Stanwell's AI platform powers shift to renewable energy

Stanwell has rolled out a new internal modelling system that uses artificial intelligence and cloud computing to support forecasting, trading, safety and asset management, as the government-owned generator expands beyond coal into wind, solar and battery storage.

The Stanwell Modelling Platform (SMP) runs on Microsoft Azure and uses Microsoft AI services. Stanwell described it as a shared environment for data and predictive models across the business.

Early results include more than 200% performance improvement for a large-scale battery asset following AI-led optimisation of charging and discharging decisions. Stanwell also reported a 30% lift in the accuracy of electricity demand forecasts for certain time frames after moving forecasting models to Azure. Some market and risk simulations now run up to 15 times faster.

From trading pilot

SMP began in Stanwell's trading function as a machine-learning experiment and is now used across forecasting, safety, trading and asset management.

Stanwell supplies more than a third of Queensland's energy needs through its generation assets, according to the company. Its portfolio is shifting as it adds renewable generation and firming assets alongside existing coal-fired plants.

That shift is reshaping how electricity is produced and traded in Australia's National Electricity Market, where prices are set in short intervals, and participants respond to weather-driven changes in supply and demand.

Stanwell Executive General Manager Business Services, Sophie Naughton, said the platform is designed for these conditions, including the growing volumes of data associated with renewable generation patterns.

"This significant increase in weather dependent energy generation, including Australia's love of solar generation, is driving significant market volatility and new trading patterns," Naughton said.

"For our trading team it's also driving the beginning of a 'data tsunami'."

"Stanwell has a proud history of selling energy from two extremely reliable coal fired energy plants, but our portfolio is growing with generation and firming assets, power purchase agreements, each one bringing with it new and more complex data for our traders to analyse."

"This data is crucial for our success. It informs our energy pricing, contracting and how much energy our assets should generate and when."

"Our traders have to move faster than ever before so as you can imagine, the right software is also now critical."

Battery operations

One of the most visible uses of SMP is a large-scale battery system, where an AI model advises when to charge or discharge. Stanwell said this approach improved asset performance by more than 200%.

Battery behaviour affects both revenue and broader system balance, particularly in states with high solar output during daylight hours and strong demand peaks later in the day. Stanwell said it is exploring ways to extend AI optimisation across its wider portfolio of generation and storage assets.

Kevin Lin, Stanwell's Chief Information Officer, said the platform has increased the speed at which teams can run complex analysis.

"Think of it as our intelligent command centre," Lin said. "SMP lets us throw complex problems at AI models and get answers in hours or seconds, where it used to take weeks. We use it across the organisation for mission critical tasks like forecasting energy demand, supporting trading decisions, and managing safety risks. It's changed how we operate."

Forecasting and risk

Stanwell also highlighted improvements after migrating its forecasting models to the Azure-based environment, reporting 30% higher accuracy in predicting electricity demand for certain time frames. It attributed the change to broader access to data sources and updated modelling approaches.

More accurate demand forecasts can influence which assets run and when, and can shape short-term purchasing and bidding decisions. Forecasting also matters for operational planning during extreme heat, when air-conditioning demand can rise sharply.

Risk modelling and market scenario planning were other areas of focus. Stanwell said simulations now run up to 15 times faster in some cases. Such modelling typically tests how changes in prices, demand, fuel costs or supply conditions could affect outcomes.

Stanwell positioned SMP as a company-wide system rather than a standalone project. Lin said teams focused on integrating it with existing processes and controls.

"We didn't want this to be a siloed science project," he said. "It had to tie into everything - data pipelines, reporting tools, user access controls - so that the insights flow to where decisions are made."

Stanwell expects to keep developing the platform as electricity systems become more complex, driven by greater use of distributed energy, electric vehicles and more dynamic grid operations.

"We're just scratching the surface," he said. "The beauty of AI is that it gets better with more data and use. Every day our platform is learning from new situations. We're excited because as the tech grows, so do the benefits for our customers."