CN116385954A - High-precision power transmission line abnormality detection method and system - Google Patents
High-precision power transmission line abnormality detection method and system Download PDFInfo
- Publication number
- CN116385954A CN116385954A CN202310113127.3A CN202310113127A CN116385954A CN 116385954 A CN116385954 A CN 116385954A CN 202310113127 A CN202310113127 A CN 202310113127A CN 116385954 A CN116385954 A CN 116385954A
- Authority
- CN
- China
- Prior art keywords
- transmission line
- module
- foreign matter
- power transmission
- model
- 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
- 230000005540 biological transmission Effects 0.000 title claims abstract description 154
- 238000001514 detection method Methods 0.000 title claims abstract description 139
- 230000005856 abnormality Effects 0.000 title claims abstract description 40
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims description 47
- 238000012549 training Methods 0.000 claims description 41
- 230000006870 function Effects 0.000 claims description 27
- 238000010586 diagram Methods 0.000 claims description 18
- 230000007246 mechanism Effects 0.000 claims description 17
- 238000011176 pooling Methods 0.000 claims description 14
- 230000006872 improvement Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 3
- 238000012886 linear function Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 238000013461 design Methods 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 235000005770 birds nest Nutrition 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 235000005765 wild carrot Nutrition 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a high-precision transmission line abnormality detection method and a high-precision transmission line abnormality detection system, which relate to the technical field of transmission line detection and aim at solving the problems that most of the existing transmission line abnormality detection is that a system uploads an acquired transmission line image to a cloud for processing, the acquired transmission line image is unrealistic, the bandwidth occupancy rate and the delay are high, and the using effect is poor. The system for detecting the foreign matters in the power transmission line is reasonable in design, the model is mainly trained in the cloud, and edge reasoning calculation is carried out according to the trained model, so that the bandwidth and the delay of uploading images to the cloud can be greatly reduced, and the using effect is good.
Description
Technical Field
The invention relates to the technical field of transmission line detection, in particular to a high-precision transmission line abnormality detection method and system.
Background
The power transmission line is realized by boosting the electric energy generated by the generator by using a transformer and accessing the electric energy into the power transmission line through control equipment such as a circuit breaker and the like. The transmission line is divided into an overhead transmission line and a cable line; the overhead transmission line is composed of a line pole tower, a wire, an insulator, line hardware, a stay wire, a pole tower foundation, a grounding device and the like, and is erected on the ground. According to the nature of the transmission current, power transmission is classified into ac power transmission and dc power transmission. The smoothness of the transmission line is one of the basic guarantees of industrial development and people living, the foreign object of the transmission line is one of the main reasons for causing the failure of the transmission line, and in order to detect the transmission line and prevent potential safety hazards and unnecessary economic loss caused by line abnormality, a high-precision transmission line abnormality detection method is needed.
However, in the existing power transmission line abnormality detection, most of the power transmission line images acquired by the system are uploaded to the cloud for processing, and this mode is not practical, so that the bandwidth occupancy rate and the delay are relatively high, and the using effect is poor.
Disclosure of Invention
The invention aims to solve the defects that the bandwidth occupancy rate and the delay are high and the using effect is poor due to the fact that most of the existing power transmission line abnormality detection is that a system uploads the acquired power transmission line image to a cloud for processing, and the method and the system for detecting the power transmission line abnormality are provided.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
in a first aspect, the present invention provides a method for detecting an abnormality of a power transmission line, including:
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
In some embodiments, the transmission line abnormality detection method adopts a transmission line abnormality detection model based on a YOLOv4 network, and the transmission line abnormality detection model comprises a feature extraction network module, a feature enhancement module and a detection module.
Further, in some embodiments, in the power transmission line abnormality detection model based on the YOLOv4 network, the CSPDarknet53 module in the YOLOv4 network is replaced by an improved Ghost module, the feature extraction layers of the trunks of the YOLOv4 network are respectively set at the 5 th, 11 th and 16 th layers, and the bottleneck layer of each GhostNetBottleNeck contains two GhostModule modules, which are respectively used for expanding and reducing the number of channels and ensuring that the number of channels connected with the input can be well provided.
