Circuit breaker fault detection method, system and equipment
By preprocessing and multi-characteristic parameter detection of the circuit breaker in the closed state, using network structures such as convolutional layer and residual module, the leakage caused by relying on a single characteristic parameter in the circuit breaker fault detection in the prior art is solved. Inspection problems improve detection accuracy and save costs.
Patent Information
- Application Number
- CN202411200580.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-29
AI Technical Summary
The fault detection of existing circuit breakers relies on a single characteristic parameter, resulting in some fault missed detection, reducing the reliability of detection.
By obtaining the action data of the circuit breaker in the closed state, data preprocessing is performed to obtain the sound feature map and the current feature map, and a pre-trained fault detection model, including convolutional layer, residual module, pooling layer and LSTM The network structure of layer and other networks performs fault detection on multiple characteristic parameters.
Improve the accuracy of circuit breaker fault detection, avoid missed detection caused by a single characteristic parameter, save detection costs, and eliminate the need to collect vibration signals.
Smart Images

Figure CN118980923A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of circuit breaker fault detection, and in particular to a circuit breaker fault detection method, system and device. Background Art
[0002] As an important equipment in the power system, the safe and reliable operation of the circuit breaker directly affects the safety and stability of the power system. Most accidents caused by circuit breaker failures are mechanical failures of the circuit breaker. Therefore, accurate detection of circuit breaker failures is of great significance to ensure the reliable operation of the power system.
[0003] At present, the detection of circuit breaker faults is based on a single characteristic parameter, such as the vibration signal generated during the operation of the high-voltage circuit breaker, to extract and analyze the characteristics to detect whether the high-voltage circuit breaker is faulty. However, in the actual operation of the circuit breaker, the fault characteristics are complex and diverse, and different fault characteristics will be manifested in different characteristic parameters. Therefore, detection based on a single characteristic parameter will miss some circuit breaker faults, reducing the reliability of circuit breaker fault detection. Summary of the invention
[0004] The present invention provides a circuit breaker fault detection method, system and device, which solve the technical problem that the current circuit breaker fault detection is based on a single characteristic parameter, which easily misses some circuit breaker faults and reduces the reliability of circuit breaker fault detection.
[0005] A first aspect of the present invention provides a circuit breaker fault detection method, comprising:
[0006] Acquiring action data of the circuit breaker in a closed state, and performing data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram;
[0007] Inputting the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network;
[0008] Performing feature extraction on the sound feature graph through the first extraction network to obtain a sound extraction feature graph;
[0009] Performing feature extraction on the current feature graph through the second extraction network to obtain a current extraction feature graph;
[0010] The detection network is used to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
[0011] Optionally, the action data includes sound data and current data, and the step of performing data preprocessing on the action data to obtain a sound characteristic graph and a current characteristic graph includes:
[0012] Using the current data to construct a closing current curve, and extracting closing current information during a current valley period in the closing current curve;
[0013] Extracting effective sound signals from the sound data according to a preset characteristic time segment;
[0014] The effective sound signal and the closing current information are respectively processed into images to obtain a sound image and a current characteristic diagram.
[0015] Optionally, the first extraction network includes a convolutional layer, a first residual module, a second residual module, a pooling layer and an LSTM layer connected in sequence, and the step of extracting features from the sound feature map through the first extraction network to obtain the sound extraction feature map includes:
[0016] Extracting features from the sound feature map through a convolutional layer to obtain a first sound feature map;
[0017] Using the first residual module to perform feature extraction on the first sound feature map to obtain a second sound feature map; wherein the first residual module includes a first extraction branch, a second extraction branch and a feature fusion layer;
[0018] Performing feature extraction on the second sound feature map through the second residual module to obtain a third sound feature map; wherein the second residual module includes a third extraction branch, a fourth extraction branch and a feature fusion layer;
[0019] Performing a pooling operation on the third sound feature map through a pooling layer to obtain a fourth sound feature map;
[0020] The LSTM layer is used to perform feature extraction on the fourth sound feature map to obtain a sound extraction feature map.
[0021] Optionally, the second extraction network includes a convolutional layer, a third residual module, a pooling layer and an LSTM layer connected in sequence, and the step of extracting features from the current feature map through the second extraction network to obtain the current extraction feature map includes:
[0022] Extracting features from the current characteristic graph through a convolution layer to obtain a first current characteristic graph;
[0023] The third residual module is used to perform feature extraction on the first current characteristic map to obtain a second current characteristic map; wherein the third residual module includes a fifth extraction branch and a feature fusion layer;
[0024] Performing a pooling operation on the second current characteristic map through a pooling layer to obtain a third current characteristic map;
[0025] The LSTM layer is used to extract features from the third current feature map to obtain a current extraction feature map.
