Baidu PaddlePaddle easydl helps improve the inspection efficiency of high-altitude photovoltaic power stations by 98%

The double carbon policy was accelerated , Photovoltaic industry has entered a period of rapid development . National Energy Administration statistics ,2021 In, China newly added photovoltaic grid connected installed capacity 53GW, Year on year growth of about 10%, continuity 9 In, it ranked first in the world . But photovoltaic power stations are usually located in the Gobi 、 In the mountains 、 Grassland and other sparsely populated areas , The environment is relatively harsh , Failure of precision instruments occurs frequently . To prevent equipment failure , Ensure the stable operation of the power station , The daily operation and maintenance inspection of photovoltaic power station is particularly important .

From passive inspection to active prediction

Photovoltaic operation and maintenance “ Patrol inspection and supervision ” demand

The traditional inspection method is that the operation and maintenance personnel hold the scanner high or check the photovoltaic panel with the help of vehicles , There are many hidden faults , It took a long time to be found , There are many invalid patrols , High labor cost , There are also many problems in the later operation and maintenance inspection .

 

In order to meet the requirements of cost reduction, efficiency increase and refined management of photovoltaic power stations , Unmanned intelligent inspection has become the general trend , This also gave birth to photovoltaic operation and maintenance “ Patrol ”、“ check ”、“ prison ”、“ Observe ” Four business needs ：

Patrol ： How to achieve efficient patrol ？

Photovoltaic power stations are generally large , Photovoltaic panels are scattered , A large number of panels are distributed per unit area , for instance , a 100MW Photovoltaic power stations in the world , Photovoltaic panel modules can be as high as 40 Ten thousand pieces , The traditional inspection basically adopts “ people + vehicle ” By , It often takes a lot of time and energy , Both personnel cost and vehicle cost are very high .

check ： How to realize refined fault judgment ？

In traditional inspection , The method of human eye discrimination extraction is often used to extract target information from image data , It is easy to make mistakes , And it's less efficient . meanwhile , The output results will also be affected by the subjective consciousness of personnel , Accuracy and standardization need to be improved .

prison ： How to realize dynamic monitoring of equipment ？

With the refined management of power stations, the demand is higher and higher , The demand for dynamic monitoring of equipment is also increasing , However, traditional patrol inspection has a large equipment base , The routine inspection cycle is long , At the same time, consider the strong wind 、 Sand and dust are not suitable for inspection , It is impossible to monitor the state of the target object anytime and anywhere .

Observe ： How to mine the deep value of data ？

Photovoltaic power stations cover a large amount of data , Including patrol data 、 Equipment power generation data 、 Terrain data, etc , How to mine the deep value of data from patrol big data , Provide customers with value-added services , For example, analyze the deterioration trend of equipment performance from the UAV thermal infrared remote sensing big data 、 Fault prediction, etc , Become the demand of the digital era of energy .

Baidu Flying propeller EasyDL

Help PV power station patrol inspection reduce cost and improve efficiency

Today, the photovoltaic industry has entered the era of meager profits , Every minute of power generation is extremely critical . How to improve the efficiency of quality control and operation and maintenance management , It is an urgent problem to be solved in the photovoltaic industry .

Beijing Digital Vision Technology Co., Ltd. is a supplier of unmanned aerial vehicle intelligent inspection system for photovoltaic power stations , Focus on the research, development and application of UAV intelligent inspection system in photovoltaic power stations . To meet the photovoltaic operation and maintenance patrol 、 check 、 prison 、 Observe needs , Digital creation is based on Baidu Flying propeller EasyDL Zero threshold AI Development platform , Develop a set of efficient inspection 、 Intelligent operation and maintenance 、 Safe and reliable intelligent inspection solution for photovoltaic power station .

The scheme is based on deep learning of thermal infrared photovoltaic panel recognition , It can quickly and intelligently analyze the thermal infrared images of photovoltaic modules obtained by UAV , Automatically identify hot spots 、 diode 、 Zero current and other faults and automatic positioning , Among them, hot spots 、 The accuracy of diode automatic recognition exceeds 95%, Fault location accuracy is better than 1m, Compared with the traditional manual inspection, the efficiency is improved by dozens of times .

With 100MW Take the DC side inspection of the power station as an example , In the past, manual inspection required 180 people / God , And UAV patrol only needs 3 people / God , Improve the efficiency of patrol inspection 98.33%. Plus fault diagnosis 、 Troubleshooting, etc , A conservative estimate , UAV patrol + Fault diagnosis can improve the efficiency of power station operation and maintenance management in 20% above .

Behind the improvement of efficiency , Cannot do without AI Blessing of technology . Speaking of AI Integration of technology and photovoltaic industry , Yang Fan, the founder of digital startups, is also wandering around on the road of exploration . Digital startups has built its own 2-3 Human doctoral team develops algorithms , after 3 After months of R & D investment , The accuracy of the discovery algorithm can only reach 90%, Unable to meet commercial needs , Then I investigated and tried many AI platform products , Finally chose Baidu Flying propeller EasyDL Zero threshold AI Development platform .“ The accuracy of the algorithm can not only be stable at 95% above , It also greatly reduces the development cost , Improved R & D efficiency , Commercialization requirements are also easy to achieve .” Yang Fan said .

Mention the benefits of the project , Yang Fan with a 20MWp Take photovoltaic power station as an example , Fault components are found through intelligent patrol inspection of UAV 287 block , Distributed in 182 Within a string . After quantitative analysis of fault diagnosis , The annual power loss is about 1,065,598kwh, Accounting for 3.15%. After defect elimination and rectification , The power of the power station increases 5% about , The cost of inspection and defect elimination in the whole process is only 5.07 Ten thousand yuan , The income is as high as 106 Ten thousand yuan .

This is the best time , It is also the most challenging era , Powerful empowerment of AI technology , It is reshaping all walks of life in an unprecedented way . As “AI Technology flow ”, Baidu Flying propeller EasyDL After years of industrial practice , It has been widely used in industry 、 Agriculture 、 In transportation and other industries . future , Baidu Flying propeller EasyDL It will enable more enterprises to transform , Help speed up the building of a new paradigm of made in China , Co climbing AI The peak of the era of industrial production .