CN111401437A - Deep learning-based power transmission channel hidden danger early warning grade analysis method - Google Patents
Deep learning-based power transmission channel hidden danger early warning grade analysis method Download PDFInfo
- Publication number
- CN111401437A CN111401437A CN202010174906.0A CN202010174906A CN111401437A CN 111401437 A CN111401437 A CN 111401437A CN 202010174906 A CN202010174906 A CN 202010174906A CN 111401437 A CN111401437 A CN 111401437A
- Authority
- CN
- China
- Prior art keywords
- early warning
- hidden danger
- power transmission
- transmission channel
- detection
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of detection of hidden troubles of a power transmission line; in particular to a power transmission channel hidden danger early warning grade analysis method based on deep learning, which comprises the following steps: s1, preprocessing the pre-collected hidden danger pictures, marking hidden danger types and early warning levels, and generating a data set; s2, training and testing the neural network model by using the data set to obtain a power transmission channel hidden danger detection and early warning level detection model; s3, deploying a power transmission channel hidden danger detection and early warning level detection model into a server, and loading model parameters by the server; s4, acquiring hidden danger pictures in the power transmission channel through the monitoring equipment and uploading the pictures to a server; and S5, the server receives the hidden danger picture, calls a power transmission channel hidden danger detection and early warning level detection model to detect the hidden danger in the picture and provides early warning level information. The method and the device can obtain the hidden danger and the early warning level of the power transmission channel, and technicians can timely handle various hidden dangers, thereby effectively ensuring the safety of the power transmission line.
Description
Technical Field
The invention relates to the technical field of detection of hidden troubles of a power transmission line; in particular to a power transmission channel hidden danger early warning grade analysis method based on deep learning.
Background
In the power industry, the safety problem of the power transmission line is always of great importance. However, with the rapid development of economy, the engineering construction is gradually increased, and the phenomena of mechanical construction in a power transmission channel due to the construction of roads, railways and buildings are more and more, so that the potential safety hazard of a power transmission line is greatly increased. Therefore, it is essential to regularly monitor the operating state of the transmission line.
In recent years, a target detection technology based on deep learning is widely applied in the industry, the rapid detection of potential safety hazards in a power transmission channel is realized, the difficulty of power transmission line inspection is greatly reduced, and the detection efficiency is improved. However, the range of the power transmission channel shot by the monitoring equipment is wide, the large and small hidden dangers can be alarmed, the unordered alarm without priority causes great trouble to the processing of the hidden dangers, and meanwhile, the working efficiency is reduced.
In summary, how to provide an efficient and reliable method and system for analyzing early warning levels of hidden dangers in a power transmission channel, classify the early warning levels of the hidden dangers, quickly know the priority of hidden danger processing, and provide technical support for timely clearing of serious hidden dangers in a power transmission line is a problem to be solved by technical staff in the field.
Disclosure of Invention
In order to solve the technical problems, the invention provides a power transmission channel hidden danger early warning grade analysis method based on deep learning, which can obtain the hidden danger and early warning grade of a power transmission channel, so that technical personnel can quickly deal with the hidden danger and process various hidden dangers in time, and the safety of a power transmission line is effectively ensured.
The invention discloses a deep learning-based power transmission channel hidden danger early warning grade analysis method, which comprises the following steps of:
s1, preprocessing the pre-collected hidden danger pictures, marking hidden danger types and early warning levels, and generating a data set;
s2, training and testing the neural network model by using the labeled data set to obtain a power transmission channel hidden danger detection and early warning grade detection model;
s3, deploying a power transmission channel hidden danger detection and early warning level detection model into a server, and loading model parameters by the server;
s4, acquiring hidden danger pictures in the power transmission channel through the monitoring equipment, and uploading the pictures to a server through a network transmission module;
and S5, the server receives the hidden danger picture uploaded by the monitoring equipment, calls a power transmission channel hidden danger detection and early warning level detection model to detect the hidden danger in the picture and provides early warning level information.