[0026] Optionally, the detection network includes a feature fusion layer, an extraction module, and an activation layer connected in sequence, and the step of using the detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker includes:
[0027] Using a feature fusion layer to perform feature fusion on the sound extraction feature map and the current extraction feature map to obtain an information fusion feature map;
[0028] Performing feature extraction on the information fusion feature map through an extraction module to obtain an information feature map; wherein the extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, a random drop layer and a fully connected layer connected in sequence;
[0029] An activation layer is used to perform fault detection on the information characteristic graph to obtain a fault detection result corresponding to the circuit breaker.
[0030] Optionally, the step of extracting features from the first sound feature map using the first residual module to obtain a second sound feature map includes:
[0031] Performing feature extraction on the first sound feature map through the first extraction branch to generate a first feature map, wherein the first extraction branch includes a 3×3 convolution layer, a batch normalization layer, a DY-Relu activation layer, a 3×3 convolution layer, a batch normalization layer, and a DY-Relu activation layer connected in sequence;
[0032] Performing feature extraction on the first sound feature map through the second extraction branch to generate a second feature map, wherein the second extraction branch includes a 3×3 convolutional layer, a batch normalization layer, and a Relu activation layer connected in sequence;
[0033] A feature fusion layer is used to perform feature fusion on the first feature map and the second feature map to obtain a second sound feature map.
[0034] Optionally, the step of extracting features from the second sound feature map by using the second residual module to obtain a third sound feature map includes:
[0035] Performing feature extraction on the second sound feature map through the third extraction branch to obtain a third feature map; wherein the third extraction branch includes a batch normalization layer, a Relu activation layer, a 3×1 convolution layer, a batch normalization layer, a Relu activation layer, a 1×3 convolution layer and a global average pooling layer connected in sequence;
[0036] Performing feature extraction on the second sound feature map through the fourth extraction branch to obtain a fourth feature map; wherein the fourth extraction branch includes a 3×3 convolution layer, a batch normalization layer, and a Swish activation layer connected in sequence;
[0037] A feature fusion layer is used to perform feature fusion on the third feature map and the fourth feature map to obtain a third sound feature map.
[0038] Optionally, the step of extracting features from the first current characteristic graph using the third residual module to obtain a second current characteristic graph includes:
[0039] Performing feature extraction on the first current feature map through the fifth extraction branch to obtain a fifth feature map; wherein the fifth extraction branch includes a batch normalization layer, a Swish activation layer, a 3×3 convolutional layer, a batch normalization layer, a Swish activation layer, and a 3×3 convolutional layer connected in sequence;
[0040] A feature fusion layer is used to perform feature fusion on the fifth feature map and the first current feature map to obtain a second current feature map.
[0041] A second aspect of the present invention provides a circuit breaker fault detection system, comprising:
[0042] The acquisition module is used to obtain the action data of the circuit breaker in the closed state, and perform data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram;
[0043] An input module, used for inputting the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network;
[0044] A first extraction module, configured to perform feature extraction on the sound feature map through the first extraction network to obtain a sound extraction feature map;
[0045] A second extraction module, used for performing feature extraction on the current characteristic graph through the second extraction network to obtain a current extraction characteristic graph;
[0046] The detection module is used to use the detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
[0047] A third aspect of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the circuit breaker fault detection method as described in any one of the above items.
[0048] It can be seen from the above technical solutions that the present invention has the following advantages:
[0049] The present invention obtains the action data of the circuit breaker in the closed state, performs data preprocessing on the action data, obtains the sound characteristic graph and the current characteristic graph, and then uses the pre-trained fault detection model to perform fault detection on the sound characteristic graph and the current characteristic graph, so that the circuit breaker fault can be detected through multiple characteristic parameters, overcoming the technical problem that a single characteristic parameter easily misses some circuit breaker faults and reduces the reliability of circuit breaker fault detection. Compared with traditional circuit breaker fault detection, the accuracy of circuit breaker fault detection is improved and there is no need to collect vibration signals to analyze the circuit breaker, saving detection costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0051] Figure 1 A flow chart of the steps of a circuit breaker fault detection method provided in Embodiment 1 of the present invention;
[0052] Figure 2 A flow chart of the steps of a circuit breaker fault detection method provided in Embodiment 2 of the present invention;
[0053] Figure 3 A schematic diagram of the structure of a fault detection model provided in the second embodiment of the present invention;
[0054] Figure 4 A structural block diagram of a circuit breaker fault detection system provided in Embodiment 3 of the present invention;
[0055] Figure 5 This is a structural block diagram of a computer device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION
[0056] The embodiments of the present invention provide a circuit breaker fault detection method, system and device, which are used to solve the technical problem that the current circuit breaker fault detection is based on a single characteristic parameter, which easily misses some circuit breaker faults and reduces the reliability of circuit breaker fault detection.
