CN117197554A - Transformer oil leakage real-time detection method and system - Google Patents
Transformer oil leakage real-time detection method and system Download PDFInfo
- Publication number
- CN117197554A CN117197554A CN202311132418.3A CN202311132418A CN117197554A CN 117197554 A CN117197554 A CN 117197554A CN 202311132418 A CN202311132418 A CN 202311132418A CN 117197554 A CN117197554 A CN 117197554A
- Authority
- CN
- China
- Prior art keywords
- oil leakage
- hyperspectral image
- outlier
- band
- transformer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000011897 real-time detection Methods 0.000 title claims abstract description 26
- 238000001514 detection method Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 48
- 230000006835 compression Effects 0.000 claims abstract description 25
- 238000007906 compression Methods 0.000 claims abstract description 25
- 230000009467 reduction Effects 0.000 claims abstract description 21
- 230000003595 spectral effect Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000002708 enhancing effect Effects 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 238000012795 verification Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004868 gas analysis Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 238000002310 reflectometry Methods 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000701 chemical imaging Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001373 regressive effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Landscapes
- Examining Or Testing Airtightness (AREA)
Abstract
The application provides a transformer oil leakage real-time detection method and a system, which relate to the technical field of fault detection, and specifically comprise the following steps: inputting the hyperspectral image acquired in real time into a trained oil leakage detection model to obtain an oil leakage detection result; the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands; the method based on hyperspectral analysis realizes high-efficiency compression of hyperspectral image characteristic wave bands by using a method for rapidly selecting outlier wave bands, and realizes real-time detection of oil leakage faults of a transformer by using a YOLOv8 network as an oil leakage detection model by combining a deep learning method.
Description
Technical Field
The application belongs to the technical field of fault detection, and particularly relates to a transformer oil leakage real-time detection method and system.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The power transformer is used as main equipment in a transformer substation, the running condition of the power transformer affects the whole power system at any time, the transformer working for a long time is easy to have oil leakage problem, and the running fault of the power system can be caused when the oil leakage problem is serious, so that the property and personal safety are endangered.
The existing oil leakage detection of the power transformer mainly comprises the modes of visual inspection, oil leakage alarm device, oil temperature monitoring, gas analysis method, ultrasonic detection and the like. However, these methods have some application limitations. Visual inspection is susceptible to subjective judgment and negligence and is insensitive to tiny or hidden oil leakage; the oil leakage alarm device can only detect the oil level rise, and cannot determine the specific oil leakage position and reason; oil temperature monitoring may be interfered by other factors and is insensitive to minor oil leakage; the gas analysis method requires special equipment, has higher cost and has limited detection on tiny oil leakage; ultrasonic detection is interfered by environmental noise, and has limited hidden oil leakage; in order to improve accuracy and efficiency, a new method for detecting oil leakage of a power transformer is required to be searched.
Disclosure of Invention
In order to overcome the defects of the prior art, the application provides a transformer oil leakage real-time detection method and system, which are based on a hyperspectral analysis method, realize high-efficiency compression of hyperspectral image characteristic wave bands by using an outlier wave band rapid selection method, and realize real-time detection of transformer oil leakage faults by using a YOLOv8 network as an oil leakage detection model in combination with a deep learning method.
To achieve the above object, one or more embodiments of the present application provide the following technical solutions:
the application provides a transformer oil leakage real-time detection method.
A real-time detection method of oil leakage of a transformer comprises the steps of inputting hyperspectral images acquired in real time into a trained oil leakage detection model to obtain an oil leakage detection result;
the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands;
the method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the integral energy mean value of all wave bands, and the energy deviation degree of the hyperspectral image is calculated, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band.
Further, the transformer oil leakage hyperspectral image set comprises the following specific acquisition methods:
using a hyperspectral camera to acquire spectral information of a transformer device image with a preset number of wavelengths in a preset spectral range, wherein the spectral resolution meets preset conditions;
and acquiring hyperspectral original photo data of oil leakage of at least a preset number.
Further, the method also comprises the enhancement operation of the transformer oil leakage hyperspectral image set, and the diversity of the transformer oil leakage hyperspectral image set is increased through the operations of cutting, rotating, random scaling and translation.
