GAN Feeder Mapping for Hidden and Overlapping Grid Assets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for mapping power grid assets, particularly feeders, face challenges such as imperfect recognition systems, lack of detailed images, and the difficulty in differentiating multiple feeders on the same pole, as well as the inability to accurately map underground assets.
Innovation Solution
A generative adversarial network (GAN) is trained using ground truth data to predict the locations and layouts of feeders, including all assets from substations to loads, using aerial, drone, or ground-level images, enabling the generation of complete maps even in areas without available ground truth data, and can identify underground assets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional recognition systems are used to map power grid assets from aerial or street level images, then mapping can be performed with existing technology, but assets will be missing from the mapping and incorrectly identified due to imperfect recognition and lack of detailed images
Solution Approach 1:
The patent introduces an intermediary process that uses multiple image sources (aerial, street-level, satellite) and combines them with recognition systems to create a more reliable mapping. The intermediary step involves cross-referencing multiple data sources to verify asset locations and reduce false positives, thereby improving both accuracy and completeness simultaneously
Solution Approach 2:
The patent merges multiple image sources and recognition systems into a unified mapping approach. By combining aerial images, street-level images, and satellite imagery, the system overcomes the limitations of any single source, achieving both high accuracy and complete asset identification
2Area of stationary object
If multiple feeders are run on the same pole to optimize grid layout, then space efficiency is improved, but differentiation between such feeders becomes difficult with existing mapping methods
Solution Approach 1:
The patent resolves the feeder differentiation problem by adding another dimension of observation. Instead of relying solely on horizontal separation that is difficult to capture in 2D images, the system uses vertical positioning data and multi-angle imaging to distinguish between multiple feeders on the same pole, maintaining space efficiency while achieving accurate differentiation
Solution Approach 2:
The patent applies local quality analysis by examining specific regions of images with enhanced detail. For areas where multiple feeders converge on the same pole, the system applies localized processing with higher resolution analysis to differentiate individual feeders, while maintaining standard processing for areas with single feeders
3Ease of manufacture
If above-ground mapping is performed using standard imaging methods, then visible assets can be identified, but underground assets appear isolated from substations and cannot be accurately mapped
Solution Approach 1:
The patent creates a universal mapping system that handles both above-ground and underground assets through a single integrated approach. The system uses multiple imaging modalities and data sources that can detect both visible above-ground assets and infer underground asset locations, eliminating the need for separate mapping processes and achieving accurate location identification for all asset types
Solution Approach 2:
The patent applies preliminary action by using above-ground asset mapping results to guide and constrain underground asset location inference. Before attempting to map underground assets, the system first establishes the above-ground context, which provides boundary conditions and spatial relationships that improve the accuracy of subsequent underground asset location predictions
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using a neural network to predict locations of feeders in an electrical power grid. One of the methods includes training a generative adversarial network comprising a generator and a discriminator; and generating, by the generator, from input images, output images with feeder metadata that represents predicted locations of feeder assets, including receiving by the generator a first input image and generating by the generator a corresponding first output image with first feeder data that identifies one or more feeder assets and their respective locations, wherein the one or more feeder assets had not been identified in any input to the generator.