With its heritage of regulated assets and limited competition, infrastructure has been slow to embrace technology when compared with similarly capital-intensive but competitive industries such as manufacturing, where tech innovation has been critical to survival. But that is changing.

The dual pressures of climate change mitigation and adaptation, and the need to create operational efficiencies, mean digitisation is now in the ascendancy. And, from the artificial intelligence underpinning the rise of autonomous vehicles to predictive maintenance of renewable generation assets, transformational technologies are at last being adopted in myriad ways across the infrastructure investment lifecycle.

“The incorporation of AI and machine learning into infrastructure investing has been steadily on the rise over the past few years and we are really starting to see its impact across the asset class,” says Misha Logvinov, managing director and head of IT strategy at EQT. “We now have access to more data than ever before and we are working hard to maximise its potential to create differentiated market leaders with new, AI-driven revenue streams.”

“The spread of machine learning and data analytics tools enable new value-creation opportunities for management teams,” adds Marion Calcine, chief investment officer at Ardian Infrastructure. “Managers that can harvest the data generated by infrastructure assets can monitor the assets’ operations more closely and find additional pockets of value creation and optimisation, thus gaining a competitive advantage over conventional investors.”

AI and origination

The use cases for data analytics, machine learning and AI begin with deal sourcing. “We use AI-based data analytics software to identify, research and monitor new investment themes and potential investments,” says Julien Bedin, co-head of infrastructure investment research at Partners Group. “The tool combs through our internal databases of over 50,000 assets and third-party resources to help identify new opportunities. AI-based software has proved effective in expanding our horizons across industries and asset classes.”

These technologies also have a role to play in construction. “Smart contracting has been made possible by the adoption of evidence-based, real-time data reporting,” says Tibor Schwartz, senior adviser at QIC. He cites drone-based inspections, which are increasingly being used to validate progress on construction projects.

“The spread of machine learning and data analytics tools enables new value creation opportunities for management teams”

Marion Calcine
Ardian Infrastructure

“Technologies like blockchain – with features such as dynamic contract updating, based on digital data obtained through multi-sensory and digital capture systems – is another example,” Schwartz adds. “These technologies help infrastructure managers to improve capital management by allocating risk more precisely across contractual boundaries, minimising contractual disputes, and promoting accountability and efficiency.”

Equally, if not more importantly, is the use of AI and digital twinning – the creation of digital representations of physical assets – to improve multiple aspects of infrastructure asset management. “Digital twinning facilitates data-driven decision making for complex asset management and operations,” says Schwartz. “AI and digital twinning together can be used to understand causal relationships across multiple, heterogeneous data sets and generate deep asset management insights that would be impossible without these critical technologies.”

Elsewhere, AI tools are also being used to support forecasting in major construction projects, thereby improving capital efficiency. Octant AI’s solution for the Queensland Department of Transport and Main Roads delivered improved early warnings by an average of four months for underruns and two months for overruns, according to Schwartz. It also improved the accuracy of portfolio cost forecasts by 87 percent; of project cost forecasts by 21 percent; of project duration forecasts by 11 percent; and is estimated to have saved an average of $148,000 per project.

Efficiency gains

In addition to improving efficiency in the construction process, operational efficiency is a major driver of AI adoption. AMP Capital partner Adam Ringer points to the use of AI in data centres: “Data centres need to keep the power on to keep the facilities cool. The use of AI sensors is massively increasing the efficiency of what are already very high margin businesses.”

EQT’s Fenix Marine Services, which provides container handling services to shipping lines in North America and plays a critical role in harbours, last year launched HONE, a first-of-its-kind AI platform that uses machine learning to increase terminal operating efficiency. “Not only has HONE optimised operations, it has also enabled new revenue-generating priority services,” says Logvinov.

Fenix Marine Services is now able to offer predictive forecasting of container availability, which has improved operating performance for customers and increased productivity across the company’s fleet of cranes.

AI also has an important role to play in optimising the efficiency of renewable generation assets by enabling investors to collect and analyse massive amounts of data from across a portfolio. “For example, we can run statistical analysis of turbine misalignment versus wind direction,” Calcine says. “This allows GPs to really leverage on the scale that they have by having access to a larger portfolio than just one asset.”

“These technologies help infrastructure managers to improve capital management by allocating risk more precisely”

Tibor Schwartz

Portfolio construction is an important area in itself. “Managers can use simulations during the acquisition process of renewable assets to assess the impact of the additional asset in terms of risk/return at portfolio level,” says Calcine. She cites Ardian’s partnership with Pexapark, which helps the firm to gain PPA pricing intelligence to quantify and manage price risk and secure cashflow across its renewable assets.

Maintenance is another critical use case for AI in infrastructure.