In some embodiments, the improved Ghost module comprises:
introducing a GhostNet network into the foreign object feature extraction module, generating a real feature layer from input features by utilizing convolution operation of the GhostNet network, obtaining the Ghost feature layer by deep convolution DW processing on the generated real feature layer, and splicing the obtained real feature layer and the Ghost feature layer to obtain an output feature layer;
the input transmission line characteristic diagram is m multiplied by n multiplied by cin, the output characteristic diagram is m multiplied by n multiplied by cout, the whole input layer is divided into s parts, the convolution kernel size is set as k, and the calculated amount through the conventional convolution operation is as follows:
m'×n'×cout×k×k×cin
the calculation amount of the whole Ghost module is expressed as:
after conventional convolution and depth convolution, the compression rate of the whole model is changed into s, so that the calculated amount of the whole model is greatly reduced;
in order to further improve the performance of the Ghost module, an ECA mechanism is introduced to replace an original SE mechanism on the basis of ensuring that the calculated amount is not increased; the ECA mechanism is mainly a channel attention model and is used for carrying out global average pooling on the feature map m multiplied by n multiplied by c, so that a vector with the size of 1 multiplied by c is obtained, and information interaction among channels is realized through one-dimensional convolution operation; the size k' of the convolution kernel in a one-dimensional convolution operation is determined mainly from an adaptive function:
where Θ (C) represents a modulus of the vector for C, a and δ represent coefficients of linear functions, respectively, and m () represents a function for finding the nearest odd number.
In some embodiments, an improved SPP module is adopted in the power transmission line abnormality detection model based on the YOLOv4 network, and average pooling is introduced to replace one of the maximum pooling so as to complete the improvement of the SPP module; and the foreign matter and the background information are distinguished by using the average pooling operation, more foreign matter information is reserved, and the accuracy of foreign matter detection is improved.
In some embodiments, the total Loss function Loss of the power transmission line anomaly detection model based on the YOLOv4 network Total (S) The method comprises the following steps:
Loss total (S) ＝Loss Regression +Loss Confidence level +Loss Classification
The total Loss function includes regression error Loss Regression Confidence error Loss Confidence level Classification error Loss Classification ；
Defining a power transmission line foreign matter detection image with m×n size, dividing the image into d×d grids, p representing the number of boundary frames to be predicted in the divided grids, Q and Q' representing actual boundary frames and prediction boundary frames, respectively, and regression error Loss Regression The method comprises the following steps:
wherein I is i,j Represents the size of the jth frame region of the ith frame, lambda represents a constant,
definition P i,j A rough representation of a foreign object-free frame in the ith frameThe rate of the product is determined by the ratio,
confidence coefficients of the actual boundary frame and the prediction boundary frame are respectively obtained, and a confidence coefficient error Loss is obtained according to a cross entropy formula Confidence level ：
Where mu represents a set constant value,
defining using cross entropy formula
And->
The probability of foreign object type in the actual boundary box and the predicted boundary box respectively, and the classification error Loss Classification The method comprises the following steps:
where c ε f represents the vector c in the f range, where f is the value found from the image.
In a second aspect, the invention provides a power transmission line abnormality detection system, which is based on a cloud-edge fusion technology and comprises an image acquisition module, an edge end module and a cloud service module;
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
In a third aspect, the present invention provides a transmission line abnormality detection apparatus, including a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to the first aspect.
In a fourth aspect, the present invention provides a storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method of the first aspect.
The beneficial effects are that: the invention provides a transmission line anomaly detection method and a transmission line anomaly detection system, wherein the detection of a transmission line anomaly target is the guarantee of ensuring the safety and the normal operation of a transmission line, and the transmission line anomaly target efficient detection technology based on an improved YOLOv4 network is provided on the basis of a cloud-edge intelligent technology aiming at the problems of low detection precision and low speed of the traditional transmission line:
the cloud-edge intelligent technology is fused, a foreign matter detection system framework of the power transmission line is designed, and abnormal target efficient detection of the power transmission line is realized;
the ECA mechanism replaces an SE mechanism in the GhostNet network to complete GhostNet network improvement, and the improved GhostNet network is utilized to improve a feature extraction module of the YOLOv4 model on the basis, so that the information calculation amount is reduced, and the foreign object detection and recognition speed is accelerated.