[0057] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0058] See also Figure 1 , Figure 1 A flowchart of the steps of a circuit breaker fault detection method provided in Embodiment 1 of the present invention.
[0059] The present invention provides a circuit breaker fault detection method, comprising:
[0060] Step 101, obtaining the action data of the circuit breaker in the closed state, and performing data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram;
[0061] In an embodiment of the present invention, current information and sound information of a circuit breaker when it is closed are obtained, and effective areas of the current information and the sound information are extracted respectively to obtain current effective area information and sound effective area information, and then the current effective area information and the sound effective area information are converted into images to obtain a sound characteristic diagram and a current characteristic diagram.
[0062] It should be noted that the current effective area information is the current information within the current valley time. The sound effective area information is the sound information from the moment when the closing electromagnet of the circuit breaker strikes to the moment when the moving and static contacts of the circuit breaker strike.
[0063] Step 102: input the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network;
[0064] In an embodiment of the present invention, the sound characteristic graph and the current characteristic graph are used as inputs of a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network.
[0065] It should be noted that the training process of the fault detection model is specifically as follows: 1. Perform a mechanical characteristic test on the target circuit breaker through a mechanical characteristic tester to obtain training action data (including sound training data and current training data). 3. Perform data preprocessing on the training action data to generate an action feature set. 4. Use the action feature set to input the preset initial fault detection model for training and output the fault detection data. 5. Calculate the training loss function value of the action feature set based on the fault detection data. 6. When the training loss function value is greater than or equal to the preset standard loss function value, use the grid search method or the random search method to adjust the network parameters of the initial fault detection model until the training loss function value is less than the preset standard loss function value. 7. When the training loss function value is less than the preset standard loss function value, a fault detection model is generated.
[0066] Step 103: extracting features from the sound feature map through a first extraction network to obtain a sound extraction feature map;
[0067] In an embodiment of the present invention, a first extraction network is used to perform feature extraction on a sound feature map to obtain a sound extraction feature map, wherein the first extraction network includes a convolutional layer, a first residual module, a second residual module, a pooling layer and an LSTM layer connected in sequence.
[0068] Step 104, extracting features from the current feature graph through a second extraction network to obtain a current extraction feature graph;
[0069] In an embodiment of the present invention, a second extraction network is used to perform feature extraction on a current feature map to obtain a current extraction feature map, wherein the second extraction network includes a convolutional layer, a third residual module, a pooling layer, and an LSTM layer connected in sequence.
[0070] Step 105: Use a detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
[0071] In an embodiment of the present invention, the detection network includes a feature fusion layer, an extraction module and an activation layer which are connected in sequence. First, the sound extraction feature map and the current extraction feature map are feature fused through the feature fusion layer, and then the fused information fusion feature map is subjected to fault detection through the extraction module and the activation layer to obtain a fault detection result corresponding to the circuit breaker.
[0072] In the embodiment of the present invention, the present invention obtains the action data of the circuit breaker in the closed state, performs data preprocessing on the action data, obtains the sound characteristic diagram and the current characteristic diagram, and then uses the pre-trained fault detection model to perform fault detection on the sound characteristic diagram and the current characteristic diagram, so that the circuit breaker fault can be detected through multiple characteristic parameters, overcoming the technical problem that a single characteristic parameter easily misses some circuit breaker faults and reduces the reliability of circuit breaker fault detection. Compared with traditional circuit breaker fault detection, the accuracy of circuit breaker fault detection is improved and there is no need to collect vibration signals to analyze the circuit breaker, saving detection costs.
[0073] See also Figure 2 , Figure 2 A flowchart of the steps of a circuit breaker fault detection method provided in Embodiment 2 of the present invention.
[0074] The present invention provides a circuit breaker fault detection method, comprising:
[0075] Step 201, obtaining sound data and current data of the circuit breaker in a closed state;
[0076] In the embodiment of the present invention, sound information and current information of the circuit breaker when it is closed are obtained.
[0077] In another embodiment, sound information and current information of the circuit breaker when it is opened are obtained.
[0078] Step 202: construct a closing current curve using current data, and extract closing current information during a current valley period in the closing current curve;
[0079] In the embodiment of the present invention, a closing current curve is constructed by using current data, and closing current information in a current valley period in the closing current curve is intercepted.
