A few months ago we had a final team meeting for our project investigating the potential benefits of neighbourhood-scale batteries, funded by the Australian Renewable Energy Agency. We felt buoyed by our results, demonstrating the technical capability of neighbourhood-scale batteries to support an increased amount of rooftop solar in our suburbs. Our social research had also given unique insights into the risks and benefits of this technology, as perceived by a range of stakeholders, including both householders and energy industry experts. We felt that our multidisciplinary approach had been a great success; the social research had informed the engineering design and vice-versa.
But then we wondered how, in practice, can we include those stakeholder perceptions of the risks and benefits around this technology, into our engineering design? Clearly, they need to be included in the design for this technology to be accepted.
It felt like a lightbulb moment when we realised that we could encode those stakeholder priorities into the battery algorithm i.e., the set of instructions that tells the battery how and when to charge and discharge. Even better, we can allow these priorities to be flexibly encoded, according to the needs and expectations of each community and each scenario. In our paper published this month in Nature Energy, we demonstrate how stakeholder-identified risks and benefits can be translated into battery algorithms, providing six different examples of how this can be done.
The carbon-savings algorithm, for example, operates the battery to minimise the carbon emissions of the neighbourhood’s energy generation mix. A consistent theme among householders was the value of batteries in enabling more renewables and the decarbonization of the electricity system. This algorithm might be chosen, for example, by a city community who choose to pay a little more for their energy in order to prioritise local decarbonisation.
For the self-sufficiency algorithm, the battery operates to maximize the energy independence of the community, by storing locally produced solar energy. Algorithm design here was influenced by citizens’ enthusiasm for the idea of local generation and local use of solar resources located in their own neighbourhoods. This algorithm might be chosen, for example, by a community from a coastal region of Australia which is at risk of isolation from the main grid because of bushfire or other natural disasters.
The timer algorithm instructs the battery to follow a simple fixed daily schedule. Although not the financially optimal choice, we tested this algorithm in response to concerns that were voiced to us about the ‘gaming’ of the energy market by incumbents. This algorithm is easy for non-experts to understand and makes it easy to monitor the distribution of benefits. Stakeholders often expressed a desire for transparency and explainability, and a desire for autonomy and control over their energy choices.
In practice, battery algorithms can be designed to optimize for more than a single objective e.g. financial costs and decarbonisation values. However, we demonstrate in our paper that multiple values cannot always simply be ‘value stacked’ in an algorithm, rather, some values are inescapably in tension with one another, and trade-offs are required.
These trade-offs will be inherently political, particularly because the values being traded off may be based on unrelated metrics (for example, dollars versus algorithm explainability) and some are not naturally quantifiable. For example, it’s easy to design an algorithm to maximise revenue for a battery owner but challenging to consider how to design an algorithm that maximises energy users’ autonomy and control. As we conclude in our paper, this will inevitably bias the design of algorithms towards the easily quantifiable.
Research on the non-neutrality of algorithms has grown substantially over the past decade, yet these are relatively new issues to energy researchers like us. Our work highlights that it will be essential to consider how these issues are explicitly encoded into the millions of devices that will underpin our future electricity grid. How can we do this in practice? In our paper, we discuss the need for digital energy technologies to be developed through an ‘algorithmic accountability in action’ approach that aligns the behaviour of these technologies with public values.
In writing this blog post, we took the opportunity to reflect on the research process behind our paper. Working in a multi-disciplinary team was certainly challenging at times. In a normally fast-paced research environment, we had to slow down to explain basic concepts to each other. This can be highly frustrating; it’s easier to simply do our own work in parallel. It can feel like we are achieving less. However, our experience was that, through really listening to each other and respecting other knowledge domains, we found that we could collectively think in a new way and produce ideas that we would not have had working in our disciplinary silos. Ultimately this has been a very rewarding research process and we will continue to adopt and promote a multi-disciplinary approach for the complex challenges we face in the energy transition, including the digitisation of our energy system.