And the average pooling is introduced to replace one of the maximum pooling to finish the improvement of the SPP module, so that more foreign matter information can be ensured to be always reserved in the process of feature scale fusion. The convergence speed of the model can be further improved while the detection precision is improved;
optimizing a YOLOv4 model loss function, and reducing the problem of low detection precision caused by color or external textures;
the method is verified through the actual power transmission line data set picture, the detection and identification accuracy of the detected foreign matters is 99.2%, and the detection speed is 218ms, so that the method has good power transmission line foreign matters detection capability and meets the requirement of real-time detection;
the system for detecting the foreign matters in the power transmission line is reasonable in design, the model is mainly trained in the cloud, and edge reasoning calculation is carried out according to the trained model, so that the bandwidth and the delay of uploading images to the cloud can be greatly reduced, and the using effect is good.
Drawings
Fig. 1 is a system architecture of a foreign matter detection system of a power transmission line based on a cloud-edge fusion technology, which is a high-precision power transmission line abnormality detection method provided by the invention;
fig. 2 is a power transmission line foreign matter detection model based on yolov4 network model of the high-precision power transmission line anomaly detection method provided by the invention;
fig. 3 is a schematic structural diagram of a Ghost module of the high-precision transmission line abnormality detection method according to the present invention;
fig. 4 is a schematic diagram of replacing an SE mechanism by an ECA mechanism of the high-precision transmission line abnormality detection method provided by the invention;
FIG. 5 is a Ghost module based on ECA mechanism improvement of the high-precision transmission line abnormality detection method provided by the invention;
fig. 6 is a schematic diagram of a Yolov4 network structure of a converged GhostNet network in the high-precision transmission line abnormality detection method provided by the invention;
FIG. 7 is a schematic diagram showing an improvement of an SPP module of a method for detecting abnormality of a high-precision power transmission line according to the present invention;
FIG. 8 is a comparison of transmission line foreign matter detection results based on different models according to the high-precision transmission line anomaly detection method provided by the invention;
fig. 9 is a graph showing the average time of foreign object identification of a power transmission line based on different models according to the high-precision power transmission line abnormality detection method provided by the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
In the description of the present invention, the descriptions of the terms "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Example 1
A transmission line abnormality detection method includes:
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
Referring to fig. 1-9, one embodiment provided by the present solution: the utility model provides a high accuracy transmission line anomaly detection method, includes the transmission line foreign matter detecting system of cloud-limit integration technique, as shown in fig. 1, the transmission line foreign matter detecting system of cloud-limit integration technique includes image acquisition module, edge end module and high in the clouds service module;
the image acquisition module installed on the power transmission line continuously acquires images and transmits the images to the edge end module, the cloud end carries out foreign matter detection model training, after the training process is completed, the cloud end unloads the model with parameters to the edge, detection and calculation can be carried out on the edge, the trained edge end firstly detects the images uploaded by the nearby image acquisition module, only the images with abnormal objects are sent to the cloud end together with alarm information, the images are reversely used as a new sample for further model training, namely so-called incremental learning, and then the model parameters stored at the edge are correspondingly updated.
In this embodiment, a high-precision transmission line anomaly detection method includes a transmission line anomaly detection model, where the transmission line anomaly detection model is mainly divided into a one-stage detection method and a two-stage detection method when a neural network is used to detect transmission line anomalies, and a YOLOv4 network is a typical one-stage detection network, and based on a YOLOv4 network design, the transmission line anomaly detection model is mainly divided into a feature extraction network module, a feature enhancement module and a detection module, and fig. 2 is a transmission line anomaly detection model based on the YOLOv4 network in the high-precision transmission line anomaly detection modeling method according to this embodiment.