[0080] In another embodiment, a tripping current curve is constructed using current data, and the tripping current information during a current valley period in the tripping current curve is intercepted.
[0081] Step 203: extracting effective sound signals from the sound data according to a preset characteristic time segment;
[0082] In the embodiment of the present invention, the sound information from the moment when the closing electromagnet of the circuit breaker strikes to the moment when the moving and static contacts of the circuit breaker strike is extracted from the sound data as the effective sound signal.
[0083] Step 204 , respectively perform image processing on the effective sound signal and the closing current information to obtain a sound image and a current characteristic diagram.
[0084] In the embodiment of the present invention, the effective sound signal and the closing current information (opening current information) are respectively processed into images to obtain the sound image and the current characteristic diagram.
[0085] Step 205: input the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network;
[0086] In the embodiment of the present invention, the specific implementation process of step 205 is similar to step 102 and will not be repeated here.
[0087] Step 206: extracting features from the sound feature map through the first extraction network to obtain a sound extraction feature map;
[0088] Further, the first extraction network includes a convolutional layer, a first residual module, a second residual module, a pooling layer and an LSTM layer connected in sequence, and step 206 includes the following sub-steps:
[0089] S11, extracting features from the sound feature map through a convolutional layer to obtain a first sound feature map;
[0090] In the embodiment of the present invention, a convolution operation is performed on the sound feature map through a convolution layer to obtain a first sound feature map.
[0091] It should be noted that the convolution layer is implemented by performing an inner product (multiplying each element and then summing them) operation on the image (or data in different data windows) and the filter matrix (a set of fixed weights).
[0092] S12, using a first residual module to perform feature extraction on the first sound feature map to obtain a second sound feature map; wherein the first residual module includes a first extraction branch, a second extraction branch and a feature fusion layer;
[0093] Furthermore, S12 includes the following sub-steps:
[0094] S121, extracting features from the first sound feature map through a first extraction branch to generate a first feature map, wherein the first extraction branch includes a 3×3 convolutional layer, a batch normalization layer, a DY-Relu activation layer, a 3×3 convolutional layer, a batch normalization layer, and a DY-Relu activation layer connected in sequence;
[0095] In this embodiment of the present invention, feature extraction is performed on the first sound feature map through a 3×3 convolution layer, a batch normalization layer, a DY-Relu activation layer, a 3×3 convolution layer, a batch normalization layer, and a DY-Relu activation layer in sequence to generate a first feature map.
[0096] S122, extracting features from the first sound feature map through a second extraction branch to generate a second feature map, wherein the second extraction branch includes a 3×3 convolutional layer, a batch normalization layer, and a Relu activation layer connected in sequence;
[0097] In the embodiment of the present invention, feature extraction is performed on the first sound feature map through a 3×3 convolution layer, a batch normalization layer, and a Relu activation layer in sequence to generate a second feature map.
[0098] S123. Use a feature fusion layer to fuse the first feature map and the second feature map to obtain a second sound feature map.
[0099] In the embodiment of the present invention, the first feature map and the second feature map are feature fused by a feature fusion layer to obtain a second sound feature map.
[0100] It should be noted that the residual processing of S121-S123 solves the problem of gradient vanishing in the fault detection model, so that the fault detection model can be trained more deeply.
[0101] S13, extracting features from the second sound feature map through a second residual module to obtain a third sound feature map; wherein the second residual module includes a third extraction branch, a fourth extraction branch and a feature fusion layer;
[0102] Furthermore, S13 includes the following sub-steps:
[0103] S131, extracting features from the second sound feature map through a third extraction branch to obtain a third feature map; wherein the third extraction branch includes a batch normalization layer, a Relu activation layer, a 3×1 convolution layer, a batch normalization layer, a Relu activation layer, a 1×3 convolution layer, and a global average pooling layer connected in sequence;
[0104] In an embodiment of the present invention, feature extraction is performed on the second sound feature map through a batch normalization layer, a Relu activation layer, a 3×1 convolution layer, a batch normalization layer, a Relu activation layer, a 1×3 convolution layer and a global average pooling layer in sequence to obtain a third feature map.
[0105] It should be noted that the setting of 3×1 convolutional layers and 1×3 convolutional layers in the third extraction branch can improve the computational efficiency of the fault detection model while ensuring the computational accuracy of the fault detection model.
