Informed AI in the Water Sector - The Roadmap from Possible to Plausible

Silvia Vlad, Jacobs

As water suppliers throughout New Zealand tackle the twin challenges of climate change and limited investment budgets, artificial intelligence (AI) and machine learning applications can help us achieve both sustainability and efficiency goals, giving utilities the insights needed to streamline and optimise operations and capital investments. While the conceptual promise of AI is often lauded, many utilities find that real-world considerations such as historical data quality and quantity can interpose significant limitations on the value of AI applications.

This paper will discuss key factors in identifying viable machine learning applications, highlighting data requirements and methods for supplementing real-world historical data, as well as outlining opportunities for tiered AI implementations that grow with an organisation as it gains digital maturity. Starting from available data and planning out future data collection, machine learningapplications can increase in sophistication and value as additional datasets become available.

By further augmenting our existing datasets with synthetic data, “informed” AI allows machine learning models to overcome the challenges of narrow historical datasets, which frequently capture only a small range of operating conditions. Leveraging both scenario analysis and operational forecasting capabilities, AI applications are giving utilities an ever-expanding set of insights into chemical and energy optimisation, greenhouse gas, and cost-reduction opportunities. Key case studies from around the world will showcase how global best practices in leveraging machine learning can be used to help New Zealand utilities achieve sustainability and cost-savings goals.