In this embodiment, in order to reduce the calculation amount, deploy fewer parameters and accelerate the foreign object detection when extracting the characteristics of the foreign object picture of the power transmission line, a GhostNet network is introduced to improve the foreign object characteristic extraction module, and FIG. 3 is a schematic structural diagram of a Ghost module of a high-precision power transmission line anomaly detection modeling method; the GhostNet network is adopted to reserve basic characteristic information and calculate less redundant complex information, the GhostNet network generates partial real characteristic layers of the input characteristics mainly by means of convolution operation, the generated real characteristic layers are processed by DW on the basis to obtain the Ghostcharacteristic layers, and finally the obtained real characteristic layers and the Ghostcharacteristic layers are spliced to output a final complete output characteristic layer;
assuming that the input transmission line characteristic diagram is m×n×cin, the output characteristic diagram is m×n×cout, the whole input layer is divided into s-divisions, and the convolution kernel size is set to k, the calculated amount through the conventional convolution operation at this time is:
m'×n'×cout×k×k×cin
the computational effort of the entire Ghost module can be expressed as:
from the above formula, it can be seen that after conventional convolution and deep convolution, the compression rate of the whole model becomes s, so that the calculated amount of the whole model is greatly reduced, and in order to further improve the performance of the Ghost module, an ECA mechanism is introduced to replace an original SE mechanism on the basis of ensuring that the calculated amount is not increased; fig. 4 is a schematic diagram of replacing an SE mechanism by an ECA mechanism of the modeling method for detecting the abnormality of the high-precision power transmission line according to the present embodiment;
FIG. 5 is a Ghost module based on ECA mechanism improvement; the ECA mechanism is mainly a channel attention model, and can carry out global average pooling on a feature map m multiplied by n multiplied by c, so as to obtain a vector with the size of 1 multiplied by c, and then information interaction among channels is realized through one-dimensional convolution operation. The size of the convolution kernel in a one-dimensional convolution operation is determined mainly from an adaptive function:
where Θ (C) represents a modulus of the vector for C, a and δ represent coefficients of linear functions, respectively, and m () represents a function for finding the nearest odd number.
In this embodiment, as shown in fig. 6, the improved Ghost module is replaced with the CSPDarknet53 module in the Yolov4 network, and the feature extraction layer of the trunk of the whole network is set to 5 th, 11 th and 16 th layers respectively, which can be seen that the bottleneck layer of each Ghost netbowtleneck mainly includes two Ghost modules, and the two modules are used for expanding and reducing the number of channels respectively, and ensuring that the number of channels connected with the input can be well provided.
In this embodiment, in the process of detecting the foreign matter in the power transmission line, the calculation amount of the feature map can be reduced by downsampling the feature map in a mode of maximizing pooling, so that the purpose of improving the training speed of the network is achieved, the problem of overfitting existing in the neural network can be prevented by the operation of maximizing pooling, the most interesting area in the feature map is obtained, and the foreign matter and the background information are distinguished by the operation of averaging pooling because the foreign matter detection in the power transmission line possibly has the similar problem of the foreign matter and the background detection, and meanwhile, more foreign matter information is reserved, so that the accuracy of the foreign matter detection is improved.
In this embodiment, the loss function of the yolov4 network is improved, and the improved algorithm is that the function mainly includes three parts, namely a classification error, a confidence coefficient and a regression error. Assuming that the size of one transmission line foreign matter detection image is m×n, dividing this image into grids of d×d on average, and p represents the number of bounding boxes to be predicted among the divided grids, Q and Q' represent actual and predicted bounding boxes, respectively, the formula of regression error can be expressed as:
wherein I is i,j Represents the size of the jth frame region of the ith frame, lambda represents a constant,
Assuming that a and a 'are the confidence levels of the actual and predicted bounding boxes, pi, j represent the probability of not containing foreign object objects in the ith frame, j's frame, then the confidence error formula that is possible according to the cross entropy formula is:
where mu represents a set constant value,
defining using cross entropy formula
And->
The probability of foreign object type in the actual boundary box and the predicted boundary box respectively, and the classification error Loss Classification The method comprises the following steps:
where c ε f represents the vector c in the f range, where f is the value found from the image.
The total loss function of the network, which can be improved according to the equation, can be defined as:
Loss total (S) ＝Loss Regression +Loss Confidence level +Loss Classification 。
To verify the performance of the method presented herein, transmission line anomaly targets are employed as test and training sets. The superiority of the methods presented herein was verified by comparison experiments with four methods that are currently more common.
