AI in Australia's electricity sector
Australia is in the midst of a major political debate over energy policy, pricing and surety of supply to meet residential and business demand. One of many levers to achieve this — which appears to hold the most promise — is the rise of the microgrid and of artificially intelligent generation and distribution systems.
It’s hard to mention artificial intelligence (AI) without some reference to long-running efforts around the smart grid. But where smart grids have been largely focused on creating intelligent energy distribution and flows within the existing grid structure, the focus of the industry is now shifting to use intelligence to restructure the grid completely by bringing in new, diverse and decentralised energy sources.
The likely future state is a complex network of generation and distribution assets that can intelligently match demand and supply and operate in a semi-autonomous fashion. A network that is capable of measuring, balancing and acting on the individual needs of customers, whether the customer is residential, commercial or industrial.
Australia is already stepping towards this future, but there is a real need to remove barriers to adoption — notably regulation, legacy thinking and entrenched business models — that could stifle progress.
Microgrids rise up
The rise of microgrids and other types of embedded energy networks is a clear sign of progress on this front.
A microgrid is a smart private energy-producing network for a commercial, industrial or residential precinct. It consists of some form of renewable energy generation (usually solar), an energy storage system, and a network distributing that power to users. It is designed to work in collaboration with the regular electricity grid, allowing communities to draw upon a mix of their own and other energy resources.
Microgrids are popping up across Australia, from the Pilbara mining town of Onslow to South Australia’s Yorke Peninsula and the far north coast of New South Wales.
For communities and organisations served by microgrids, it is envisioned that they will draw power from both their microgrid and the regular grid. Which source they favour at any one time will depend on a number of factors, such as price, and balance of supply and demand. This will be managed dynamically, and is likely to involve an increasing amount of AI. AI will be used to manage much of the orchestration, both of components within the microgrid and the interaction between different microgrids and the grid itself.
Adding AI to coordinate the pieces
New AI tools are expected to be created to aid this orchestration and management. In this context, AI refers to a convergence of the several key computer science trends and, in particular, machine learning, optimisation, forecasting and operations reseach.
By building on its global leadership in this space, Monash University’s key grid research centre, the Monash Grid Innovation Hub, and Indra hope to contribute to this through the Smart Energy City project, which will see the development of a microgrid at Monash’s Clayton campus. Using Indra’s InGRID Active Grid Management (AGM) software platform, the microgrid will enable control of various distributed energy resources (DER), including a minimum of 1 MW of solar panels, 20 buildings, electric vehicle charging stations and 1 MWh of energy storage. No single command and control system will schedule all of these different components to work together. The project aims to use applied AI and optimisation techniques to make that work.
While we know how this should be done, we’re now working out how best to do it in practice. This includes the kind of price signals that might be required to mediate optimisation decisions — for example, how does the market reward those who are more flexible with their energy load requirements than others and are willing to dial their use down in periods of high value to the grid?
These kinds of questions sometimes fall under a subset of smart grid research known as transactive energy, which aims to find ways for energy producers and consumers to balance supply and demand, and for each side to be appropriately compensated. Analyst firm Navigant Research expects Australia to be one of only a handful of markets worldwide to be in the position to shift from trials to large-scale deployments in this area.
Beyond transactive energy, Monash University plans to use its microgrid to create more novel uses of AI to aid converged grid-microgrid operations.
If the future of Australia’s electricity supply is going to involve a higher amount of solar and other renewable resource inputs, it will be critical to accurately forecast how much power can be produced by these sources, improving significantly on what is currently possible. We also need to be able to forecast the output of renewables better, as well as optimise the scheduling of the energy resources in a way that is robust to the residual uncertainty. It is these optimisation and forecasting methodologies that are still being developed.
Monash University is hoping to aggregate data from a variety of different places to improve forecasting and operational planning. These places may include satellite data as well as a network of cameras pointed towards the sky to track cloud cover, and estimate what impact this will have on the output of nearby renewable energy sources. This kind of intelligence could allow participants to optimise the energy mix at any one point. A similar sky camera set-up is already in use for the University of California San Diego’s (UCSD) microgrid, and Monash University is keen to test a similar set-up in Australia on its own microgrid.
The capacity to forecast renewable generation with a 24–48 h lead time and a high degree of precision, and then to dynamically adapt this forecast and optimise operation decisions to changing conditions, would represent a major breakthrough for the sector, and for the rise of Australia’s intelligent grid.
It’s past time for industry to get behind the research
To shorten the odds further of Australia achieving this endgame and maintaining its recognised lead in the transactive energy space and the orchestration of DER, it is critical that more parts of the electricity sector get behind such research efforts and test different approaches to reach the same destination. It will be important for all players to keep an open mind and enthusiastically embrace the progress of the technology.
While there is clear evidence of some operators trying to engineer any transition to suit their existing models (and sunk investments), this approach is fraught with risk, particularly as it is difficult to anticipate all possible effects as industry evolves.
In the same way, backing a single path to embedding intelligence in the electricity network is currently problematic; operators that succeed will likely be those that are open to multiple paths and flexible in allowing pivoting in response to critical changes in technology, since it is not at all clear which of the many available and often overlapping technologies will go on to make the biggest mark.
The shift to artificially intelligent electricity networks is underway and it will totally transform the way value is shared between consumers, producers and transporters of energy. Like any disruptive change, it is unlikely to be as orderly a transition as most of us would like it to be. However, the sooner the industry can begin fully embracing the technology and the changes, and the sooner regulations are loosened, the more an orderly transition to Australia’s energy future is likely to be possible.
An electrical enclosure's ability to protect the controls contained within is impacted by...
New 'time reversal' technology enables faster fault-finding within power networks, as...
New research is redesigning how electricity is distributed within power grids, with the aim of...