[0106] S132, extracting features from the second sound feature map through a fourth extraction branch to obtain a fourth feature map; wherein the fourth extraction branch includes a 3×3 convolutional layer, a batch normalization layer, and a Swish activation layer connected in sequence;
[0107] In the embodiment of the present invention, feature extraction is performed on the second sound feature map through a 3×3 convolution layer, a batch normalization layer, and a Swish activation layer in sequence to obtain a fourth feature map.
[0108] It should be noted that replacing the Relu activation layer with the Swish activation layer in the fourth extraction branch can solve some limitations of ReLU, such as the "neuron death" problem. The Swish function, through its unique mathematical expression, can generate a non-zero gradient when the input is negative, thus avoiding the problem of the ReLU function's gradient disappearance when the input is negative.
[0109] S133. Use a feature fusion layer to perform feature fusion on the third feature map and the fourth feature map to obtain a third sound feature map.
[0110] In the embodiment of the present invention, the third feature map and the fourth feature map are feature fused through the feature fusion layer to obtain a third sound feature map with high semantic information and high resolution.
[0111] S14, performing a pooling operation on the third sound feature map through a pooling layer to obtain a fourth sound feature map;
[0112] S15. Use the LSTM layer to extract features from the fourth sound feature map to obtain a sound extraction feature map.
[0113] In an embodiment of the present invention, a pooling operation is performed on the third sound feature map through a pooling layer to obtain a fourth sound feature map, and a LSTM layer is used to perform feature extraction on the fourth sound feature map to obtain a sound extraction feature map.
[0114] Step 207, extracting features from the current feature graph through a second extraction network to obtain a current extraction feature graph;
[0115] Further, the second extraction network includes a convolutional layer, a third residual module, a pooling layer and an LSTM layer connected in sequence, and step 207 includes the following sub-steps:
[0116] S21, extracting features from the current characteristic graph through a convolution layer to obtain a first current characteristic graph;
[0117] In the embodiment of the present invention, a convolution operation is performed on the current characteristic map through a convolution layer to obtain a first current characteristic map.
[0118] S22, using a third residual module to extract features from the first current feature map to obtain a second current feature map; wherein the third residual module includes a fifth extraction branch and a feature fusion layer;
[0119] Further, S22 includes the following sub-steps:
[0120] S221. Perform feature extraction on the first current feature map through a fifth extraction branch to obtain a fifth feature map; wherein the fifth extraction branch includes a batch normalization layer, a Swish activation layer, a 3×3 convolutional layer, a batch normalization layer, a Swish activation layer, and a 3×3 convolutional layer connected in sequence;
[0121] In an embodiment of the present invention, feature extraction is performed on the first current feature map through a batch normalization layer, a Swish activation layer, a 3×3 convolutional layer, a batch normalization layer, a Swish activation layer, and a 3×3 convolutional layer in sequence to obtain a fifth feature map.
[0122] S222. Use a feature fusion layer to perform feature fusion on the fifth feature map and the first current feature map to obtain a second current feature map.
[0123] In the embodiment of the present invention, the fifth feature map and the first current feature map are feature fused by a feature fusion layer to obtain a second current feature map.
[0124] S23, performing a pooling operation on the second current characteristic map through a pooling layer to obtain a third current characteristic map;
[0125] S24. Use the LSTM layer to extract features from the third current feature map to obtain a current extraction feature map.
[0126] In an embodiment of the present invention, a pooling operation is performed on the second current characteristic map through a pooling layer to obtain a third current characteristic map, and a LSTM layer is used to perform feature extraction on the third current characteristic map to obtain a current extraction feature map.
[0127] Step 208: Use a detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
[0128] Further, the detection network includes a feature fusion layer, an extraction module, and an activation layer connected in sequence, and step 208 includes the following sub-steps:
[0129] S31, using a feature fusion layer to perform feature fusion on the sound extraction feature map and the current extraction feature map to obtain an information fusion feature map;
[0130] In the embodiment of the present invention, a feature fusion layer is used to fuse the features of the sound extraction feature map and the current extraction feature map to obtain an information fusion feature map.
[0131] S32, extracting features from the information fusion feature map through an extraction module to obtain an information feature map; wherein the extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, a random drop layer and a fully connected layer connected in sequence;
[0132] In the embodiment of the present invention, feature extraction is performed on the information fusion feature map in sequence through a fully connected layer, a Swish activation layer, a batch normalization layer, a random dropout layer and a fully connected layer to obtain an information feature map.
[0133] S33. Use the activation layer to perform fault detection on the information characteristic graph to obtain a fault detection result corresponding to the circuit breaker.
[0134] In the embodiment of the present invention, a PLTanh activation layer or a Soflmax activation layer is used to perform fault detection on the information characteristic graph to obtain a fault detection result corresponding to the circuit breaker (the fault detection result is divided into fault or normal).
