Company Introduction
Located in Chongqing — China's "mountain city" — the State Grid's Chongqing Yongchuan Electric Power Company operates and manages 45 substations, 1611 km of power transmission lines, and 5126 km of power distribution lines.
To improve O&M efficiency and overcome staff shortages, the Company actively explores novel O&M methods, such as using UAVs for intelligent power grid inspection. Currently, UAVs are widely used to inspect transmission lines, substations, and distribution lines as well as to remove foreign objects from operating power lines and to build a 3D digital grid.
What are the challenges when developing developing AI models for smart power grid inspection?
-
Labeling massive datasets is both labor-intensive and time-consuming
Labeling massive datasets is both labor-intensive and time-consuming
-
Several auto recognition models are needed for the many types of power grid defects; iterative development is challenging.
Several auto recognition models are needed for the many types of power grid defects; iterative development is challenging.
Industrialized AI boosts AI adoption in the electric power industry
With the Pangu model's auto data augmentation and adaptive loss function optimization, one model with extraordinary generalization ability can recognize hundreds of defect types, replacing over 20 smaller models.
The Huawei Cloud Pangu CV model is pre-trained using massive unlabeled electric power data. With only a few labeled samples plus minimum fine tuning, a custom, pretrained model can be quickly developed to accurately identify defects on the power grid.
Customer Benefits
-
more efficient sample selection
-
higher sample screening quality
-
higher average precision
-
less costly model development
"We are working with Huawei Cloud to explore and research intelligent ways to detect defects on the power grid. Through our joint efforts, we have developed a large model, which not only has extraordinary generalization ability but is also much more accurate than smaller models."
"We are working with Huawei Cloud to explore and research intelligent ways to detect defects on the power grid. Through our joint efforts, we have developed a large model, which not only has extraordinary generalization ability but is also much more accurate than smaller models."