Further, the dimension reduction compression of the hyperspectral image specifically includes:
(1) Calculating the band energy value of the hyperspectral image;
energy value of kth band of hyperspectral imageThe calculation is as follows:
wherein x (i, j) represents the value of the image at the pixel point (i, j), the total wave band number of the hyperspectral image is L, the width of the image is W, and the height of the image is H;
(2) Calculating the overall energy mean value of all wave bands of hyperspectral imageThe formula is as follows:
wherein L represents the total band number of the hyperspectral image,an energy value representing an i-th band;
(3) Calculating energy deviation value E of hyperspectral image k The formula is:
wherein,represents the energy value of the kth band, +.>Representing the overall energy mean of all bands;
respectively calculating energy deviation values for L wave bands to obtain an energy deviation degree set { E } of the hyperspectral image 1 ,E 2 ,…,E k ,…,E L }；
(4) And determining an outlier band of the hyperspectral image, and performing dimension reduction compression on the hyperspectral image by reserving the outlier band.
Further, determining an outlier band of the hyperspectral image, and performing dimension reduction compression on the hyperspectral image by reserving the outlier band specifically comprises:
energy deviation degree set { E for hyperspectral image 1 ,E 2 ,…,E k ,…,E L Sequencing, wherein a wave band with a front preset proportion and a wave band with a rear preset proportion are used as outlier wave bands to form a compressed hyperspectral image only retaining discrete wave bands.
Further, the oil leakage detection model is based on YOLOv8, and comprises three parts: backbone, neck and head;
and (3) performing a detection task on the 30-band compressed hyperspectral image, modifying the YOLOv8 channels, and changing the conventional 3 channels into the number of channels consistent with the number of the bands of the outlier bands, wherein the channels are used for extracting the characteristics of each outlier band and detecting the targets.
Further, in the training process, the F1 fraction, mAP, the accuracy rate and the recall rate are used as evaluation indexes of the model, the F1 fraction is defined as a harmonic mean of the accuracy rate and the recall rate, the mAP is defined as an average value of curve area values enclosed by the accuracy rate and the recall rate in all categories, the accuracy rate is defined as the ratio of a real oil leakage area in a sample detected as oil leakage, and the recall rate is the ratio of the detected oil leakage area samples.
The application provides a transformer oil leakage real-time detection system.
The transformer oil leakage real-time detection system comprises a model training module and a real-time detection module:
the real-time detection module is configured to: inputting the hyperspectral image acquired in real time into a trained oil leakage detection model to obtain an oil leakage detection result;
the model training module is configured to: the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands;
the method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the integral energy mean value of all wave bands, and the energy deviation degree of the hyperspectral image is calculated, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band.
A third aspect of the present application provides a computer readable storage medium having stored thereon a program which when executed by a processor performs steps in a method for detecting oil leakage of a transformer according to the first aspect of the present application.
A fourth aspect of the present application provides an electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, the processor implementing the steps in a method for detecting oil leakage of a transformer according to the first aspect of the present application when executing the program.
The one or more of the above technical solutions have the following beneficial effects:
according to the application, the hyperspectral image is utilized to detect the transformer oil leakage target, the morphological characteristics of the detected target are broken through, and the physical structure and chemical composition difference of the target are detected, so that the accuracy and efficiency of transformer oil leakage detection are improved.
The application provides a method for rapidly selecting outlier wavebands, which is used for selectively compressing wavebands of a hyperspectral image based on the deviation value of the waveband energy value of the hyperspectral image and the overall energy mean value of all wavebands, and solves the system performance problem caused by large data volume and redundant information of original hyperspectral data.
The application reforms the YOLOv8, and ensures that the channel number is consistent with the band number of the outlier bands, thereby ensuring that the feature extraction and the target detection are fully carried out on each outlier band, and improving the accuracy of oil leakage detection of the transformer.
Additional aspects of the application will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application.
Fig. 1 is a flow chart of a method of a first embodiment.
Fig. 2 is a block diagram of an oil leakage detection model according to the first embodiment.
Fig. 3 (a), (b), (c), (d) and (e) are the curves of box_loss, cls_loss, dfl _loss, training accuracy, mAP value, recall, mAP95 value, respectively, of the model on the training set and the validation set during the training of the first embodiment.
FIGS. 4 (a) and (b) are the F1 score curve and the P-R curve, respectively, during training of the first embodiment.