[0135] In the embodiment of the present invention, the present invention obtains the action data of the circuit breaker in the closed state, performs data preprocessing on the action data, obtains the sound characteristic diagram and the current characteristic diagram, and then uses the pre-trained fault detection model to perform fault detection on the sound characteristic diagram and the current characteristic diagram, so that the circuit breaker fault can be detected through multiple characteristic parameters, overcoming the technical problem that a single characteristic parameter easily misses some circuit breaker faults and reduces the reliability of circuit breaker fault detection. Compared with traditional circuit breaker fault detection, the accuracy of circuit breaker fault detection is improved and there is no need to collect vibration signals to analyze the circuit breaker, saving detection costs.
[0136] See also Figure 4 , Figure 4 This is a structural block diagram of a circuit breaker fault detection system provided in Embodiment 3 of the present invention.
[0137] The present invention provides a circuit breaker fault detection system, comprising:
[0138] The acquisition module 301 is used to obtain the action data of the circuit breaker in the closed state, and perform data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram;
[0139] An input module 302, used to input the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network;
[0140] A first extraction module 303, configured to extract features from the sound feature map through a first extraction network to obtain a sound extraction feature map;
[0141] A second extraction module 304 is used to extract features from the current feature graph through a second extraction network to obtain a current extraction feature graph;
[0142] The detection module 305 is used to perform fault detection on the sound extraction feature graph and the current extraction feature graph using a detection network to obtain a fault detection result corresponding to the circuit breaker.
[0143] Furthermore, the action data includes sound data and current data, and the acquisition module 301 includes:
[0144] A current extraction submodule is used to construct a closing current curve using current data and extract closing current information during a current valley period in the closing current curve;
[0145] A sound extraction submodule is used to extract effective sound signals from sound data according to a preset characteristic time segment;
[0146] The image conversion submodule is used to respectively process the effective sound signal and the closing current information into images to obtain the sound image and the current characteristic diagram.
[0147] Furthermore, the first extraction network includes a convolutional layer, a first residual module, a second residual module, a pooling layer and an LSTM layer connected in sequence, and the first extraction module 303 includes:
[0148] A first sound feature submodule, used for extracting features from the sound feature map through a convolution layer to obtain a first sound feature map;
[0149] A second sound feature submodule, used to extract features from the first sound feature map using the first residual module to obtain a second sound feature map; wherein the first residual module includes a first extraction branch, a second extraction branch and a feature fusion layer;
[0150] A third sound feature submodule, used for extracting features from the second sound feature map through the second residual module to obtain a third sound feature map; wherein the second residual module includes a third extraction branch, a fourth extraction branch and a feature fusion layer;
[0151] a fourth sound feature submodule, configured to perform a pooling operation on the third sound feature map through a pooling layer to obtain a fourth sound feature map;
[0152] The fifth sound feature submodule is used to use the LSTM layer to extract features from the fourth sound feature map to obtain a sound extraction feature map.
[0153] Further, the second extraction network includes a convolutional layer, a third residual module, a pooling layer and an LSTM layer connected in sequence, and the second extraction module 304 includes:
[0154] A first current feature submodule, used for extracting features from the current feature map through a convolution layer to obtain a first current feature map;
[0155] A second current feature submodule, used for extracting features from the first current feature map using a third residual module to obtain a second current feature map; wherein the third residual module includes a fifth extraction branch and a feature fusion layer;
[0156] A third current characteristic submodule, used for performing a pooling operation on the second current characteristic map through a pooling layer to obtain a third current characteristic map;
[0157] The fourth current feature submodule is used to use the LSTM layer to extract features from the third current feature map to obtain a current extraction feature map.
[0158] Further, the detection network includes a feature fusion layer, an extraction module and an activation layer connected in sequence, and the detection module 305 includes:
[0159] A fusion submodule, used for performing feature fusion on the sound extraction feature map and the current extraction feature map by using a feature fusion layer to obtain an information fusion feature map;
[0160] The information feature submodule is used to extract features from the information fusion feature map through the extraction module to obtain the information feature map; wherein the extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, a random drop layer and a fully connected layer connected in sequence;
[0161] The detection submodule is used to perform fault detection on the information feature graph using the activation layer to obtain the fault detection result corresponding to the circuit breaker.