The method comprises the steps of firstly, repeatedly training and testing a pre-collected hidden danger picture through a neural network model to obtain a power transmission channel hidden danger detection and early warning level detection model, embedding the model into a server, then analyzing the hidden danger picture in the power transmission channel obtained by a monitoring device through the server to obtain hidden danger information and give an early warning level, reporting the obtained hidden danger and early warning level information to a technician in time, and determining the priority of hidden danger processing according to the level of the early warning level by the technician.
Preferably, the preprocessing in step S1 is specifically: the image is processed by the PCA method, the dimensionality of the feature vector can be reduced by the PCA method, the feature vector with large dispersion between small dispersion classes in the class, such as vector representation of features such as color, texture and the like, is extracted, and matrix multiplication is carried out on the feature vector and original data, so that the dimensionality is reduced, the calculated amount is reduced, and meanwhile, most of original data information is saved, and the training accuracy can be good on the premise of saving space and time.
Preferably, the method for marking the hidden danger types and the early warning levels in step S1 is as follows: and (4) framing hidden danger positions in the hidden danger pictures, and labeling labels corresponding to the hidden dangers, wherein the hidden danger early warning grades are classified into slight early warning, general early warning and serious early warning, the early warning grades are labeled according to the distance between the hidden dangers and the power transmission line and the size of the hidden dangers in the pictures, and the labeling mode is manual labeling by a labeling person.
Preferably, in step S4, the monitoring device includes a camera, a control unit, and a network access module, where the network access module is connected to the network transmission module, and the control unit uploads the monitoring picture captured by the camera to the server through the network transmission module. The network transmission module adopts a 4G network module, namely a wireless network operated by operators such as current mobile, communication and telecommunication, and has wide coverage and stable data transmission.
Preferably, in step S3, the method for generating the power transmission channel hidden danger detection and early warning level detection model includes:
the marked hidden danger pictures are used as a training data set, a deep learning framework is used for training a neural network model, test set data is used for testing the model effect, the iteration times, the learning rate (learning rate) or the optimization method of the model are adjusted according to the test result, for example, the iteration times are gradually increased, a larger initial learning rate is used, the iteration times are gradually reduced, the batch gradient is reduced in the optimization method, and the performance of a power transmission channel hidden danger detection and early warning level detection model is improved.
Preferably, the neural network model is based on an L light-head R-CNN object detection algorithm, a branch prediction early warning grade is added while classification and frame regression are performed, a double-layer full-connection network is used for predicting early warning grade branches, 800-class 1200 nodes are arranged in the first layer, feature extraction is performed on hidden dangers of all early warning grades, N types, namely N types of early warning grades, are arranged and output in the second layer, wherein N is a positive integer and represents the number of the classes of the early warning grades.
Preferably, the L light-head R-CNN object detection algorithm is composed of two parts, namely an ROI (candidate region) generation layer and an R-CNN sub-network, and the algorithm flow is that a feature map subjected to 16-time down-sampling in a feature extraction network is used as the input of the ROI to generate a candidate region, then the candidate region and the last layer of feature map of the feature extraction network are jointly used as the input of the R-CNN sub-network, and candidate region classification and border regression are completed through a single-layer fully-connected network.
Compared with the prior art, the invention has the following beneficial effects:
the method and the device can obtain the hidden danger and the early warning grade of the power transmission channel, so that technicians can quickly deal with the hidden danger and timely process various hidden dangers, and the safety of the power transmission line is effectively guaranteed. Firstly, repeatedly training and testing a pre-collected hidden danger picture through a neural network model to obtain a power transmission channel hidden danger detection and early warning level detection model, and placing the model into a server, then analyzing the hidden danger picture in the power transmission channel obtained by a monitoring device through the server to obtain hidden danger information and give an early warning level, timely reporting the obtained hidden danger and early warning level information to a technician, and determining the priority of hidden danger processing according to the level of the early warning level by the technician.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
Example 1:
as shown in fig. 1, the method for analyzing the warning level of the hidden danger of the power transmission channel based on deep learning in this embodiment includes the following steps:
s1, preprocessing the pre-collected hidden danger pictures, marking hidden danger types and early warning levels, and generating a data set;
s2, training and testing the neural network model by using the labeled data set to obtain a power transmission channel hidden danger detection and early warning grade detection model;
s3, deploying a power transmission channel hidden danger detection and early warning level detection model into a server, and loading model parameters by the server;
s4, acquiring hidden danger pictures in the power transmission channel through the monitoring equipment, and uploading the pictures to a server through a network transmission module;
and S5, the server receives the hidden danger picture uploaded by the monitoring equipment, calls a power transmission channel hidden danger detection and early warning level detection model to detect the hidden danger in the picture and provides early warning level information.