Fig. 5 (a) and (b) are schematic diagrams of result prediction and detection region merging on a test set for the model of the first embodiment, respectively.
Fig. 6 is a system configuration diagram of the second embodiment.
Detailed Description
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Example 1
In one or more embodiments, a transformer oil leakage real-time detection method is disclosed, and hyperspectral images acquired in real time are input into a trained oil leakage detection model to obtain an oil leakage detection result.
The following describes the implementation process of a transformer oil leakage real-time detection method in detail from the aspects of image acquisition, image compression, model training and real-time detection, as shown in fig. 1, and includes the following steps:
step S1: aiming at a transformer oil leakage scene, a hyperspectral camera is used for collecting a plurality of hyperspectral images, and the hyperspectral images are subjected to enhancement operation to obtain a transformer oil leakage hyperspectral image set.
The hyperspectral imaging technology combines imaging and spectrum information, can acquire space information and continuous narrow-band spectrum information of a target at the same time, and draws target reflectivity in all wavelength ranges according to different reflectivities of different substances so as to acquire spectrum characteristics of the target, so that the target detection by using hyperspectral images can break through the morphological characteristics of the detected target, and physical structure and chemical composition differences of the target are detected.
Using a hyperspectral camera to acquire spectral information of 300 wavelengths of an image of the transformer equipment in a spectral range of 400 nm-1000 nm, wherein the spectral resolution is higher than 2.18nm; it is recommended to obtain not less than 1000 pieces of hyperspectral raw photograph data of leaked oil.
In consideration of the problem that the model overfitting easily occurs due to the small number of data set samples of the later oil leakage detection model, data enhancement operations such as cutting, rotating, random scaling, translation and the like can be performed on hyperspectral original photo data, so that diversity of the data sets is increased, model training effects are guaranteed, and finally hyperspectral data of about 5000 oil leakage are obtained.
Step S2: and performing dimension reduction compression on the hyperspectral image in the transformer oil leakage hyperspectral image set based on the outlier band rapid selection method to obtain a training data set.
The method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the overall energy mean value of all wave bands, and calculates the energy deviation degree of the hyperspectral image, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band, and specifically comprises the following steps:
(1) Calculating the band energy value of the hyperspectral image;
each hyperspectral image is composed of multiple wave band images, and a set { X } for hyperspectral images is arranged 1 ,X 2 ,…,X k ,…,X L Expressed by X k Representing an image of the kth band, then the energy value of the kth band of the hyperspectral imageThe calculation is as follows:
where x (i, j) represents the pixel value of the image at the pixel point (i, j), the total band number of the hyperspectral image is L, the width of the image is W, and the height of the image is H.
Respectively calculating energy values of the images of the L wave bands to obtain a hyperspectral image energy value set
(2) Calculating the overall energy mean value of all wave bands of hyperspectral imageThe formula is as follows:
wherein L represents the total band number of the hyperspectral image,representing the energy value of the i-th band.
(3) Calculating energy deviation value E of hyperspectral image k The formula is:
wherein,represents the energy value of the kth band, +.>Representing the overall energy mean of all bands.
The energy deviation value of the k-th wave band belongs to positive deviation if the value is regular energy higher than the average value, and belongs to negative deviation if the value is negative energy lower than the average value.
Calculating energy bias for L wave bandsThe separation value is used for obtaining an energy deviation degree set { E (E) of the hyperspectral image 1 ,E 2 ,…,E k ,…,E L }。
(4) Determining an outlier band of the hyperspectral image, and performing dimension reduction compression on the hyperspectral image by reserving the outlier band, wherein the method specifically comprises the following steps:
energy deviation degree set { E for hyperspectral image 1 ,E 2 ,…,E k ,…,E L Sorting, wherein the first 5% of the bands and the last 5% of the bands are used as outlier bands, if the hyperspectral image before compression has 300 bands, the outlier bands are 30 bands, and a compressed hyperspectral image only retaining the 30 discrete bands is formed
Step S3: and training an oil leakage detection model constructed based on YOLOv8 by using the training data set, wherein the number of channels of the oil leakage detection model is consistent with the number of bands of the outlier bands.
