CN109784672A - A kind of warning system for real time monitoring and method for power grid exception - Google Patents
A kind of warning system for real time monitoring and method for power grid exception Download PDFInfo
- Publication number
- CN109784672A CN109784672A CN201811587468.XA CN201811587468A CN109784672A CN 109784672 A CN109784672 A CN 109784672A CN 201811587468 A CN201811587468 A CN 201811587468A CN 109784672 A CN109784672 A CN 109784672A
- Authority
- CN
- China
- Prior art keywords
- sample
- region
- time monitoring
- power grid
- convolutional neural
- 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
Classifications
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- 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
- Image Analysis (AREA)
Abstract
The present invention provides a kind of warning system for real time monitoring for power grid exception for recommending convolutional network based on region, comprising: fixed photographic device, region recommend convolutional neural networks, sample to generate system, alarm system.Various pieces cooperate, and are automatically performed the real-time monitoring work to power grid designated position.Region recommends convolutional neural networks to be made of two networks, and recommendation network prediction in region is likely to occur abnormal region, and convolutional neural networks are accurately positioned the type and its size and location of exception object.It is alarmed immediately by modes such as mail, short messages after it found the abnormal situation, notifies inspector.A kind of real-time monitoring alarming method of above-mentioned warning system for real time monitoring for power grid exception is provided simultaneously.The present invention realizes the automation real-time monitoring to the specified abnormal conditions of power grid designated position, saves a large amount of manpowers.
Description
Technical field
It is specifically a kind of the present invention relates to a kind of system for carrying out real-time monitoring alarming to power grid designated position
Recommend the warning system for real time monitoring and method for power grid exception of convolutional neural networks based on region.
Background technique
Electric inspection process is run concerning the safety and steady of power grid.Guo Wang company puts into a large amount of manpower and material resources every year to guarantee electric power
Inspection work.Wherein, some transmission lines of electricity and substation in city, due to close to civil plantation area, rubbish, kite etc.
The main reason for foreign matter often causes power transmission failure.The harm of problems should not be underestimated, but due to it with contingency and
Paroxysmal feature, it is virtually impossible to predict its time occurred, therefore Utilities Electric Co. has to send someone to do timed patrol.
However, often even several weeks a few days just will appear once the failure as caused by foreign body intrusion, artificial timed patrol
Not only wasting manpower and material resources, and floor manager is easy to psychologically become to relax after not finding the problem for a period of time.
In addition, must have certain interval between inspection each time, failure can not probably be found in time when occurring.Therefore, by counting
Calculation machine cooperates sensor to replace manual inspection, realizes that real time monitoring is inexorable trend.
Artificial intelligence technology after rising and falling for several times, finally in outburst comprehensively in recent years.It is widely answered in deep learning
With before, feature extraction and the matching of characteristic quantity generally require to be designed according to practical problem in image recognition, and threshold is higher,
Versatility is poor.Designer is needed to have the sensibility of higher mathematics standard and data.Recommend convolutional neural networks in region
Using simplifying feature extraction and the matched work of characteristic quantity.In addition, the rare of sample probably I guess at present introduces deep learning
The biggest obstacle of electric utility.Sufficiently large suitable sample with mark is collected into for specific scene, in the short time
It is the task of almost impossible completion.Therefore, how to utilize computer to synthesize sample using appropriate mode and carry out sample generation
And expand, to realize the real-time monitoring alarming for being directed to power grid exception, become this field urgent problem to be solved.
Currently without the explanation or report for finding technology similar to the present invention, it is also not yet collected into money similar both at home and abroad
Material.
Summary of the invention
Aiming at the above shortcomings existing in the prior art, the purpose of the present invention is to propose to one kind recommends convolution net based on region
The power grid exception warning system for real time monitoring and method of network, the system and method synthesize sample using computer using appropriate mode
The generation of this progress sample and expansion, application region are recommended convolutional neural networks to carry out the identification of abnormal image, are realized for electricity
Net abnormal real-time monitoring alarming.
