Why Utility Companies Are Using Synthetic Data to Optimize Real-Time Asset Monitoring

Extreme weather events and the trend towards remote management makes synthetic data perfect for powering the revolution in utility inspections using geospatial images, artificial intelligence and machine learning algorithms.

With the real-time monitoring of assets now a critical component in the management of utility infrastructure, the use of synthetic images to train machine learning algorithms offers huge advantages. 

Geospatial imagery, collected from satellite, aerial and drone data can provide the input to allow remote monitoring. This can be done via the manual interpretation of the images. However, the analysis can also be done automatically using machine learning algorithms. These algorithms can be used, for example, for the detection of the wear and tear of infrastructure, overgrown vegetation near power lines, and other such scenarios that require action to prevent potential damage and disruption to services. 

However, in pursuit of superior turnaround times, improved accuracy and better results, synthetic data is helping companies supercharge their geospatial imagery analysis processes, in the utilities sector

Remote Management

The inspection of assets, such as power lines, pipelines, solar installations, roads, structures, etc, has become even more essential given the rise in severe weather events. The difficulty of sending ground crews during and in the aftermath of such events means assets are at greater risk of being compromised, damaged and even destroyed. 

AI-powered machine learning offers utility companies the supply of accurate, realtime data that can identify the location of infrastructure vulnerable to possible damage, and anticipate areas which may be affected by adverse weather conditions. Support crews can then be dispatched in advance with a full diagnosis of the problem they need to resolve or the protection they need to put in place on the ground before a severe weather event takes place. Importantly, this allows for predictive maintenance, which can avert costly disasters to vital infrastructure before they happen.

However, as with all machine learning algorithms, the accuracy of these models relies heavily on the amount, variety and correctly annotated data used to train them. Put simply, the more geospatial imagery and associated data used to train an algorithm, the more accurate the results will be. 

But gathering images and annotating them is a burdensome and time-consuming task prone to human error. Remember, it is not only sufficient to capture an image of an object of interest, such as a pylon or powerline, once. Each object potentially needs a minimum of thousands or more images to train the algorithm properly. This enables it to recognize the object regardless of angle, weather condition, geographic location, encroaching overgrowth of vegetation, etc. Each of these images also needs to be annotated, which as the number of images increases will become an even more laborious, time-consuming and error-prone task.

Speed, Accuracy and Efficiency

The use of synthetic imagery, which has evolved from the gaming and film industry, adds three important elements to the collection of images and data to train machine learning algorithms — speed, accuracy and efficiency. Indeed, synthetic data is the perfect solution to resolve the data bottleneck associated with the poor performance of these algorithms.

OneView’s synthetic images can be quickly developed to accommodate any use-case scenario, replacing the need for realworld imagery. These images also come with the associated annotations making them immediately ready for the training of machine learning algorithms. This not only saves vast amounts of time and resources, but also improves accuracy. If an algorithm is found to be underperforming in certain scenarios its performance can also be boosted through additional training using synthetic data. 

Optimizing data analytics through the use of synthetic data can benefit utility companies by:

1. Cutting costs associated with image gathering and annotation

2. Improving detection accuracy 

3. Providing meticulous real-time insights

4. Gaining instant feedback from the field

5. Planning maintenance and quality control

6. Enabling the ability to react immediately to issues as they arise

Challenges

In the aftermath of the fires that recently ravaged California, the failure of the electricity grid in Texas in the winter of 2021, and the requirement to meet governmental sustainability goals, the utilities sector is currently experiencing unprecedented challenges. 

These challenges are set to continue given that the Fourth National Climate Assessment report predicts energy systems across the U.S. will be increasingly affected by extreme weather including wildfires, hurricanes, and changing rainfall patterns due to climate change. And while improvements to infrastructure are being made, the scope and scale of these changes need to increase in order to safeguard reliable energy and utility services into the future. 

Technology is widely recognized as key in helping to facilitate this transition to future-proofing assets. However, as with any investment, choosing the correct solution is paramount.

The use of synthetic data can provide the highly accurate detection of infrastructure changes, damage and construction development, including pipelines, solar installations and various structures, allowing asset monitoring across multi-site, large-scale operations. When compared to real world images, the speed and flexibility of synthetic data makes it a cutting edge solution that can power the assets management needs of utility companies well into the future.