The preprocessing in step S1 specifically includes: the picture is processed by the PCA method, the dimensionality of the feature vectors can be reduced by the PCA method, only some feature vectors with obvious features are taken to perform matrix multiplication with the original data, so that the dimensionality is reduced, the calculated amount is reduced, meanwhile, most of original data information is saved, and on the premise of saving space and time, the training accuracy can still be good.
The method for marking the hidden danger types and the early warning levels in the step S1 comprises the following steps: and (4) framing hidden danger positions in the hidden danger pictures, and labeling labels corresponding to the hidden dangers, wherein the hidden danger early warning grades are classified into slight early warning, general early warning and serious early warning, the early warning grades are labeled according to the distance between the hidden dangers and the power transmission line and the size of the hidden dangers in the pictures, and the labeling mode is manual labeling by a labeling person.
In step S4, the monitoring device includes a camera, a control unit, and a network access module, the network access module is connected to the network transmission module, and the control unit uploads a monitoring picture captured by the camera to the server through the network transmission module. The network transmission module adopts a 4G network module, namely a wireless network operated by operators such as current mobile, communication and telecommunication, and has wide coverage and stable data transmission.
In step S3, the method for generating the power transmission channel hidden danger detection and early warning level detection model includes:
the marked hidden danger pictures are used as a training data set, a deep learning framework is used for training a neural network model, test set data is used for testing the model effect, the iteration times, the learning rate (learning rate) or the optimization method of the model are adjusted according to the test result, for example, the iteration times are gradually increased, a larger initial learning rate is used, the iteration times are gradually reduced, the batch gradient is reduced in the optimization method, and the performance of a power transmission channel hidden danger detection and early warning level detection model is improved.
The neural network model is based on L light-head R-CNN object detection algorithm, a branch prediction early warning grade is added while classification and border regression are carried out, a double-layer full-connection network is used for predicting early warning grade branches, 800-plus 1200 nodes are arranged in the first layer, feature extraction is carried out on hidden dangers of each early warning grade, N types of early warning grades are arranged and output in the second layer, wherein N is a positive integer and represents the number of classes of the early warning grades, the double-layer full-connection network of the L light-head R-CNN object detection algorithm is composed of two parts, namely an ROI generation layer and an R-CNN sub-network, the algorithm flow is that a feature map sampled by 16 times in the feature extraction network is used as the input of an ROI to generate a candidate region, then the candidate region and the last layer of the feature extraction network are used as the input of the R-CNN sub-network together, and candidate region classification and border regression are completed through the single-layer full-connection network.
The method comprises the steps of firstly, repeatedly training and testing a pre-collected hidden danger picture through a neural network model to obtain a power transmission channel hidden danger detection and early warning level detection model, embedding the model into a server, then analyzing the hidden danger picture in the power transmission channel obtained by a monitoring device through the server to obtain hidden danger information and give an early warning level, reporting the obtained hidden danger and early warning level information to a technician in time, and determining the priority of hidden danger processing according to the level of the early warning level by the technician.
In the primary detection of the hidden danger of the power transmission channel, 44000 on-site pictures are uploaded in a certain time period by adopting the method for analyzing the early warning level of the hidden danger of the power transmission channel based on the deep learning, wherein the pictures comprise 18596 pictures containing hidden danger construction machinery and 25404 pictures not containing the construction machinery, the detection and early warning level analysis are carried out on the hidden danger in all the pictures, finally, the recognition missing report rate of the hidden danger of the construction machinery is 5.2%, the recognition false report rate is 7.2% and the accuracy rate is 88.1%, and the early warning level of the hidden danger in the pictures is given at the same time: and (3) making a slight early warning prompt for the construction machinery far away from the power transmission line, and making a serious early warning prompt for the construction machinery close to the power transmission line, wherein the prediction accuracy of the early warning grade reaches 86%, and the technical requirement is met.