(1) Comparative analysis of transmission line identification accuracy
The accuracy of the foreign matter identification of the transmission line by the method provided herein is compared with the identification accuracy of the conventional method at present, and the specific effect is shown in fig. 8. It can be seen from the figure that after the method proposed herein is improved twice on the basis of the traditional Yolov4 network model, the average accuracy of the obtained foreign matter identification is up to about 99%, which is improved by about 15% compared with the traditional Yolov4 network before improvement, meanwhile, compared with other models in the same stage, the method proposed herein still has higher identification accuracy, the method proposed herein aims at the common six foreign matter situations, wherein the identification accuracy of a plastic bag on a power transmission line is 97.95%, the identification accuracy of the plastic bag is lower than that of other four foreign matters, and particularly, the identification accuracy after improvement can be seen, compared with the Yolov4 before improvement, the identification accuracy of the whole model is obviously improved by about 20%, 17% and 21% respectively. The average accuracy of the method for identifying six common foreign matters of the power transmission line is shown in table 3, and as can be seen from the table, the accuracy for identifying the foreign matters of the bird nest is highest, and the accuracy for identifying the plastic bag is lowest.
TABLE 3 Table 3
(2) Transmission line identification speed comparison analysis
In order to verify the recognition speed of the transmission line based on the cloud edge technology, the recognition speeds of the six different models are compared with the recognition speeds of the common models, the recognition time of the six different models on the foreign matter recognition of the transmission line is shown in fig. 9, the recognition speed of the method is 218ms, and the recognition speed is increased by 71% compared with that of the YOLOv4 network before improvement. Meanwhile, the recognition speeds of the method on different foreign matters are compared, and particularly the recognition speeds of different foreign matters can be clearly seen from the table, the time spent in the recognition process of the tower crane and the bird nest is minimum, and the time required for recognizing the smoke and the fire is maximum.
(3) Comparison results with different loss functions
To further verify the effect of the loss function presented herein, the loss function presented herein is compared to the currently common MSE, CIOU, focal loss function, which ultimately has the effect on the accuracy of transmission line target detection as shown in table 4. As can be seen from the data in the table, when the foreign matter of the transmission line is detected, the identification accuracy of the CIOU loss function is about 11% higher than that of the Focal function. When the loss function is adopted, the final average recognition accuracy is about 10% higher than that of the CIOU loss function, and the effectiveness of the loss function is fully verified.
TABLE 4 Table 4
Example 2
In a second aspect, the invention provides a power transmission line abnormality detection system, which is based on a cloud-edge fusion technology and comprises an image acquisition module, an edge end module and a cloud service module;
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
Example 3
In a third aspect, the present embodiment provides a transmission line abnormality detection apparatus, including a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is operative according to the instructions to perform the steps of the method according to embodiment 1.
Example 4
In a fourth aspect, the present embodiment provides a storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method described in embodiment 1.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
The preferred embodiments of the invention disclosed above are intended only to assist in the explanation of the invention. The preferred embodiments are not exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. The invention is limited only by the claims and the full scope and equivalents thereof.
Claims (9)
1. The power transmission line abnormality detection method is characterized by comprising the following steps:
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
2. The transmission line abnormality detection method according to claim 1, characterized in that a transmission line abnormality detection model based on a YOLOv4 network is adopted, and the transmission line abnormality detection model includes a feature extraction network module, a feature enhancement module, and a detection module.
3. The transmission line anomaly detection method according to claim 2, wherein in the transmission line anomaly detection model based on the YOLOv4 network, a CSPDarknet53 module in the YOLOv4 network is replaced by an improved Ghost module, feature extraction layers of a trunk of the YOLOv4 network are respectively arranged at layers 5, 11 and 16, and a bottleneck layer of each GhostNetBottleNeck comprises two GhostModule modules which are respectively used for expanding and reducing the number of channels and ensuring that the number of channels connected with an input can be well provided.
4. A transmission line anomaly detection method according to claim 3, wherein the modified Ghost module comprises:
introducing a GhostNet network into the foreign object feature extraction module, generating a real feature layer from input features by utilizing convolution operation of the GhostNet network, obtaining the Ghost feature layer by deep convolution DW processing on the generated real feature layer, and splicing the obtained real feature layer and the Ghost feature layer to obtain an output feature layer;
the input transmission line characteristic diagram is m multiplied by n multiplied by cin, the output characteristic diagram is m multiplied by n multiplied by cout, the whole input layer is divided into s parts, the convolution kernel size is set as k, and the calculated amount through the conventional convolution operation is as follows:
m'×n'×cout×k×k×cin
the calculation amount of the whole Ghost module is expressed as:
after conventional convolution and depth convolution, the compression rate of the whole model is changed into s, so that the calculated amount of the whole model is greatly reduced;
in order to further improve the performance of the Ghost module, an ECA mechanism is introduced to replace an original SE mechanism on the basis of ensuring that the calculated amount is not increased; the ECA mechanism is mainly a channel attention model and is used for carrying out global average pooling on the feature map m multiplied by n multiplied by c, so that a vector with the size of 1 multiplied by c is obtained, and information interaction among channels is realized through one-dimensional convolution operation; the size k' of the convolution kernel in a one-dimensional convolution operation is determined mainly from an adaptive function:
where Θ (C) represents a modulus of the vector for C, a and δ represent coefficients of linear functions, respectively, and m () represents a function for finding the nearest odd number.