The present invention is achieved by the following technical solutions:
According to an aspect of the invention, there is provided a kind of warning system for real time monitoring for power grid exception, including take the photograph
As device, region recommend convolutional neural networks, sample to generate system and alarm system；Wherein:
The photographic device, collection site image, and reach region and recommend convolutional neural networks；
The sample generates system, training sample is generated according to the live image of photographic device acquisition, via gaussian filtering
Convolutional neural networks are recommended in training region after processing；
Convolutional neural networks are recommended in the region, carry out Anomaly target detection to image, and by abnormal image at
Function identification triggering alarm system.
Preferably, live image is transmitted back to region recommendation convolutional neural networks by 4G module by the photographic device.
Preferably, it includes region recommendation network and convolutional neural networks that convolutional neural networks are recommended in the region；Wherein:
The region recommendation network, the convolutional layer for being 3 × 3 comprising 13 convolution kernel sizes, 4 pond ranges are 2 × 2
Pond layer, 13 activation primitive layers, the activation primitive layer use ReLU activation primitive, are likely to occur abnormal area for predicting
Domain；
The convolutional neural networks use VGG16 network, the accurate positionin for abnormal area.
Preferably, the sample generate system by exception object individual in photo site with it is normal when image pass through pool
The mode of pine fusion merges automatically, generates training sample.
Preferably, the sample generates system and zooms in and out rotation transformation to individual exception object, wherein scaling multiple
It is 0.5~2 times, rotation angle is ± 5~± 30 °.
Preferably, the training sample includes the sample of synthesis and the location tags of target.
Preferably, the alarm system is using following any one or any a variety of type of alarms:
Mail alarm；
SMS alarm；
Wechat alarm.
According to another aspect of the present invention, a kind of real-time monitoring alarming method for power grid exception is provided, including
Following steps:
S1, sample generate system and generate training sample by material of photographic device live image collected,；
S2 mixes acquired original sample and generates training sample, and convolutional Neural is recommended in training region after gaussian filtering
Network；
S3 recommends convolutional neural networks to carry out abnormal inspection to the live image that photographic device acquires using trained region
It surveys:
If abnormal, pass through alarm system Realtime Alerts；
Otherwise, then S3 is returned, the abnormality detection of live image is continued.
Preferably, the sample generates system and generates training sample, includes the following steps:
Use the corresponding Pixel Labeling in Binary Images target position of the profile of abnormal object；
Input abnormal object sample and background sample；
Rotation transformation is zoomed in and out according to setting rule to abnormal object；
Transformed abnormal object and the normal picture as background sample are subjected to graph cut；
Export training sample and the label with the training sample type and position.
Preferably, further include following any one or any multinomial feature:
The rule of the scaling rotation transformation of setting are as follows: scaling multiple is 0.5~2 times, and rotation angle is ± 5~± 30 °；
The training sample includes the sample of synthesis and the position mark file of target.
Compared with prior art, the invention has the following beneficial effects:
1, the warning system for real time monitoring and method provided by the invention for power grid exception, simplifies the work of patrol officer
It measures, and improves the real-time of abnormal conditions alarm.
2, the warning system for real time monitoring and method provided by the invention for power grid exception is returned from the picture pick-up device at scene
It passes, to gaussian filtering process is passed through, is to be automatically performed to early warning when identification and exception, omits human involvement factor, mention
The high accuracy of monitoring result.
Detailed description of the invention
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon:
Fig. 1 is that sample generates systematic sample product process figure；
Fig. 2 is the warning system for real time monitoring work flow diagram provided by the present invention for power grid exception.
Specific embodiment
Elaborate below to the embodiment of the present invention: the present embodiment carries out under the premise of the technical scheme of the present invention
Implement, the detailed implementation method and specific operation process are given.It should be pointed out that those skilled in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.
Embodiment
A kind of warning system for real time monitoring for power grid exception is present embodiments provided, which is based on region and recommends volume
Product neural network, comprising: fixed photographic device, region recommend convolutional neural networks, sample to generate system and alarm system
System.Wherein, live image is transmitted back to server end by 4G module by photographic device, is recommended by the region that server end is disposed
Convolutional neural networks carry out Anomaly target detection to image, are directly patrolled by alarm system prompt if detecting exception in image
Inspection person.
