Renewable energies represent a significant opportunity to meet the world’s energy needs while maintaining a viable environment for generations to come. The future electricity grid will feature millions of intermittent and distributed generation sources, will support a significant penetration of electric vehicles, and will give greater incentives and control to consumers to optimise their energy usage. It will also require an unprecedented level of automation, to self-manage, self-reconfigure and self-heal. This vision challenges the human-controlled, top down management style of the current grid which relies on the existence of few, predictable and rapidly adjustable fossil fuel generators. It calls for a fundamental paradigm shift in the way power systems are planned and operated, underpinned by a new generation of communication, control, data analytics, and optimisation technologies.
The Future Energy Systems Team is building new technologies based on mathematical optimisation and artificial intelligence to support the future of energy systems and the transition from today’s power systems. Members of the team are affiliated with Data61’s Optimisation Research Group and with the ANU College of Engineering and Computer Science. The team is also associated with the ANU Energy Change Institute.
Our research is focused on four major aspects of the emerging grid, and on the scientific and technological challenges which arise in enabling disciplines of artificial intelligence and optimisation.
We automate and optimise planning and operational decisions to enable the future grid to operate economically and reliably under the dynamic and unpredictable conditions arising with renewable energy.
A vision of the future grid is as a network of carbon neutral communities called microgrids, which balance their own renewable generation, storage, and loads, and provide ancillary services to the grid. We are investigating the optimal design and operations of microgrids.
We automate power system restoration following incidents ranging from minor outages to natural disasters. This reduces outage time, fines for utilities, and costs to society.
We design and evaluate incentive mechanisms that encourage consumers to shift demand to enable renewable generation and reduce network peaks. We also develop optimisation software that helps consumers to make optimal decisions about their energy consumption and reduce their electricity bills.
Our research is powered by a host of different technologies.
Networks of power components form complex systems governed by the laws of electromagnetism. We strive to understand and characterize these complex systems by developing novel approximations and relaxations, which exploit the physical properties of power networks and are efficiently computable using cutting-edge optimisation technologies.
Optimisation problems arising in Energy Systems are both discrete and nonlinear: integer variables are needed to model combinatorial aspects of decision making problems while nonlinear functions accurately describe physical properties and operational constraints. We attack the challenges in modelling and solving relevant classes of non-convex Mixed-Integer NonLinear Programs (MINLPs).
To support studies of the optimal development and operations of future electricity grids, we are developing SmartGridToolbox, an event-driven, agent-based simulation Application Programming Interface (API). It incorporates power flow analysis whilst allowing customised energy sources, end-uses and control strategies to be simulated in a flexible, extensible and scalable manner.
This general artificial intelligence technology is key in building optimised control plans and schedules in areas such as home and building automation systems and power systems operations. We develop energy-aware planning and scheduling techniques to solve problems of this type.
It is important to provide solutions that are robust to variations in load, to intermittent generation, and to uncertainty in weather conditions and the network state. With this goal, we research stochastic optimisation methods and their integration with forecasting.
Faults regularly occur on energy systems, due to unfavourable weather conditions, ageing equipment, overloading situations, or human error. We develop new model-based diagnosis techniques that locate faults and intelligently filter alarm flows to help system operators take adequate remedial actions quickly.
Here are a couple of videos to demonstrate the scope of our research.