[0162] Furthermore, the second sound feature submodule includes:
[0163] A first feature unit, configured to extract features from the first sound feature map through a first extraction branch to generate a first feature map, wherein the first extraction branch includes a 3×3 convolutional layer, a batch normalization layer, a DY-Relu activation layer, a 3×3 convolutional layer, a batch normalization layer, and a DY-Relu activation layer connected in sequence;
[0164] A second feature unit, used for performing feature extraction on the first sound feature map through a second extraction branch to generate a second feature map, wherein the second extraction branch includes a 3×3 convolution layer, a batch normalization layer, and a Relu activation layer connected in sequence;
[0165] The third feature unit is used to use a feature fusion layer to perform feature fusion on the first feature map and the second feature map to obtain a second sound feature map.
[0166] Furthermore, the third sound feature submodule includes:
[0167] A fourth feature unit is used to extract features from the second sound feature map through a third extraction branch to obtain a third feature map; wherein the third extraction branch includes a batch normalization layer, a Relu activation layer, a 3×1 convolution layer, a batch normalization layer, a Relu activation layer, a 1×3 convolution layer and a global average pooling layer connected in sequence;
[0168] A fifth feature unit, configured to extract features from the second sound feature map through a fourth extraction branch to obtain a fourth feature map; wherein the fourth extraction branch includes a 3×3 convolutional layer, a batch normalization layer, and a Swish activation layer connected in sequence;
[0169] The sixth feature unit is used to use a feature fusion layer to perform feature fusion on the third feature map and the fourth feature map to obtain a third sound feature map.
[0170] Furthermore, the second current characteristic submodule includes:
[0171] A seventh feature unit, configured to extract features from the first current feature map through a fifth extraction branch to obtain a fifth feature map; wherein the fifth extraction branch includes a batch normalization layer, a Swish activation layer, a 3×3 convolutional layer, a batch normalization layer, a Swish activation layer, and a 3×3 convolutional layer connected in sequence;
[0172] The eighth feature unit is used to use a feature fusion layer to perform feature fusion on the fifth feature map and the first current feature map to obtain a second current feature map.
[0173] See also Figure 5 , Figure 5 This is a structural block diagram of a computer device provided in Embodiment 4 of the present invention.
[0174] An electronic device according to an embodiment of the present invention includes: a memory 401 and a processor 402, wherein the memory 402 stores a computer program; when the computer program is executed by the processor 402, the processor 402 executes a circuit breaker fault detection method according to any of the above embodiments.
[0175] The memory 401 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory 401 has a storage space 403 for a program code 413 for executing any method step in the above method. For example, the storage space 403 for the program code may include individual program codes 413 for implementing the various steps in the above method, respectively. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. The program code may be compressed, for example, in an appropriate form. When these codes are run by a computing and processing device, the computing and processing device performs the various steps in the method described above. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. The program code may be compressed, for example, in an appropriate form. When these codes are executed by a computing and processing device, the computing and processing device is caused to execute the various steps in the circuit breaker fault detection method described above.
[0176] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0177] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0178] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0179] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0180] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0181] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A circuit breaker fault detection method, characterized in that: include: Acquiring action data of the circuit breaker in a closed state, and performing data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram; Inputting the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network; Performing feature extraction on the sound feature graph through the first extraction network to obtain a sound extraction feature graph; Performing feature extraction on the current feature graph through the second extraction network to obtain a current extraction feature graph; The detection network is used to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
2. The circuit breaker fault detection method according to claim 1, characterized in that: The action data includes sound data and current data, and the step of performing data preprocessing on the action data to obtain a sound characteristic graph and a current characteristic graph includes: Using the current data to construct a closing current curve, and extracting closing current information during a current valley period in the closing current curve; Extracting effective sound signals from the sound data according to a preset characteristic time segment; The effective sound signal and the closing current information are respectively processed into images to obtain a sound image and a current characteristic diagram.
3. The circuit breaker fault detection method according to claim 1, characterized in that: The first extraction network includes a convolutional layer, a first residual module, a second residual module, a pooling layer and an LSTM layer connected in sequence, and the step of extracting features from the sound feature map through the first extraction network to obtain the sound extraction feature map includes: Extracting features from the sound feature map through a convolutional layer to obtain a first sound feature map; Using the first residual module to extract features from the first sound feature map to obtain a second sound feature map; wherein the first residual module includes a first extraction branch, a second extraction branch and a feature fusion layer; Performing feature extraction on the second sound feature map through the second residual module to obtain a third sound feature map; wherein the second residual module includes a third extraction branch, a fourth extraction branch and a feature fusion layer; Performing a pooling operation on the third sound feature map through a pooling layer to obtain a fourth sound feature map; The LSTM layer is used to perform feature extraction on the fourth sound feature map to obtain a sound extraction feature map.