Further, it includes two nets of region recommendation network and convolutional neural networks that convolutional neural networks are recommended in the region
Network, the former includes 13 convolutional layers, and convolution kernel size is 3 × 3,4 pond layers, and pond range is 2 × 2,13 activation primitives
Layer, activation primitive use ReLU activation primitive, are likely to occur abnormal region for predicting, the latter uses VGG16 network, is used for
Realize accurate abnormal area positioning.Identification model used need to train in advance, and training sample used comes from collection in worksite and sample
Generation system generates, then via gaussian filtering process.
Further, the sample generates system and passes through individual exception object and image of surveillance area when normal
The mode of graph cut merges automatically, wherein needs to zoom in and out exception object rotation transformation, rotation angle regards different objects
And it is different, value range is ± 5 ° to ± 30 °, and scaling multiple regards different positions, and value range is 0.5 to 2 times.
Further, the alarm system using any one in mail alarm, SMS alarm and wechat alarm or is appointed
Meaning various ways are triggered by the successful identification of abnormal image.
The warning system for real time monitoring of power grid exception, real-time alarming method by monitoring, packet are directed to provided by the present embodiment
Include following steps:
S1, sample generate system and generate training sample by material of photographic device live image collected；
S2 mixes acquired original sample and generates sample, for training region to recommend convolutional Neural after gaussian filtering
Network；
S3 recommends convolutional neural networks to carry out abnormal inspection to the live image that photographic device acquires using trained region
It surveys:
If abnormal, pass through alarm system Realtime Alerts；
Otherwise, then S3 is returned, the abnormality detection of live image is continued.
In the present embodiment:
Sample generates system and generates sample using picture pick-up device collection site image, for training network after filtering.It uses
When by photographic device acquisition live image be sent into region recommend convolutional neural networks performance objective detect work, to find image
In abnormal conditions, and carry out Realtime Alerts by modes such as mails in time after noting abnormalities.
Exception object is merged with live image using graph cut, generates sample, while generating exception object
The mark file of position and type.
It is returned from the picture pick-up device at scene, to filtering, alarm when to identification and occurring abnormal is to be automatically performed.
Alarm signal will not be issued when normal, pass through the modes such as mail/short message/App/ wechat when abnormal conditions occur
Auto-alarming is carried out, during which without manual intervention.
Sample for training region to recommend convolutional neural networks all is from collection in worksite, and automatic using graph cut
It generates.
The above embodiment of the present invention is further described with reference to the accompanying drawing.
As shown in Figure 1, sample, which generates system, uses graph cut, exception object marks it using the bianry image of its profile
Position in original image, the object in the position will be merged with the live image as background sample.
As shown in Fig. 2, being directed to the warning system for real time monitoring of power grid exception, work provided by the above embodiment of the present invention
Make method, first generate system using sample and generate certain amount sample, gaussian filtering is carried out to sample, is then fed into region recommendation
Convolutional neural networks train network weight, frequency of training 70000 times, initial learning rate be 0.001,50000 times after be 0.0001.
The convolution kernel size of convolutional layer is 3 × 3 in the recommendation network of region, and the number of plies is 13 layers, and 4 layers of pond layer, pond range is 2 × 2, letter
Several layers 13 layers, activation primitive uses ReLU activation primitive, and convolutional neural networks use VGG16 network.Two networks pass through approximation
Joint training training simultaneously.
When system is run, picture pick-up device passes live image back server end, server end deployment by built-in 4G module
It has trained the region completed to recommend convolutional neural networks, has automatically grabbed the image that picture pick-up device is passed back and execute identification operation, occur
When abnormal, alarm system is triggered, sends alarm mail to the mailbox of the inspector set in advance.This is that system is completely instructed
Practice, the course of work.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited to above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring substantive content of the invention.
Claims (10)
1. a kind of warning system for real time monitoring for power grid exception, which is characterized in that recommend convolution including photographic device, region
Neural network, sample generate system and alarm system；Wherein:
The photographic device, collection site image, and reach region and recommend convolutional neural networks；
The sample generates system, training sample is generated according to the live image of photographic device acquisition, via gaussian filtering process
Region is trained to recommend convolutional neural networks afterwards；
Convolutional neural networks are recommended in the region, carry out Anomaly target detection to image, and pass through the successful knowledge to abnormal image
It Chu Fa not alarm system.
