Abstract: This group of “young” people with technology “tamed” drone inspection transmission lines.
In June 2019, a city was suddenly turned into a “gray city” because of the failure of its power interconnection system. Traffic lights stopped working, subways, intercity railways, buses and other public transportation stopped, and the water supply system failed to operate properly. The blackout affected more than 48 million people.
Today, when we are highly dependent on the power industry, power failure has become a disaster for the city, with vehicles unable to operate, signal lights not on and shops closed… As the lifeblood of the city’s normal operation and rapid development, it has become an urgent problem to improve the efficiency of transmission line inspection and ensure the reliable power supply of civil and enterprise electricity.
With the popularity of UAVs, circuit inspection is gradually getting rid of manpower inspection, but there are still problems such as single technical means and not being able to synchronize circuit in real time.
A group of teachers and students from North China Electric Power University saw the blue ocean market of AI inspection for power transmission lines in the future and invested in research and development. Through uav intelligent inspection and AI image detection and analysis, the accuracy of transmission line defect detection has been continuously improved.
Human pain point is difficult to solve, change machine to try
Traditional power transmission line inspection, mostly rely on operation maintenance personnel to the naked eye or handheld devices in the circuit breakdown, according to the experience to potential hidden danger, however, pure human detection itself is a huge hidden trouble, inspection is difficult to a don’t rule out all the hidden danger, and in power grid operation, any tiny safety problems are likely to cause accidents, lead to the paralysis of large area, Cause huge economic losses; Sometimes encounter rain, snow and other bad weather, manpower inspection is more difficult to move.
Most transmission lines are located in sparsely populated suburbs, far from the city center, and special inspection and maintenance lines account for 20% of all lines. The difficulty of transportation and inspection and high quality requirements make fewer and fewer people willing to engage in inspection work. According to data, the annual growth rate of current transmission line inspection personnel in China is less than 3%, and the shortage of personnel at the end of 2019 is as high as 34%. The problems of small number of inspection professionals, low labor efficiency and high risk coefficient are becoming increasingly prominent.
According to incomplete statistics, from 2014 to 2018, there were 238 cases of personal injury and death caused by overhead line inspection in China, and the economic loss caused by transmission line failure reached billions of yuan. In view of the increasing contradiction between the number of transmission professionals and the scale of equipment, the state vigorously promotes uav inspection. Compared with manual inspection, UAV has advantages of high efficiency, high safety and high defect detection rate, but there are still many problems.
Unmanned aerial vehicle (uav) inspection most homework without supporting machine nest, cannot move, defects inspection to only part of the circuit, and the inspection after the completion of the unmanned aerial vehicle (uav) is still need to manually image defect analysis, according to statistics, in the unmanned aerial vehicle (uav) power transmission line inspection in China each year more than billions of copies inspection pictures, to the consumption of human resources is still large. Taking Shandong Power Grid as an example, it is equipped with 600 unmanned aerial vehicles (UAVs), which produce about 1.2 million pictures annually, and 45,000 sets of visual monitoring devices, which produce 180 million pictures annually. According to the eight-hour daily working system, it takes 92 people per year to judge human images.
Such a heavy workload provides a huge space for the improvement of circuit inspection mechanism.
After a comprehensive market survey, the team learned that uav inspection currently in use still has problems such as high rate of missed inspection and error inspection, high labor cost and difficulty in ensuring inspection quality. In 2019, the “lightning bird – power visual” entrepreneurial teams with north China electric power university (baoding) intelligent visual computer institute, big power visual data laboratory has developed a set of complete and practical joint unmanned aerial vehicle (uav) overhead transmission line edge cloud synergy intelligent inspection tools, named it “lightning bird – unmanned aerial vehicle (uav) AI inspection platform”.
“Taming” drones is difficult because of technological innovation
When designing the scheme, the team had a clear idea that uav should be used as the eyes of inspectors, who could know the situation of each transmission line only through the Lightning Bird platform. The idea is full, the reality is very skinny, in order to realize the real “unmanned” inspection, there are still many technologies to overcome.
First of all, in the inspection of UAV, there will be limitations in battery, weather, photos not as good as the naked eye and other aspects. If these problems can not be solved, then the later technology can not really identify the defects and hidden dangers.
For this group of young people studying electricity, drones are not their area of expertise, and many of the restrictions on flying are imposed by the industry ceiling. But they are unwilling, because if these problems are not solved, drone inspection will become useless, outdoor inspection, once there is no electricity and crash will cause losses in vain.
For these problems, team members have had many internal discussions, but it is still difficult to come up with a good solution, because some limitations are the ceiling of the entire industry, and they cannot break them alone. To this end, they specifically contacted a number of drone companies, consulting the corresponding solutions.
After a bit, they would be found to extend the time of unmanned aerial vehicle (uav) flight – will charge unmanned aircraft hangar in power transmission tower, each charge hangar is equipped with multiple unmanned aerial vehicle (uav), when the drone fast without electricity charge can go to the nearest garage, charged electric unmanned aerial vehicle (uav) to take the “inspection”, rushed out of the “bird’s nest” continue to fly.