4. The circuit breaker fault detection method according to claim 1, characterized in that: The second extraction network includes a convolutional layer, a third residual module, a pooling layer and an LSTM layer connected in sequence, and the step of extracting features from the current feature map through the second extraction network to obtain the current extraction feature map includes: Extracting features from the current characteristic graph through a convolution layer to obtain a first current characteristic graph; The third residual module is used to perform feature extraction on the first current characteristic map to obtain a second current characteristic map; wherein the third residual module includes a fifth extraction branch and a feature fusion layer; Performing a pooling operation on the second current characteristic map through a pooling layer to obtain a third current characteristic map; The LSTM layer is used to extract features from the third current feature map to obtain a current extraction feature map.
5. The circuit breaker fault detection method according to claim 1, characterized in that: The detection network includes a feature fusion layer, an extraction module, and an activation layer connected in sequence. The step of using the detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker includes: Using a feature fusion layer to perform feature fusion on the sound extraction feature map and the current extraction feature map to obtain an information fusion feature map; Performing feature extraction on the information fusion feature map through an extraction module to obtain an information feature map; wherein the extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, a random drop layer and a fully connected layer connected in sequence; An activation layer is used to perform fault detection on the information characteristic graph to obtain a fault detection result corresponding to the circuit breaker.
6. The circuit breaker fault detection method according to claim 3, characterized in that: The step of extracting features from the first sound feature map using the first residual module to obtain a second sound feature map includes: Performing feature extraction on the first sound feature map through the first extraction branch to generate a first feature map, wherein the first extraction branch includes a 3×3 convolution layer, a batch normalization layer, a DY-Relu activation layer, a 3×3 convolution layer, a batch normalization layer, and a DY-Relu activation layer connected in sequence; Performing feature extraction on the first sound feature map through the second extraction branch to generate a second feature map, wherein the second extraction branch includes a 3×3 convolutional layer, a batch normalization layer, and a Relu activation layer connected in sequence; A feature fusion layer is used to perform feature fusion on the first feature map and the second feature map to obtain a second sound feature map.
7. The circuit breaker fault detection method according to claim 3, characterized in that: The step of extracting features from the second sound feature map by using the second residual module to obtain a third sound feature map includes: Performing feature extraction on the second sound feature map through the third extraction branch to obtain a third feature map; wherein the third extraction branch includes a batch normalization layer, a Relu activation layer, a 3×1 convolution layer, a batch normalization layer, a Relu activation layer, a 1×3 convolution layer and a global average pooling layer connected in sequence; Performing feature extraction on the second sound feature map through the fourth extraction branch to obtain a fourth feature map; wherein the fourth extraction branch includes a 3×3 convolution layer, a batch normalization layer, and a Swish activation layer connected in sequence; A feature fusion layer is used to perform feature fusion on the third feature map and the fourth feature map to obtain a third sound feature map.
8. The circuit breaker fault detection method according to claim 4, characterized in that: The step of using the third residual module to extract features from the first current characteristic graph to obtain a second current characteristic graph includes: Performing feature extraction on the first current feature map through the fifth extraction branch to obtain a fifth feature map; wherein the fifth extraction branch includes a batch normalization layer, a Swish activation layer, a 3×3 convolutional layer, a batch normalization layer, a Swish activation layer, and a 3×3 convolutional layer connected in sequence; A feature fusion layer is used to perform feature fusion on the fifth feature map and the first current feature map to obtain a second current feature map.
9. A circuit breaker fault detection system, characterized in that: include: The acquisition module is used to obtain the action data of the circuit breaker in the closed state, and perform data preprocessing on the action data to obtain a sound characteristic diagram and a current characteristic diagram; An input module, used for inputting the sound characteristic graph and the current characteristic graph into a pre-trained fault detection model, wherein the fault detection model includes a first extraction network, a second extraction network and a detection network; A first extraction module, configured to perform feature extraction on the sound feature map through the first extraction network to obtain a sound extraction feature map; A second extraction module, used for performing feature extraction on the current characteristic graph through the second extraction network to obtain a current extraction characteristic graph; The detection module is used to use the detection network to perform fault detection on the sound extraction feature graph and the current extraction feature graph to obtain a fault detection result corresponding to the circuit breaker.
10. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the circuit breaker fault detection method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Automatic identification method and device for mechanical fault of high-voltage switch
CN111796180A
Power distribution system, high-voltage circuit breaker fault detection model and method
CN114841222A
Method for processing CT (Computed Tomography) reconstructed image
CN117726754A
Method and device for generating posture of object in image
CN117894038A
Cited By
Online monitoring method and system for fault state of high-voltage parallel cable line
CN119535066A
