Australian utilities explore potential of AI

While Australia often enjoys a reputation as an early adopter of new technologies, the utilities sector’s uptake of artificial intelligence (AI) appears to be progressing at a more measured pace.

Recent research conducted for Itron, a technology provider for energy and water utilities, revealed that only 16% of Australian utilities have fully integrated AI into their operations, compared with a global average of 22%.

That said, a significant 51% have made substantial investments and are progressing with several mature AI projects, slightly trailing the worldwide figure of 60%.

This data, sourced from Itron’s 2024 resourcefulness report, also revealed that 8% of Australian utilities are still in the exploratory phase of AI adoption. This contrasts with countries such as India, where the corresponding figure is a mere 2%, and the US, where all surveyed organisations have moved beyond initial exploration.

But just because Australian utilities aren’t leading the way, it would not be fair to label them as laggards. Itron is an established provider to these organisations, with more than 75 customers around the country but especially in Victoria and Western Australia, according to Alex Beveridge, Itron’s Australia, New Zealand, Pacific Islands and strategic markets area vice-president.

Noting that the definition of AI itself has broadened in recent years, Beveridge said the current focus on generative AI (GenAI) often overshadows the longer-established subsets of AI such as machine learning (ML), which have been integral to Itron’s portfolio and the utilities sector for some time. “It’s not a new concept,” he said.

A key driver for AI adoption is safety. One early application of ML by Victoria’s United Energy, powered by Itron’s SensorIQ technology, involved analysing data from advanced metering infrastructure (AMI) meters to identify potential neutral integrity issues. This enables ongoing surveillance instead of relying on mandatory 10-yearly inspections of connection points.

Another long-established customer is Australian Energy Market Operator (AEMO), which has been using Itron MetrixIDR to merge data from meters and other sources including weather reports to generate short-term forecasts of the load on the entire network. With a high penetration of rooftop solar, access to solar irradiance data is an important part of the process.

Thanks to solar panels, Australia has millions of distributed generation assets so it is not practical for humans to control the overall system – but MetrixIDR’s combination of neural networks and mathematical models can do the job. Being able to process power quality data from electricity meters has helped customers, said Beveridge.

The next step is to move the intelligence from the back office to the edge of the electricity grid, he said.

One part of the problem is to balance supply and demand behind each transformer. On the supply side, electricity providers know about the solar panels in an area, but they lack visibility of new sources of demand such as electric vehicles (EVs), which can have a substantial operational impact – especially in circumstances such as several people in a neighbourhood taking delivery of electric cars on the same day.

One of the factors driving the investigation of AI in the utilities sector is that while humans were able to control electricity networks with thousands or tens of thousands of assets, in another decade or two, Australia will have millions of distributed generation assets, Beveridge predicted, and managing that will require the use of ML and AI. Moving intelligence to the edge allows utilities to react in close to real time, whereas centralised processing involves hours of latency.

The sooner changes are detected, the sooner providers can take steps to mitigate the effects, such as encouraging controlled events such as cycling air-conditioning compressors or adjusting EV charging profiles so that neighbours don’t all charge their cars at the same time.

Itron has even developed a product specifically designed to assist electric bus operators in optimising their fleet’s charging patterns to remain within the capacity limitations of their depot’s transformer.

The need to balance supply and demand to safeguard critical infrastructure remains a concern for the entire electricity grid. New winter peak demand records were set in Queensland and Victoria in July 2024, followed by record low demand periods in subsequent months. Furthermore, September 9 saw a peak renewable contribution record of 72.2%, with distributed (mostly rooftop) solar accounting for 38.5% of total electricity generation, according to AEMO.

Around half of Australian homes are expected to have rooftop solar by 2040, so the problem isn’t getting simpler. Indeed, AEMO recently said that it needed last-resort emergency powers to turn rooftop solar installations down or off for short periods to maintain the necessary balance between generation and consumption. This power already exists in Victoria, South Australia, Western Australia and parts of Queensland, but AEMO believes it is needed in the rest of the country.

In Victoria, where rooftop solar systems installed or upgraded since October 2024 are required to have a utility compatible inverter and an internet connection, energy distributor Jemena uses Itron’s LV DERMS software to collect and process almost real-time solar generation data from such systems, so it can moderate their output as part of the process of keeping the energy grid in balance.

“While solar generation is beneficial to customers and a sustainable way to power homes, excess solar power can cause supply-demand imbalance within the energy system and potentially overload it,” said Georgia Cronin, Jemena’s general manager of customer energy solutions. “Itron’s LV DERMS helps us manage minimum demand and contributes to keeping the lights on for our customers.”

Utilities must also balance safety and criticality against the need to move reasonably quickly to meet challenges such as decarbonisation targets. “That’s a tricky balance. So, to move with new technology, we always discuss with our customers the need to think big, [but] start small,” Beveridge said.

While acknowledging that trials require time, Beveridge noted that there are substantial projects in Australia that are demonstrating the potential of AI and ML to integrate renewable energy sources effectively, paving the way for large-scale deployments.