Claims (7)
1. A power transmission channel hidden danger early warning grade analysis method based on deep learning is characterized by comprising the following steps:
s1, preprocessing the pre-collected hidden danger pictures, marking hidden danger types and early warning levels, and generating a data set;
s2, training and testing the neural network model by using the labeled data set to obtain a power transmission channel hidden danger detection and early warning grade detection model;
s3, deploying a power transmission channel hidden danger detection and early warning level detection model into a server, and loading model parameters by the server;
s4, acquiring hidden danger pictures in the power transmission channel through the monitoring equipment, and uploading the pictures to a server through a network transmission module;
and S5, the server receives the hidden danger picture uploaded by the monitoring equipment, calls a power transmission channel hidden danger detection and early warning level detection model to detect the hidden danger in the picture and provides early warning level information.
2. The deep learning-based power transmission channel potential risk early warning level analysis method according to claim 1, wherein the preprocessing in step S1 specifically comprises: the data set is processed using the PCA method.
3. The deep learning-based power transmission channel potential hazard early warning grade analysis method as claimed in claim 1, wherein the method for marking the types and the early warning grades of the potential hazards in step S1 is as follows: and (4) framing hidden danger positions in the hidden danger pictures, and labeling labels corresponding to the hidden dangers, wherein the hidden danger early warning grades are divided into a slight early warning grade, a general early warning grade and a serious early warning grade, and the early warning grades are labeled according to the distance between the hidden dangers and the power transmission line and the size of the hidden dangers in the pictures.
4. The deep learning-based power transmission channel potential risk early warning level analysis method according to claim 1, wherein in step S4, the monitoring device includes a camera, a control unit, and a network access module, the network access module is connected to the network transmission module, and the control unit uploads a monitoring picture captured by the camera to the server through the network transmission module.
5. The deep learning-based power transmission channel hidden danger early warning level analysis method according to claim 1, wherein in step S3, the power transmission channel hidden danger detection and early warning level detection model generation method comprises:
the marked hidden danger pictures are used as a training data set, a deep learning framework is used for training a neural network model, test set data is used for testing the model effect, the iteration times and the learning rate of the model are adjusted according to the test result, or the optimization method is used for improving the performance of the power transmission channel hidden danger detection and early warning level detection model, such as increasing the iteration times and increasing the learning rate in the initial training stage.
6. The deep learning-based power transmission channel potential hazard early warning grade analysis method as claimed in claim 5, wherein the neural network model is based on L light-head R-CNN object detection algorithm, a branch prediction early warning grade is added while classification and border regression are performed, a double-layer full-connection network is used for branch prediction early warning grade, 800-1200 nodes are arranged on a first layer to perform feature extraction on potential hazards of each early warning grade, and N types, namely N types of early warning grades, are arranged and output on a second layer, wherein N is a positive integer and represents the number of classes of the early warning grade.