2. the warning system for real time monitoring according to claim 1 for power grid exception, which is characterized in that the camera shooting dress
It sets and live image is transmitted back to by recommendation convolutional neural networks in region by 4G module.
3. the warning system for real time monitoring according to claim 1 for power grid exception, which is characterized in that the region pushes away
Recommending convolutional neural networks includes region recommendation network and convolutional neural networks；Wherein:
The region recommendation network, the convolutional layer for being 3 × 3 comprising 13 convolution kernel sizes, the pond that 4 pond ranges are 2 × 2
Layer, 13 activation primitive layers, the activation primitive layer use ReLU activation primitive, are likely to occur abnormal region for predicting；
The convolutional neural networks use VGG16 network, the accurate positionin for abnormal area.
4. the warning system for real time monitoring according to claim 1 for power grid exception, which is characterized in that the sample is raw
At system exception object individual in photo site is merged by way of graph cut automatically with image when normal, is generated
Training sample.
5. the warning system for real time monitoring according to claim 4 for power grid exception, which is characterized in that the sample is raw
Rotation transformation is zoomed in and out to individual exception object at system, wherein scaling multiple is 0.5~2 times, and rotation angle is ± 5
~± 30 °.
6. the warning system for real time monitoring according to claim 4 for power grid exception, which is characterized in that the trained sample
This includes the sample of synthesis and the location tags of target.
7. the warning system for real time monitoring according to claim 1 for power grid exception, which is characterized in that the alarm system
System is using following any one or any a variety of type of alarms:
Mail alarm；
SMS alarm；
Wechat alarm.
8. a kind of real-time monitoring alarming method for power grid exception, which comprises the steps of:
S1, sample generate system and generate training sample by material of photographic device live image collected,；
S2 mixes acquired original sample and generates training sample, and convolutional neural networks are recommended in training region after gaussian filtering；
S3 is carried out abnormality detection using the live image that trained region recommends convolutional neural networks to acquire photographic device:
If abnormal, pass through alarm system Realtime Alerts；
Otherwise, then S3 is returned, the abnormality detection of live image is continued.
9. the real-time monitoring alarming method according to claim 8 for power grid exception, which is characterized in that the sample is raw
Training sample is generated at system, is included the following steps:
Use the corresponding Pixel Labeling in Binary Images target position of the profile of abnormal object；
Input abnormal object sample and background sample；
Rotation transformation is zoomed in and out according to setting rule to abnormal object；
Transformed abnormal object and the normal picture as background sample are subjected to graph cut；
Export training sample and the label with the training sample type and position.
10. the real-time monitoring alarming method according to claim 9 for power grid exception, which is characterized in that further include as
Lower any one or any multinomial feature:
The rule of the scaling rotation transformation of setting are as follows: scaling multiple is 0.5~2 times, and rotation angle is ± 5~± 30 °；
The training sample includes the sample of synthesis and the position mark file of target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811587468.XA CN109784672A (en) | 2018-12-25 | 2018-12-25 | A kind of warning system for real time monitoring and method for power grid exception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811587468.XA CN109784672A (en) | 2018-12-25 | 2018-12-25 | A kind of warning system for real time monitoring and method for power grid exception |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109784672A true CN109784672A (en) | 2019-05-21 |
Family
ID=66498127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811587468.XA Pending CN109784672A (en) | 2018-12-25 | 2018-12-25 | A kind of warning system for real time monitoring and method for power grid exception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109784672A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334612A (en) * | 2019-06-19 | 2019-10-15 | 上海交通大学 | Electric inspection process image object detection method with self-learning capability |
| CN110807353A (en) * | 2019-09-03 | 2020-02-18 | 国网辽宁省电力有限公司电力科学研究院 | Transformer substation foreign matter identification method, device and system based on deep learning |
