AI-Enhanced Visual Intelligence Moves Utility Asset Inspection Forward

The grid is the most complex system in the world to manage and maintain. As the grid requires increased investments, optimizing resource usage for asset maintenance is a critical aspect of smarter and more efficient grid operations. Yet grid assets are widely spread over the utilities' service territory, which can be challenging for visual inspection and vegetation management.

This article shares some of the information shared in GE Vernova’s discussion of Mitigating Threats and Strengthening the Grid with AI-based Inspection Programs during a recent webinar for Distributech. First, we’ll examine the challenges, then consider best practices and the business value of integrating visual inspection with a data management platform.

Managing Visual Inspection Challenges

Legacy survey methods (land-based line patrol) are very expensive. Driving out to get eyes on every asset is also time -consuming and can pose unacceptable liability risk. At the same time, the images collected using this approach have a lot of variation and integrity issues due to inconsistent perspectives, focal inaccuracies, and distance from the target objects.

Asset managers want to optimize their overall condition-based asset risk management programs. New technology for surveying and AI-enabled data management solutions are available. They offer electric power utilities a clear upgrade path from their legacy inspection and risk management programs.

Using helicopters, satellites, drones, or other aerial methods is faster, cheaper, and safer. Plus, a wide range of data can be gathered including images, videos, thermal, hyperspectral and multispectral information.

The size of visual inspection data can be measured in 10s or 100s of petabytes per year. This requires a data management platform to efficiently centralize and process the raw data, associate images (or point clouds) with specific assets and extract the risk classifications. That’s why the efficiency and optimization of AI-enabled data management is so appealing for vegetation management.

Industrial AI-Enablement

Traditional approaches for Asset Management that are not necessarily efficient or consistently reliable can lead to costly O&M investments. It’s also not useful to have thousands of individual jpeg files without any associated information about that asset. A human clicking through each picture one by one faces an arduous task looking to identify any red flags.

Meanwhile, an Artificial Intelligence (AI) solution can streamline the process. GE Vernova’s AI-enabled data management platform can:

With all of these capabilities on the GE Vernova platform, the utility can easily generate insights that will be turned into action. By integrating the visual inspection data with the system of record, overlaying the data with the GIS, the utility gains essential context.

Utilities that adopt these new solutions see improvements in reliability, safety, costs and service. At GE Vernova we’ve seen use cases in:

The AI-enabled platform can take the visual inspection data and identify what it is looking at and even recognize defects. The computer vision can automatically detect the asset and its component parts. Additionally, the utility might run a change detection program to see what it is missing or if it needs repair. This can benefit maintenance planning and optimize the scheduling of field crews (who go out into the field knowing what to expect and what tools and equipment will be needed).

The technology also optimizes network planning. For instance, a utility doing land survey can use the visual inspection data and change detection to identify someone building a house or other type of right of way impediment.

The emergency response application is also growing more prevalent. With aerial inspection integrated with the network, the utility can quickly send out drones to collect data on a storm or wildfire. The inspection data is then efficiently ingested, inspected, and contextualized to enable more efficient (and safer) outage response.

Visual data inspection management can also help minimize the costs associated with vegetation encroachment. With automated point cloud classification and configurable risk assessment, the utility can optimize its trimming plans.

Finally, visual data inspection can be leveraged to build (or reconcile) systems of records by sending status updates to GIS.

Injecting Value in Visual Inspection Management

Intelligent analytics facilitates visual inspection data management from end-to-end. Extracting the data to improve the quality of the system of record improves inspection planning and management. Whether it is field or aerial inspection data, Enterprise Integration Services can ingest, quickly review, and assess risk with machine learning, always improving the process along the way.

Finally, the platform allows utilities to generate reports and export the AI-optimized data to drive action. This triggers downstream actions such as creating a work order to address vegetation or repair an asset in need of maintenance.

Digitalization of the distribution grid is leading utilities to look to AI and big data development. Smart sensors have been adding observability and automation. Integrating the visual inspection data management platform further supports data mining, image recognition, and predictive modeling. Adopting AI has helped our clients to scale up successfully, shorten time-to-market, and continuously improve with greater structuring and sharing of the utility data.