Multi-sensory inspection systems allow needs-based maintenance to replace fixed or periodic maintenance cycles, thereby reducing costs and downtime. “This technology has been adopted deeply in the inspection of some assets, such as powerlines,” says Schwartz.
He cites technology company Sharper Shape, which has pioneered this capability for transmission and distribution inspections. It has developed a system using multi-sensory aerial inspection data to create an AI-based maintenance programme with automated fault

“Utilities globally are increasingly adopting this type of inspection,” Schwartz adds. “In situations like the California wildfires, where vegetation management is a key challenge for utilities, AI can predict the rate of growth and encroachment of vegetation impacting assets.”

Another example is the development of advanced structural monitoring systems for bridges, runways, roads and other transport infrastructure. Schwartz continues: “As these systems gradually replace periodic inspections, the result will be significantly safer assets with longer lives.”

Partners Group is trialing third-party software on two of its Australian wind farms. “The software analyses myriad unique data points from the wind farms’ systems to identify trends and outliers in performance, as well as early-warning signals of potential issues,” says Bedin. “The more detailed information allows maintenance crews to plan for and address issues before they lead to unplanned outages and unnecessary loss of generation. Successful implementation of these tools could potentially achieve increases in generation of up to 5 percent for underperforming farms.”

Yet there are challenges. “We remain at an early stage of adoption,” Bedin adds. “The infrastructure industry is complex and changes within the industry take time. In addition, comprehensive data analytics programmes can be difficult to implement and are not necessarily accretive in all circumstances.”

He explains that, as of 2019, only an estimated 20 percent of wind farms globally were using data analytics for operations and maintenance. However, the benefits of data analytics could extend beyond just asset maintenance optimisation in the future. For instance, AI-based software is increasingly being used for forecasting intermittent renewable generation to help with electric grid management or for power trading and hedging purposes.

Self-adjusting assets

One particularly interesting evolution in the use of AI in infrastructure involves self-adjusting assets. This includes the ability to fine-tune operational characteristics, such as the positioning of satellites in response to weather conditions. This use of AI is still in the earliest stages but could ultimately be used in the renewables industry as well.

“We can already monitor the alignment of each wind turbine and also the behaviour of wind turbines during high wind events,” says Calcine. “We can also gain more insight into turbine manufacturers because we can really monitor the availability of turbines and the actual performance vis à vis the power curves we have contractually agreed.”

For merchant renewable assets, the objective is also to capture the best market price window or to best predict production forecasts in order to minimise penalties linked to intermittency. Ardian is therefore also developing AI models to achieve short-term price projections and network congestion. “Our medium-to-long-term vision includes the objective of having our assets reacting and optimising according to various data signals, notably those coming from the markets,” Calcine says.

Schwartz adds that sophisticated transmission and distribution network operators are already combining weather forecasting with their own sensory data positioned at assets such as power exchanges: “By using multiple data sets, it is possible to predict the areas of maximum weather impact on these assets and position maintenance crews in advance to minimise the severity and duration of outages due to abnormal weather events.”

Renewable power generators are embracing cloud sensor technology to more accurately predict solar farm outputs and enable better bidding into electricity markets. They are also using infrared drone camera scanning to scan for solar panel underperformance. “This type of intelligent asset management will become even more important in the future as the need to manage changing climatic conditions grows,” says Schwartz.

Climate change considerations are among the most powerful drivers of AI adoption, and this is an area where digital and data capabilities have already proven their efficiency. “A great example is Ardian’s Air Carbon project to monitor carbon emissions at Italian airports in real time,” says Calcine.

There is no doubt that the use of AI will continue to proliferate across the infrastructure industry.

“Comprehensive data analytics programmes can be difficult to implement and are not necessarily accretive in all circumstances”

Julien Bedin
Partners Group

“We see a huge opportunity in the application of AI in the field of infrastructure – particularly for predictive means,” says Logvinov. “Whether these technologies are used to monitor assets, detect patterns in behaviours or mitigate failures, if applied at the right time, they can help morph a non-technical company into a tech-enabled infrastructure business.”

“This is the beginning of the adoption curve and there are a still a lot of unexplored avenues of developments,” says Calcine. “Current projects are built from zero by early adopters of these technologies. Most of the work is focused on data access and data infrastructure while AI applications are only scratching the surface.”

Ringer goes further, suggesting that these AI technology businesses could ultimately become infrastructure investments in themselves. “Is this AI simply going to be used to make portfolio companies better or are there business models where infrastructure businesses can invest in these technology companies in their own right? It is an interesting question,” he says. “I think AI will primarily be used to improve the efficiency of infrastructure assets.

“But if these SaaS business models achieve network effects, providing essential services to their underlying clients, they will have created that important barrier to entry. I look forward to a point in the future where these companies could become credible infrastructure investments in themselves.”