Research shows that if energy professionals and customers agree on what they expect a community battery to do, engineers can write performance algorithms to suit those objectives.

Can you trust a battery to make the best decisions about when to charge and discharge? It depends who owns it, for a start, but most of all it depends who wrote the code that is its book of commands.

As community batteries are deployed to manage solar exports and calm grid disturbances, researchers at the Australian National University wanted to understand the degree to which these assets can be bent to serve their owners or the communities they are plonked in the middle of.

“How an electric vehicle or a battery operates in your home, these things are governed by algorithms coded up by humans,” says ANU battery storage and grid integration research leader Bjorn Sturmberg. “They are not governed by the physics of spinning machines, which is traditionally what we have built our energy system around.”

Let’s face it, the choices imbedded in algorithms are “relatively arbitrary”, Sturmberg says. Designers may want to test the constraints of a technology while maximising revenue (or minimising cost), but their decisions will be biased. Because they are human, of course, those biases will be obscure to them. That’s just how we are.

Systemic concerns

The researchers shared three systemic concerns about how algorithms are deployed: bias of considerations towards the easily quantifiable; inhibition of explainability, and; the undermining of trust and inclusion, as well as energy users’ autonomy and control.

To demonstrate the consequence of those concerns, the team first had to write some algorithms itself – and run the risk of pursuing team-members’ own biases. To minimise the chance of that, they sought opinions among the energy sector and consumers in separate rounds of deep qualitative research, where respondents had their say about how they thought a neighbourhood-scale battery should behave, why it should behave that way and who it should serve.

“We wanted to know who they might benefit, who might be left out and what values they encode,” says Sturmberg, who co-authored the paper with Hedda Ransan-Cooper, Marnie Shaw and Lachlan Blackhall.

With the hopes and expectations of the two groups noted, it was time to code up six algorithms that satisfied a set of five objectives (acknowledging, of course, that an algorithm cannot satisfy everyone). The six algorithms targeted: battery profit; communal savings; carbon savings (where grid power is graded on levels of renewables); co-optimised (maximise battery revenue while minimising the neighbourhood’s carbon emissions); self-sufficiency (use as much locally-produced solar as possible), and; “timer” (the battery charges from 6am to 6pm and discharges for the next 12 hours – wonderfully simple).

Us and them

The “timer” model was like a cat among the pigeons for the energy professionals, Sturmberg tells EcoGeneration. “Even though it goes against so much of what appeals to the economic rationality of energy professionals, a lot of our social research found [consumers] like transparency and trust, things that are valued by citizens but are often ignored.”

Translating qualitative results – opinions, essentially – into quantitative code was something to grapple with, he says. “We don’t claim to have all the answers in this paper; we’re trying to alert the energy sector to the fact that maybe there is more to what people care about than just money.”

Pulling against this egalitarian ideal is a theme heard in the research sessions that locals should benefit from a neighbourhood battery. “There’s an inherent contradiction that you can achieve equity within that local community but not necessarily guarantee equity between local regions,” he says. A follow-up paper aims to explore the design of tariffs such that networks, solar-owning and non-solar-owning customers are all satisfied to varying degrees.

The algorithms were applied to a scenario where a 500kW/1MWh battery is installed within a neighbourhood of 100 households with 6kW of solar each, all connected to the grid.

Fair or square

The purpose of research is not always to make firm recommendations, but the ANU paper includes a spider and radar chart (shown above) that plots the six algorithms against five objectives: customer savings, battery revenue, simplicity, carbon reductions and self-sufficiency.

Of the five, it is the target of carbon reductions that appears to have been satisfied by most of the algorithms. That’s maybe not be surprising, as prioritising clean grid energy and solar self-sufficiency should logically lead to communal savings and battery profit (thanks to surplus solar energy).

“Citizens’ primary benefits and risk were decarbonisation, local independence and simplicity. Of our six algorithms, none is best across all domains, demanding that trade-offs must be made,” the researchers wrote.

The citizen and energy professional groups both recognised that a wayward algorithm would only increase inequality, but how can consumers trust a load of code drawn up by engineers? Sturmberg hopes the research encourages providers to communicate clearly and perhaps favour simplicity in their decisions. The timer and self-sufficiency models are easy to explain, he points out. It could be a case that a jury of non-experts is recruited to have their say on algorithm designs, “to represent the common good”.

Consumers might say that they don’t trust energy providers, and sometimes it’s hard to know if owners of solar are happy with their lot or not, but the ANU paper, Applying Responsible Algorithm Design to Neighbourhood-Scale Batteries in Australia, will hopefully light up some minds in the energy industry. It shows there are many shared benefits and concerns. The demand and supply side are starting to understand each other.