blocking process. Fault diagnosis and result output unit: Outputs diagnostic results including whether there is a blocking fault in the iron core and the classification of its severity.

8. The circuit breaker opening and closing coil core blockage diagnosis device based on feature fusion according to claim 7, Features: The fault simulation and data acquisition unit performs fault simulation by winding different numbers of turns of tape on the opening and closing coil of the circuit breaker, and collects the core motion image and current signal by a high-speed camera and a Hall open current clamp.

Images