Yolov8 is a currently mainstream oil leakage detection model, which is the latest version of YOLO (You Only Look Once) series models; YOLOv8 adopts dark net53 as a backbone network and improves on the basis; the entire model structure includes three parts: backbone, neck and head, wherein the backbone is typically a pre-trained convolutional neural network, such as ResNet, darkNet, etc.; the neg is a feature enhancement module used for further processing the feature map extracted by the backstone; the head comprises a classifier and a regressive device for classifying and regressing the detection frame.
The backbond of the YOLOv8 adopts the DarkNet53, and the network structure has higher accuracy and good speed performance; the neg adopts a spatial pyramid pooling network and a path aggregation network, so that the feature representation capability can be enhanced, and the accuracy of target detection is improved; the head adopts an anchor-based detection mechanism similar to that of YOLOv3, and simultaneously introduces an adaptive fusion strategy and multi-scale training, so that the performance of the model is further improved.
On the basis, in order to adapt to the detection task of the compressed hyperspectral images of 30 outlier bands in the embodiment, the channels of the YOLOv8 are modified, and the channels are changed into 30 channels from 3 channels, so that the training and actual detection tasks of the hyperspectral images are realized, and finally, the oil leakage detection model shown in fig. 2 is obtained.
After determining the structure of the oil leakage detection model, training the model is needed to adapt the model to the oil leakage hyperspectral related data. The preparation work is as follows: the hyperspectral data in the training dataset were labeled using LabelImage software and the dataset was labeled with 1:2:7, the validation set, the test set and the training set are divided. Setting relevant training parameters, wherein the size of batch size for each iterative training is 4, the learning rate is 0.001, the category number is 2, and selecting a yolov8n pre-training model trained on a large-scale public data set to perform migration learning. After the relevant parameters are set, model training is started.
In the training process, F1 fraction (F1-Score), mAP, accuracy (precision) and recall (recall) are used as evaluation indexes of the model, wherein the F1 fraction is defined as a harmonic average of the accuracy and the recall, the mAP is defined as an average value of curve area values defined by the accuracy and the recall of all categories, the accuracy is defined as a ratio of a real oil leakage area in a sample detected as oil leakage, the recall is a ratio of the detected oil leakage area sample, and the larger the four evaluation index values are, the better the recognition effect of the representative model is, the calculation formula is as follows:
TP represents an oil leakage area, and oil leakage is detected; FP denotes that the target is not an oil leakage area, and is detected as oil leakage; FN indicates that the target is an oil leak region, and the detection is not oil leak.
The loss function of the YOLOv8 is divided into three parts, namely cls_loss, box_loss and dfl_loss, wherein the cls_loss is used for supervising classification, the box_loss is used for supervising regression of a detection frame, and the dfl _loss is used for supervising whether an object exists in grid; FIG. 3 shows the variation curves of loss rate and accuracy, recall rate and mAP on the training set and verification set during the training process, wherein the curves of box_loss on the training set and verification set of FIG. 3 (a), the defects of cls_loss on the training set and verification set of FIG. 3 (b), the curves of df_l oss on the training set and verification set of FIG. 3 (c), the curves of training accuracy and mAP value of FIG. 3 (d), and the curves of recall rate and mAP95 value of FIG. 3 (e); therefore, as the training times are increased, the network gradually converges, and the precision reaches more than 90%.
Fig. 4 (a) and (b) show an F1 score curve and a P-R curve in the training process, respectively, after the training is finished, the result of the trained model is predicted on the test set, the predicted result is shown in fig. 5 (a), and the detection areas are further combined, as shown in fig. 5 (b), so that the detection of the oil leakage area can be realized.
Step S4: and inputting the hyperspectral image acquired in real time into a trained oil leakage detection model to detect an oil leakage area.
Before being input into a trained oil leakage detection model, the hyperspectral image acquired in real time is subjected to dimension reduction and compression to form a compressed hyperspectral image with only 30 discrete wave bands reserved, and then oil leakage area detection is carried out.
Example two
In one or more embodiments, a transformer oil leakage real-time detection system is disclosed, as shown in fig. 6, including a model training module and a real-time detection module:
the real-time detection module is configured to: inputting the hyperspectral image acquired in real time into a trained oil leakage detection model to obtain an oil leakage detection result;
the model training module is configured to: the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands;
the method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the integral energy mean value of all wave bands, and the energy deviation degree of the hyperspectral image is calculated, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band.