5. The transmission line abnormality detection method according to claim 4, wherein an improved SPP module is adopted in the YOLOv4 network-based transmission line abnormality detection model, and an average pooling is introduced to replace one of the maximum pooling to complete the improvement of the SPP module; and the foreign matter and the background information are distinguished by using the average pooling operation, more foreign matter information is reserved, and the accuracy of foreign matter detection is improved.
6. The method for detecting an abnormality of a power transmission line according to claim 5, characterized in that,
YOLOv4 network-based total Loss function Loss of transmission line anomaly detection model Total (S) The method comprises the following steps:
Loss total (S) ＝Loss Regression +Loss Confidence level +Loss Classification
The total Loss function includes regression error Loss Regression Confidence error Loss Confidence level Classification error Loss Classification ；
Defining a power transmission line foreign matter detection image with m×n size, dividing the image into d×d grids, p representing the number of boundary frames to be predicted in the divided grids, Q and Q' representing actual boundary frames and prediction boundary frames, respectively, and regression error Loss Regression The method comprises the following steps:
wherein I is i,j Represents the size of the jth frame region of the ith frame, lambda represents a constant,
definition P i,j Indicating the probability of not containing foreign object objects in the jth frame of the ith frame,
confidence coefficients of the actual boundary frame and the prediction boundary frame are respectively obtained, and a confidence coefficient error Loss is obtained according to a cross entropy formula Confidence level ：
Where mu represents a set constant value,
defining using cross entropy formula
And->
The probability of foreign object type in the actual boundary box and the predicted boundary box respectively, and the classification error Loss Classification The method comprises the following steps:
where c ε f represents the vector c in the f range, where f is the value found from the image.
7. The power transmission line abnormality detection system is characterized by comprising an image acquisition module, an edge end module and a cloud service module, wherein the power transmission line abnormality detection system is based on a cloud-edge fusion technology;
and an image acquisition module: the device is arranged on a power transmission line and is configured to continuously collect images and transmit the images to an edge end module;
the cloud service module is configured to: acquiring a training data set, training the foreign matter detection model of the power transmission line by using the training data set, and unloading the foreign matter detection model of the power transmission line with the trained parameters to an edge end module after the training process is completed;
an edge module configured to: inputting the image acquired by the image acquisition module into a trained transmission line foreign matter detection model for detection to obtain a transmission line foreign matter detection result;
the edge end module detects the image which is uploaded recently by the image acquisition module, and sends the image with the detection result of the abnormal object and the detection result to the cloud service module as a new sample;
the cloud service module acquires a new sample updating training data set for further model training, performs incremental learning, updates parameters of the transmission line foreign matter detection model, and sends the transmission line foreign matter detection model with updated parameters to the edge end module.
8. The power transmission line abnormality detection device is characterized by comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor being operative according to the instructions to perform the steps of the method according to any one of claims 1 to 6.