7. The deep learning-based power transmission channel potential hazard early warning level analysis method as claimed in claim 6, wherein the L light-head R-CNN object detection algorithm comprises two parts, namely an ROI generation layer and an R-CNN sub-network, and the algorithm process comprises the steps of generating a candidate region by taking a feature map subjected to 16-time down-sampling in the feature extraction network as the input of the ROI, then taking the candidate region and the last layer of feature map of the feature extraction network as the input of the R-CNN sub-network, and completing candidate region classification and frame regression through a single-layer full connection network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010174906.0A CN111401437A (en) | 2020-03-13 | 2020-03-13 | Deep learning-based power transmission channel hidden danger early warning grade analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010174906.0A CN111401437A (en) | 2020-03-13 | 2020-03-13 | Deep learning-based power transmission channel hidden danger early warning grade analysis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111401437A true CN111401437A (en) | 2020-07-10 |
Family
ID=71434425
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010174906.0A Withdrawn CN111401437A (en) | 2020-03-13 | 2020-03-13 | Deep learning-based power transmission channel hidden danger early warning grade analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111401437A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101181A (en) * | 2020-09-10 | 2020-12-18 | 湖北烽火平安智能消防科技有限公司 | Automatic hidden danger scene recognition method and system based on deep learning |
| CN112906454A (en) * | 2020-12-22 | 2021-06-04 | 安徽康能电气有限公司 | Power transmission channel AI visual monitoring method and device |
| CN113486808A (en) * | 2021-07-08 | 2021-10-08 | 核工业井巷建设集团有限公司 | Convolutional neural network-based distribution box hidden danger identification method |
| CN114550902A (en) * | 2022-01-13 | 2022-05-27 | 重庆医科大学附属第一医院 | Hospital power utilization equipment management system and method based on RFID |
-
2020
- 2020-03-13 CN CN202010174906.0A patent/CN111401437A/en not_active Withdrawn
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101181A (en) * | 2020-09-10 | 2020-12-18 | 湖北烽火平安智能消防科技有限公司 | Automatic hidden danger scene recognition method and system based on deep learning |
| CN112906454A (en) * | 2020-12-22 | 2021-06-04 | 安徽康能电气有限公司 | Power transmission channel AI visual monitoring method and device |
| CN113486808A (en) * | 2021-07-08 | 2021-10-08 | 核工业井巷建设集团有限公司 | Convolutional neural network-based distribution box hidden danger identification method |
| CN114550902A (en) * | 2022-01-13 | 2022-05-27 | 重庆医科大学附属第一医院 | Hospital power utilization equipment management system and method based on RFID |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111401437A (en) | Deep learning-based power transmission channel hidden danger early warning grade analysis method | |
| WO2022160413A1 (en) | Electric power production anomaly monitoring method and apparatus, and computer device and storage medium | |
| CN111754498A (en) | Conveyor belt carrier roller detection method based on YOLOv3 | |
| CN113592828A (en) | Nondestructive testing method and system based on industrial endoscope | |
| CN113344852A (en) | Target detection method and device for power scene general-purpose article and storage medium | |
| CN112541393A (en) | Transformer substation personnel detection method and device based on deep learning | |
| CN113569956B (en) | Mountain fire disaster investigation and identification method based on AI algorithm | |
| CN110458794B (en) | Quality detection method and device for accessories of rail train | |
| CN112116594A (en) | Wind floating foreign matter identification method and device based on semantic segmentation | |
| CN113888514A (en) | Method and device for detecting defects of ground wire, edge computing equipment and storage medium | |
| CN114863118A (en) | Self-learning identification system and method based on external hidden danger of power transmission line | |
| CN115019209A (en) | Method and system for detecting state of electric power tower based on deep learning | |
| CN115082813A (en) | Detection method, unmanned aerial vehicle, detection system and medium | |
| CN113095160B (en) | Power system personnel safety behavior identification method and system based on artificial intelligence and 5G | |
| CN116310922A (en) | Petrochemical plant area monitoring video risk identification method, system, electronic equipment and storage medium | |
| CN115984158A (en) | Defect analysis method and device, electronic equipment and computer readable storage medium | |
| CN117274197A (en) | PCB defect detection method based on YOLO v5 algorithm improvement | |
| CN114821486B (en) | Personnel identification method in power operation scene | |
| CN113592864B (en) | Transformer monitoring method, device, system and medium based on convolutional neural network | |
| CN115631153A (en) | Pipe gallery visual defect detection method and system based on perception learning structure | |
| CN115187880A (en) | Communication optical cable defect detection method and system based on image recognition and storage medium | |
| CN114373116A (en) | Method, device, equipment and product for screening operation and maintenance site images | |
| CN113139476A (en) | Data center-oriented human behavior attribute real-time detection method and system | |
| CN117522872B (en) | Real-time abnormality detection method and device for photovoltaic power station | |
| CN113928947B (en) | Elevator maintenance process detection method and device |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200710 |
|
| WW01 | Invention patent application withdrawn after publication |