Yarra Valley Water betting on AI to predict asset failures

Yarra Valley Water is planning to use generative AI to predict failures across its water supply infrastructure in a bid to drive cost out of its maintenance operations.

The water authority is currently working on a proof-of-concept for the new system, which could be in operation as soon as next year, according to its cloud and devops lead Murali Manohar Shunmugaraja.

Shunmugaraja, who spoke to iTnews at Elastic’s Elasticon Tour in Sydney, said the new AI-based system was expected to use a large language model (LLM) inference engine to interrogate data coming from sensors embedded in the supply network to monitor millions of assets.

“It gives me predictive analysis to say, ‘Hey, you don't need to inspect all your millions of assets, but you can just look these 5000 assets – which are the sensors – and just focus on [them] because they are going to get brought into this,” Shunmugaraja said.

Yarra Valley Water, which serves around 2 million premises, has maintenance contracts with partners such as Ventia and Downer Group to provide inspection services for its water and sewer networks.

Arguably, the new predictive AI-system might allow it to call on their services in a more economical manner.

Shunmugaraja said it was still too early to provide a complete picture of the technology suppliers likely to be involved in getting the new system operational.  

The project brings with it a dilemma over which LLMs to use and how to host them.

As a regulated entity, Shunmugaraja said Yarra Valley Water had little appetite for feeding its data into a public LLM.

“We are very, very particular about what we send to an LLM. We would like to prefer to have an on-premises LLM so that LLM data stays nice and secure,” Shunmugaraja said.

However, that could carry considerable expense, he said.

“We may have to host the GPU within our data centre. That's not cheap. So, that's a difficult question that we're trying to answer,” he said.

The answer might lay in private cloud hosting which would see the data placed behind firewalls and stored strictly in adherence with the sector’s compliance and governance standards, according to Shunmugaraja.

Or, he said, Victoria’s water sector could follow the example of other sectors, like energy, where providers have entered partnerships with global cloud companies like Amazon Web Services, Microsoft (for Azure) and SAP.

Yarra Valley Water’s project could sit neatly within the Victoria’s Intelligent Water Networks (IWN) program overseen by, VicWater, the peak industry association representing the state’s 18 publicly funded water management corporations.