9. A storage medium having stored thereon a computer program, which when executed by a processor performs the steps of the method according to any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310113127.3A CN116385954A (en) | 2023-02-15 | 2023-02-15 | High-precision power transmission line abnormality detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310113127.3A CN116385954A (en) | 2023-02-15 | 2023-02-15 | High-precision power transmission line abnormality detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116385954A true CN116385954A (en) | 2023-07-04 |
Family
ID=86964599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310113127.3A Pending CN116385954A (en) | 2023-02-15 | 2023-02-15 | High-precision power transmission line abnormality detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116385954A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110827251A (en) * | 2019-10-30 | 2020-02-21 | 江苏方天电力技术有限公司 | Power transmission line locking pin defect detection method based on aerial image |
| CN113658136A (en) * | 2021-08-17 | 2021-11-16 | 燕山大学 | Conveyor belt defect detection method based on deep learning |
| CN114359167A (en) * | 2021-12-15 | 2022-04-15 | 湖北工业大学 | Insulator defect detection method based on lightweight YOLOv4 in complex scene |
| CN115327301A (en) * | 2022-08-30 | 2022-11-11 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line circuit foreign matter identification detection method, equipment and storage medium |
| CN115410087A (en) * | 2022-08-30 | 2022-11-29 | 南京航空航天大学 | Transmission line foreign matter detection method based on improved YOLOv4 |
-
2023
- 2023-02-15 CN CN202310113127.3A patent/CN116385954A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110827251A (en) * | 2019-10-30 | 2020-02-21 | 江苏方天电力技术有限公司 | Power transmission line locking pin defect detection method based on aerial image |
| CN113658136A (en) * | 2021-08-17 | 2021-11-16 | 燕山大学 | Conveyor belt defect detection method based on deep learning |
| CN114359167A (en) * | 2021-12-15 | 2022-04-15 | 湖北工业大学 | Insulator defect detection method based on lightweight YOLOv4 in complex scene |
| CN115327301A (en) * | 2022-08-30 | 2022-11-11 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line circuit foreign matter identification detection method, equipment and storage medium |
| CN115410087A (en) * | 2022-08-30 | 2022-11-29 | 南京航空航天大学 | Transmission line foreign matter detection method based on improved YOLOv4 |
Non-Patent Citations (3)
| Title |
|---|
| KAI HAN等: "GhostNet: More Features from Cheap Operations", 《ARXIV》, 13 May 2020 (2020-05-13), pages 3 * |
| 唐翔翔等: "基于改进YOLOv4的输电线路异物检测算法", 《安徽大学学报(自然科学版)》, vol. 45, no. 05, 30 September 2021 (2021-09-30), pages 1 * |
| 张鋆等: "基于边缘智能的输电线路异常目标高效检测方法研究", 《电网技术》, vol. 46, no. 05, 31 May 2022 (2022-05-31), pages 1 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353413B (en) | Low-missing-report-rate defect identification method for power transmission equipment | |
| CN110135319B (en) | Abnormal behavior detection method and system | |
| CN114022432B (en) | Insulator defect detection method based on improved yolov5 | |
| CN109492561B (en) | Optical remote sensing image ship detection method based on improved YOLO V2 model | |
| CN111161315B (en) | Multi-target tracking method and system based on graph neural network | |
| CN112541508A (en) | Fruit segmentation and recognition method and system and fruit picking robot | |
| CN110533022B (en) | Target detection method, system, device and storage medium | |
| CN114266891A (en) | Railway operation environment abnormity identification method based on image and laser data fusion | |
| CN115240075B (en) | Construction and training method of electric power vision multi-granularity pre-training large model | |
| CN115861853A (en) | Transmission line bird nest detection method in complex environment based on improved yolox algorithm | |
| CN114842363B (en) | Identification method and system for key power equipment in digital twin platform area | |
| CN114926667B (en) | Image identification method based on cloud edge cooperation | |
| CN115424016A (en) | High-voltage transmission line identification method | |
| CN114677357A (en) | Model, method and equipment for detecting self-explosion defect of aerial photographing insulator and storage medium | |
| CN113052136B (en) | Pedestrian detection method based on improved Faster RCNN | |
| CN114359167A (en) | Insulator defect detection method based on lightweight YOLOv4 in complex scene | |
| CN113536944A (en) | Distribution line inspection data identification and analysis method based on image identification | |
| CN114219763A (en) | Infrared picture detection method for abnormal heating point of power distribution equipment based on fast RCNN algorithm | |
| CN117474883A (en) | Insulator state detection method based on residual error offset network self-adaptive optimization | |
| Liu et al. | Object detection algorithm based on lightweight YOLOv4 for UAV | |
| CN112132207A (en) | Target detection neural network construction method based on multi-branch feature mapping | |
| CN111666822A (en) | Low-altitude unmanned aerial vehicle target detection method and system based on deep learning | |
| CN116385954A (en) | High-precision power transmission line abnormality detection method and system | |
| CN116052024A (en) | Power line inspection method based on light-weight target recognition neural network model | |
| CN116739963A (en) | Power grid equipment defect detection method based on multi-level multi-scale 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 |