There is enormous industrial electricity demand growth potential in Australia – from electrifying loads such as steelmaking, from traditional industrial sectors such as mining and from new sources such as data centres, water desalination, and minerals processing (and eventually possible hydrogen production). But there is also a lot of uncertainty about how and when this growth may happen, making forecasting industrial demand a challenge.
The Australian and South Australian governments recently asked us to explore ways to improve the forecasting of industrial demand for electricity, particularly in industrial regions entering a period of energy transition and growth.
Below, we share the method we developed, which accesses rich new data to improve the accuracy of forecasts and to overcome “optimism bias” about the likely timing and success of new projects.
Industrial load is hard to predict and optimism bias in the forecasts is a problem
Residential and smaller business customer electricity demand can be forecast based on well understood drivers such as:
- population growth,
- economic activity,
- weather, and
- new technology trends including rooftop solar, batteries and electric vehicles.
Forecasting large industrial electricity demand has always been far more reliant on project specific information, and a lot of this information needs to come directly from project developers. This is where optimism bias can come in, as developers may be overly optimistic about project timing and success. Naturally they also want their projects included in forecasts as soon as possible to allow time for network capacity to support their project to be built.
The official electricity demand forecasts produced by the Australian Energy Market Operator (AEMO) do account for some optimism bias, particularly the short-term industrial demand forecasts. However, the inclusion of more speculative industrial loads across their broader range of forecasts has the potential to introduce optimism bias.
Independent evidence on industrial project development can improve forecasts
To improve the quality of industrial demand forecasts, information is needed on actual project experience, including:
- the proportion of projects that progress from an announcement to successful development,
- what the typical project development stages are, and
- how long each of these stages typically take.
Fortunately, the Commonwealth Government has been compiling this project development data since 2012 in its Resources and Energy Major Projects (REMP) database. The database includes around 1,150 unique projects.
Our method to use this data included:
- Combining the separated annual data to accurately track projects through time as they progressed through development stages.
- Using the collated data to produce estimates of projects’ probabilities of success and timing, and from this a probability-based distribution of industrial demand outcomes, using a Monte-Carlo simulation model. The project probabilities can be produced across all of the projects in the dataset or for sub-sets such as individual industries or regions.
As the dataset is historical, it estimates the timing and success of existing resource and energy projects but has more limited information on new and emerging industries such as data centres and hydrogen production. This means it may not be as helpful to estimate the likely timing and success of projects in these new industries. However, it was still useful to inform sensitivity analysis on potential demand from these sectors.
Results from implementing new data processes/frameworks
We trialled and applied this method successfully for the Upper Spencer Gulf region of South Australia, working with the Commonwealth and South Australian governments, the South Australian electricity networks (ElectraNet and SA Power Networks) and AEMO.
The report of this demand forecasting pilot can be found here: Upper Spencer Gulf energy demand pilot report
The use of richer data and more probabilistic forecasting approaches is expected to improve the accuracy of forecasts, helping to avoid the substantial cost of over- or under-investment in electricity assets.
Extension and knowledge transfer
Governments and project developers are likely to find this data on industrial project development experience, and the tool we developed to make it user friendly, helpful well beyond improving forecasts of industrial electricity demand. For example, this information can inform broader industry and infrastructure planning, investment decisions and project/risk management.
An important element of our project delivery was ensuring that our government clients were able to continue to use the data, tools and methods developed. Training in how to develop and continue to use the data and forecasting tools (including training in the underlying ‘R’ programming) were essential to embedding this method into their future forecasting models.
As our energy systems rapidly transform, producing better informed forecasts and better understanding the probability of different outcomes is critical to decision making.
To find out more about the project, contact our team.