| CN112348306A (en) * | 2020-09-09 | 2021-02-09 | 北京华商三优新能源科技有限公司 | TitanOS artificial intelligence development method and device for power distribution operation inspection |
| CN115190277A (en) * | 2022-09-08 | 2022-10-14 | 中达安股份有限公司 | Safety monitoring method, device and equipment for construction area and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778472A (en) * | 2016-11-17 | 2017-05-31 | 成都通甲优博科技有限责任公司 | The common invader object detection and recognition method in transmission of electricity corridor based on deep learning |
| CN108734143A (en) * | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
-
2018
- 2018-12-25 CN CN201811587468.XA patent/CN109784672A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778472A (en) * | 2016-11-17 | 2017-05-31 | 成都通甲优博科技有限责任公司 | The common invader object detection and recognition method in transmission of electricity corridor based on deep learning |
| CN108734143A (en) * | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
Non-Patent Citations (1)
| Title |
|---|
| 狄亚柳: ""基于深度学习的高铁接触网悬挂目标检测图像处理研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334612A (en) * | 2019-06-19 | 2019-10-15 | 上海交通大学 | Electric inspection process image object detection method with self-learning capability |
| CN110807353A (en) * | 2019-09-03 | 2020-02-18 | 国网辽宁省电力有限公司电力科学研究院 | Transformer substation foreign matter identification method, device and system based on deep learning |
| WO2021042682A1 (en) * | 2019-09-03 | 2021-03-11 | 国网辽宁省电力有限公司电力科学研究院 | Method, apparatus and system for recognizing transformer substation foreign mattter, and electronic device and storage medium |
| CN110807353B (en) * | 2019-09-03 | 2023-12-19 | 国网辽宁省电力有限公司电力科学研究院 | Substation foreign matter identification method, device and system based on deep learning |
| CN112348306A (en) * | 2020-09-09 | 2021-02-09 | 北京华商三优新能源科技有限公司 | TitanOS artificial intelligence development method and device for power distribution operation inspection |
| CN115190277A (en) * | 2022-09-08 | 2022-10-14 | 中达安股份有限公司 | Safety monitoring method, device and equipment for construction area and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784672A (en) | A kind of warning system for real time monitoring and method for power grid exception | |
| CN110850723B (en) | Fault diagnosis and positioning method based on transformer substation inspection robot system | |
| CN110648364B (en) | Multi-dimensional space solid waste visual detection positioning and identification method and system | |
| CN102013147B (en) | High voltage power transmission tower intelligent anti-theft method for supervising and device | |
| CN103096121B (en) | A kind of camera movement detection method and device | |
| CN102915638A (en) | Surveillance video-based intelligent parking lot management system | |
| CN103514694A (en) | Intrusion detection monitoring system | |
| CN102446390A (en) | Method and system for safety detection and early warning of monitoring areas near power transmission lines | |
| CN110012268A (en) | Pipe network AI intelligent control method, system, readable storage medium storing program for executing and equipment | |
| CN109684976A (en) | Door state monitoring method, device, equipment and system and storage medium | |
| CN111259855A (en) | Mobile safety helmet wearing detection method based on deep learning | |
| CN105929392A (en) | Radar and video multi-system interaction offshore platform system | |
| CN114648708A (en) | State detection method and device of photovoltaic module, storage medium and electronic equipment | |
| CN110390288B (en) | Target intelligent searching, positioning and tracking system and method based on computer vision | |
| Manninen et al. | Multi-stage deep learning networks for automated assessment of electricity transmission infrastructure using fly-by images | |
| CN106600905A (en) | Effective geological disaster monitoring system | |
| CN112818806A (en) | Transformer substation inspection robot auxiliary navigation method based on deep learning | |
| CN115190277B (en) | Safety monitoring method, device and equipment for construction area and storage medium | |
| CN116385959A (en) | Belt corridor global smoke real-time detection method based on deep learning | |
| TW201113835A (en) | Intelligent image detecting method, intelligent image detecting system and intelligent image detecting software | |
| CN112712604A (en) | Machine vision-based on-site inspection quality non-inductive detection method | |
| CN204463161U (en) | A kind of image detecting processing system | |
| CN109271828A (en) | The method and system of construction segregator barriers condition intelligent detection based on deep learning | |
| CN206546417U (en) | A kind of GIS switch fault automatic recognition systems based on picture recognition | |
| Lin et al. | A Garbage Recognition Algorithm Based on YoloX Using in Flood Control System |
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 |