Example III
An object of the present embodiment is to provide a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of a method for real-time detection of oil leakage of a transformer according to an embodiment of the present disclosure.
Example IV
An object of the present embodiment is to provide an electronic apparatus.
The electronic device comprises a memory, a processor and a program stored in the memory and capable of running on the processor, wherein the processor realizes the steps in the transformer oil leakage real-time detection method according to the first embodiment of the disclosure when executing the program.
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. A real-time detection method of oil leakage of a transformer is characterized in that hyperspectral images acquired in real time are input into a trained oil leakage detection model to obtain an oil leakage detection result;
the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands;
the method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the integral energy mean value of all wave bands, and the energy deviation degree of the hyperspectral image is calculated, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band.
2. The method for detecting the oil leakage of the transformer in real time according to claim 1, wherein the method for collecting the hyperspectral image set of the oil leakage of the transformer is as follows:
using a hyperspectral camera to acquire spectral information of a transformer device image with a preset number of wavelengths in a preset spectral range, wherein the spectral resolution meets preset conditions;
and acquiring hyperspectral original photo data of oil leakage of at least a preset number.
3. The method for detecting the oil leakage of the transformer according to claim 1, further comprising the step of enhancing the hyperspectral image set of the oil leakage of the transformer, wherein the diversity of the hyperspectral image set of the oil leakage of the transformer is increased through the operations of cutting, rotating, random scaling and translation.
4. The method for detecting oil leakage of a transformer in real time according to claim 1, wherein the dimension reduction compression of the hyperspectral image is specifically as follows:
(1) Calculating the band energy value of the hyperspectral image;
energy value X of kth wave band of hyperspectral image k The calculation is as follows:
wherein x (i, j) represents the value of the image at the pixel point (i, j), the total wave band number of the hyperspectral image is L, the width of the image is W, and the height of the image is H;
(2) Calculating the overall energy mean value of all wave bands of hyperspectral imageThe formula is as follows:
wherein L represents the total band number of the hyperspectral image,an energy value representing an i-th band;
(3) Calculating energy deviation value E of hyperspectral image k The formula is:
wherein,represents the energy value of the kth band, +.>Representing the overall energy mean of all bands;
respectively calculating energy deviation values for L wave bands to obtain an energy deviation degree set { E } of the hyperspectral image 1 ,E 2 ,…,E k ,…,E L }；
(4) And determining an outlier band of the hyperspectral image, and performing dimension reduction compression on the hyperspectral image by reserving the outlier band.
5. The method for detecting oil leakage of a transformer according to claim 4, wherein the determining an outlier band of the hyperspectral image, and performing dimension reduction compression on the hyperspectral image by reserving the outlier band, specifically comprises:
energy deviation degree set { E for hyperspectral image 1 ,E 2 ,…,E k ,…,E L Sequencing, wherein a wave band with a front preset proportion and a wave band with a rear preset proportion are used as outlier wave bands to form a compressed hyperspectral image only retaining discrete wave bands.
6. The transformer oil leakage real-time detection method as set forth in claim 1, wherein the oil leakage detection model is based on YOLOv8 and comprises three parts: backbone, neck and head;
and (3) performing a detection task on the 30-band compressed hyperspectral image, modifying the YOLOv8 channels, and changing the conventional 3 channels into the number of channels consistent with the number of the bands of the outlier bands, wherein the channels are used for extracting the characteristics of each outlier band and detecting the targets.
7. The method for detecting oil leakage of a transformer according to claim 1, wherein in the training process, an F1 score, an mAP, an accuracy rate and a recall rate are used as evaluation indexes of a model, the F1 score is defined as a harmonic mean of the accuracy rate and the recall rate, the mAP is defined as an average value of curve area values enclosed by the accuracy rate and the recall rate in all categories, the accuracy rate is defined as a ratio of a sample detected as oil leakage to a real oil leakage area, and the recall rate is a ratio of the sample detected as oil leakage area.
