The undersea robots driving offshore wind generation

As wind farms grow, a new fleet of AI-enabled robots is helping keep the lights on.
Written by Greg Nichols, Contributing Writer

Wind farms are now a reality in the U.S., heralding a new chapter in the country's sustainable energy production ambitions. But new technologies come with new challenges, and for offshore wind generation, inspection is one of the biggest.

In much the same way as energy companies operate and maintain oil and gas subsea assets, wind farm cables, structural foundations, and all other components of the turbines need continuous monitoring and maintenance. That's dangerous work for humans, but it's a job tailor made for underwater robots and smart AI-powered analytics.

Given the bright future and growing (albeit still small) footprint of offshore wind in the nation's energy power generation infrastructure, I reached out to Harry Turner, a machine learning specialist for Vaarst, a business driving the future of marine robotics, to discuss how robots and machine learning are changing the game for energy creation.

GN: Can you explain some of the challenges of undersea inspection, particularly for offshore wind turbines?

Harry Turner: To build and maintain wind farm assets, you need a clear understanding of the subsea environment and the condition of your infrastructure. These assets include everything from the structures that turbines sit on, to the cabling that carries electricity back to the mainland. At these depths regular inspections are usually carried out with remotely-operated underwater vehicles (ROV). But the teams that pilot those ROVs and interpret the data they collect, work on large vessels which they live on for anything from two weeks to three months. And these vessels require large crews to run, use huge quantities of fuel and are incredibly expensive. 

Another challenge is capturing and managing the vast quantity of unique data required. The data volumes involved in this process are huge, think 4k video streamed continuously by more than 10 cameras for one to three months - plus positioning information, multibeam sonar data, and 20-30 other data streams, that update up to hundred times per second. It can also take many hundred hours to review and analyse video images collected. Manually interpreting potential risk factors and recognising changes in the seabed has, to date, only been done by placing tens of people offshore on each vessel to do this work.

Finally, accurate underwater measurement is incredibly difficult, but also critically important. Often the original CAD data is unavailable for subsea assets plus there can be substantial marine growth or damage over time, so to be able to properly maintain and repair them, pinpoint measurement accuracy is key.

GN: What technologies are currently used in seabed inspection? What are the limits of the current technologies and how does that impact adoption of green energy solutions?

Harry Turner: Seabed surveys are carried out from vessels deploying sonars that map the seabed. For closer inspections, the majority of companies are using manually operated ROVs collecting video data. Each ROV needs at least two pilots to operate it. And then the data collected is inspected manually by an additional team. The more people you need, the bigger the ships you then require. This is not only expensive but obviously these ships have an environmental impact as well. 

The marine robotics industry is ripe for innovation and AI is undoubtedly going to change the landscape, by decarbonising marine operations with data-driven automation of marine robotics.

GN: Please explain how Vaarst uses AI to aid undersea inspection. What's new and novel about this approach?

Harry Turner: For some time, AI has been lauded as a game-changer for many industries. It has huge potential in a number of applications, but right now, every industry is grappling with how to become more sustainable. It's in this area that AI may help reap the best rewards. The future of marine robotics lies in using 3D computer vision and machine learning to help improve efficiency and ease the transition to greener, renewable energy sources, and ways of working in offshore environments.

The use of robotics in the energy industry isn't new – as far as industries go, they were relatively early adopters – but the use of more advanced technologies, such as simultaneous localisation and mapping (SLAM), machine learning, and increasingly autonomous ROVs, presents an opportunity that too few are seizing. By leveraging such technologies energy companies can reap significant benefits.

There are three key areas Vaarst's technology is making a significant impact:

Firstly, ROVs are run by pilots who perform all the control tasks. Vaarst has built a platform which retrofits various layers of autonomy to ROVs. These layers go from advanced assistance to autonomous control. Supporting the operator to do the job safely.

While an ROV would normally run on a predefined path that the operator would follow, the autonomy technology allows it to take the SLAM information and analyse "on the go", presenting alternative options to the operator to complete its strategy whilst navigating obstacles, or course correcting for currents. The operator can then make informed, one touch decisions.

By enabling autonomy, fewer pilots are needed, and they can be located on shore, in a supervisory role thereby eliminating the need for bigger vessels offshore.

Secondly, Vaarst is innovating Computer Vision, that is to say, the way a computer sees. Vision is about giving understanding and context to images. To do this Vaarst has developed technology that captures 3D point clouds to create accurate images and accompanying measurements in real time. This allows the ROV to "orient itself" in its environment. 

Finally, Vaarst's Machine Learning (ML) Platform processes video feeds in discrete frames. The platform can recognise key features and anomalies, automatically tag them, and grade them according to confidence levels – enabling human operators to check the work and confirm the findings, which vastly expedites the process. This again, can be completed onshore thus removing people from hazardous environments and reducing vessel sizes for a positive environmental impact.

For example, in the past pipeline surveys (that is following the length of a pipeline to check its condition) may have taken hundreds of hours and meant taking additional crew members on survey vessels to carry out this time consuming, manual work. Vaarst's technology makes it possible to reduce not only the time needed to carry out this task, but the need to take these crew members on the vessels at all, enabling the work to be done from onshore.

GN: Who are Vaarst's customers (generally or specifically, either fine)? What's the pitch to prospective customers in terms of advantages, capability, and cost savings?

Harry Turner: We work with a number of leading energy suppliers on some of the biggest renewable projects in Europe, from the energy operators themselves through to the many companies operating within the supply chain. All see the huge benefits that can be brought through future-proofing their data sets for ongoing analysis, and of being able to store and maintain their data digitally. 

The immense cost savings seen from reduced rework, and large time savings in data collection and analysis are appealing. As are the reduced days at sea, which can afford dramatic cost savings, reduced CO2 emissions and the removal of humans from hazardous conditions.

Improved life/work balance is also key. Younger generations are choosing lifestyles that often do not match the demands of pursuing a career offshore on vessels, so enabling work to be performed onshore is a key way to attract and retain talent. Equally, the gamification of technology software holds appeal to this generation and takes advantage of their skillsets.

GN: What lessons are being learned about undersea inspection utilizing your process? What other applications or opportunities might your technology open up?

Harry Turner: The main lesson being learnt is that there is an effective and practical way to streamline what has been a cumbersome and expensive process up until now. The energy sector is ready for innovation, but it needs to permeate the entire maintenance and inspection supply chain.

As we continue to build and innovate, there is no doubt that the lessons we learn in marine robotics will drive innovation in AI into new and exciting territories. The vision and autonomy technology we have designed along with our analysis platforms can be applied to any robotics, not just undersea ROVs. It can be utilised in any environment, from the deepest sea trenches to hostile environments such as nuclear facilities, in air using drones or even in interplanetary discovery!

Editorial standards