Inspection personnel only need to directly announce inspection locations and routes through the application. According to the built-in smart weather sensing station, when the weather is judged to be suitable for flight, “birds” will fly out of the nest to perform tasks. During the flight of UAV, inspection personnel can get real-time information about equipment or routes through mobile phone feedback. When the transmission line defects are found, they can be quickly located and solved in time.
To train the drone to be an eye for hidden dangers, the team used high-definition footage from visible light cameras to take a close look at towers and lines, simulating human eyes. It is because of the team’s optimization and adjustment in all aspects that uav, which is rushing to the front line of inspection, can solve the shortage behind the manpower problem with science and technology and collect enough information to pave the way for the next analysis.
Smart use of Huawei cloud to solve big problems of the project
In the past years of accumulation, the team developed the core technology — AI visual processing and analysis technology, which laid a good foundation for the progress of the project.
Even so, the team also understood that in order to truly achieve “unmanned” inspection, they still faced great challenges. If they could not solve the intelligent image recognition function, they still needed to spend a lot of manpower on image analysis and processing. How can it be solved? This requires the GPU server with powerful computing power in the cloud and powerful edge computing capacity to realize intelligent image recognition and find the problems of transmission lines in an early time.
Since 2019, North China Electric Power University has established a cooperative relationship with Huawei to introduce Huawei’s technology and artificial intelligence platform into course teaching to realize talent cultivation. On the other hand, for scientific research projects, Huawei provides computing power algorithm support and jointly invests in the research and development of the project to facilitate the completion of the project. As one of the important RESEARCH and development projects of North China Electric Power University, Huawei invested in the project support early, escorting the team. Faced with the problems during the project, the team completed the whole project with the help of the super storage capacity and computing power provided by Huawei cloud technology.
Drones out inspection, processing and analyzing technology based on full stack AI huawei vision intelligent recognition of image, image and video produced by the camera, substation internal fixation through the cloud platform deep learning combined with cloud computing, automatic analysis of image and video data, can be part of image compression, storage, One side for all kinds of fault detection, foreign body analysis; Based on the results of the server analysis, the staff can immediately troubleshoot, without waiting by the host.
Combined with Huawei full-stack AI technology, on the one hand, huawei Cloud ModelArts platform is used to develop intelligent inspection cloud applications for power transmission lines to achieve unified management of inspection data and model algorithms as well as efficient utilization of storage resources and computing resources. The key parts of transmission lines and large scale defects such as bird’s nest and insulator drop are detected. On the other hand, huawei HiLens end-cloud collaborative management platform is used to develop end-side application of intelligent inspection of power transmission lines to realize end-side device model optimization and application deployment, so as to meet the real-time requirements of critical defect detection of power transmission lines, and to detect bolts missing pins and other problems in power transmission lines.
In the case of transmission line emergencies, once the UAV detects critical defects, the pictures will not be returned to the cloud-side platform as usual, but directly to the UAV terminal for detection, and the identification results will be sent to the staff as soon as possible, greatly saving the emergency repair time for crisis defects.
Deep learning algorithm was adopted to improve the inspection accuracy of key parts for transmission lines and large scale defect testing, the testing data set is formed after image annotation, use huawei cloud ModelArts platform of “intelligent data marked with” * * * * function further expansion of data set, with intelligent data labeling method improved the efficiency of annotation, Accelerated the project process.
With the technical support of Huawei Cloud, these young people have developed products with great competitive advantages in the market.
Lightning Bird uav is not affected by terrain and climate, so it can quickly achieve a wider coverage of inspection, improving the inspection efficiency of pipeline; To make inspection intelligent and automatic, the situation of transmission lines and substations can be learned without unnecessary manpower; Through ** “intelligent inspection + deep learning + edge computing/cloud computing” ** greatly improve the detection accuracy of all kinds of defects, greatly reduce all kinds of risks and economic losses.
Open the era of intelligent inspection, the future is predictable
If there is one word to describe the development process, the team chose “youth”. They say it was waves of young people who poured into the lab to work on a project that produced the lightning bird.
Lightning Bird has significantly improved detection rates and addressed industry pain points compared to humans, but the team continues to work in the lab to improve the algorithm so that the inspection platform can identify more complex defect crises.
For the future, the team wants to expand the application field of Lightning Bird not limited to electricity, but to photovoltaic new energy, oil pipeline, transportation and other inspection scenarios. If possible, it hopes to be applied in all aspects of life, and use intelligence to make up for the defects and deficiencies of artificial.
Perhaps, in the near future, our life is like the general science fiction, we can go out and see all kinds of drones in the sky, some of them are directing the road traffic, some are monitoring the public security situation, some are testing the national health level; When you cross the road, you will see the traffic department, police, doctors are out.
I believe that when more and more high-end talents come out of the intelligent base, sci-fi blockbusters will be closer and closer to us. All walks of life can solve pain points through science and technology and reduce unnecessary losses.