8. The transformer oil leakage real-time detection system is characterized by comprising a model training module and a real-time detection module:
the real-time detection module is configured to: inputting the hyperspectral image acquired in real time into a trained oil leakage detection model to obtain an oil leakage detection result;
the model training module is configured to: the oil leakage detection model is used for carrying out dimension reduction compression on the hyperspectral image based on the collected transformer oil leakage hyperspectral image set by a method of fast selecting outlier bands, constructing a training data set and carrying out model training; the number of channels of the oil leakage detection model is consistent with the number of wave bands of the outlier wave bands;
the method for rapidly selecting the outlier wave band is based on the deviation value of the wave band energy value of the hyperspectral image and the integral energy mean value of all wave bands, and the energy deviation degree of the hyperspectral image is calculated, so that the outlier wave band of the hyperspectral image is determined, and the hyperspectral image is subjected to dimension reduction compression by reserving the outlier wave band.
9. An electronic device, comprising:
a memory for non-transitory storage of computer readable instructions; and
a processor for executing the computer-readable instructions,
wherein the computer readable instructions, when executed by the processor, perform the method of any of the preceding claims 1-7.
10. A storage medium, characterized by non-transitory storing computer-readable instructions, wherein the instructions of the method of any one of claims 1-7 are performed when the non-transitory computer-readable instructions are executed by a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311132418.3A CN117197554A (en) | 2023-09-04 | 2023-09-04 | Transformer oil leakage real-time detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311132418.3A CN117197554A (en) | 2023-09-04 | 2023-09-04 | Transformer oil leakage real-time detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117197554A true CN117197554A (en) | 2023-12-08 |
Family
ID=88989857
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311132418.3A Pending CN117197554A (en) | 2023-09-04 | 2023-09-04 | Transformer oil leakage real-time detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117197554A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118209260A (en) * | 2024-05-21 | 2024-06-18 | 南通世睿电力科技有限公司 | Oil leakage monitoring and early warning system for transformer oil storage cabinet |
-
2023
- 2023-09-04 CN CN202311132418.3A patent/CN117197554A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118209260A (en) * | 2024-05-21 | 2024-06-18 | 南通世睿电力科技有限公司 | Oil leakage monitoring and early warning system for transformer oil storage cabinet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112967243B (en) | Deep learning chip packaging crack defect detection method based on YOLO | |
| CN109447977B (en) | Visual defect detection method based on multispectral deep convolutional neural network | |
| CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
| CN105678788B (en) | A kind of fabric defect detection method based on HOG and low-rank decomposition | |
| CN107563433B (en) | Infrared small target detection method based on convolutional neural network | |
| CN101236608B (en) | Human face detection method based on picture geometry | |
| CN109117876A (en) | A kind of dense small target deteection model building method, model and detection method | |
| CN115731164A (en) | Insulator defect detection method based on improved YOLOv7 | |
| CN112766195B (en) | Electrified railway bow net arcing visual detection method | |
| CN103488974A (en) | Facial expression recognition method and system based on simulated biological vision neural network | |
| CN114973032B (en) | Deep convolutional neural network-based photovoltaic panel hot spot detection method and device | |
| CN115953666B (en) | Substation site progress identification method based on improved Mask-RCNN | |
| CN114255403A (en) | Optical remote sensing image data processing method and system based on deep learning | |
| CN113920400A (en) | Metal surface defect detection method based on improved YOLOv3 | |
| CN109359661A (en) | A kind of Sentinel-1 radar image classification method based on convolutional neural networks | |
| CN114240822A (en) | Cotton cloth flaw detection method based on YOLOv3 and multi-scale feature fusion | |
| CN113674225A (en) | Power equipment fault detection method based on convolutional neural network | |
| CN116485802B (en) | Insulator flashover defect detection method, device, equipment and storage medium | |
| CN113327253A (en) | Weak and small target detection method based on satellite-borne infrared remote sensing image | |
| CN117237599A (en) | Image target detection method and device | |
| CN117218545A (en) | LBP feature and improved Yolov 5-based radar image detection method | |
| CN117237736A (en) | Daqu quality detection method based on machine vision and deep learning | |
| CN117115675A (en) | Cross-time-phase light-weight spatial spectrum feature fusion hyperspectral change detection method, system, equipment and medium | |
| Wang et al. | High-Voltage Transmission Line Foreign Object and Power Component Defect Detection Based on Improved YOLOv5 | |
| CN116229182A (en) | Small particle coffee defect green bean detection method based on feature fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |