20 years since my engineering thesis on energy storage, did i get anything right?
I was horrified to realise recently that my engineering thesis assessing energy storage technologies is now 20 years old. Once I overcame the nausea at re-reading my own work 20 years on I found some interesting tidbits, and not many correct forecasts I’ve got to say. But how they were wrong is interesting, and sets us up for lessons in future.
Background
Studying at UTS in 2006 I was lucky to get roped into a project supporting a United Nations initiative to provide renewables to Pacific Islands. My project was centered on the island of Niue, and what would be required to run the island from a single 2MW wind turbine. “Energy Storage” is what’s required, and so my thesis was an assessment of the energy storage technologies available at the time and which were the most appropriate for this application.
My conclusion was pretty clear: sodium-sulfur was the most promising technology for large-scale storage.
Lithium-ion, meanwhile, was disregarded entirely as too expensive in my decision matrix. I estimated lithium-ion at more than $600/kWh, and concluded that it couldn't economically provide the roughly 11.5 MWh of storage I was looking for.
Twenty years later, that looks rather different.
Here are five reflections on my analysis and where it went wrong.
Photo of the conclusion table from the original. I’ve learned how to put lines in tables since then.
Sodium-sulphur batteries did not take off
“Of the energy storage options available, Sodium-Sulfur batteries represent the most reliable, mature and energy capable solution to the problem of energy storage associated with the Niue Renewable Energy project.”
That was all correct, they were the most mature, but they didn’t progress much beyond there. These batteries use liquid sulphur and liquid sodium as their electrolyte, which need to be kept at over 300 degrees C. That sentence sounded like a death-knell to me, but at that point there were around 20MW/160MWh deployed, so maybe I was over-estimating those risks? If the risk is manageable though, sodium-sulphur is a really robust battery that can handle awful charge-discharge cycles, with high efficiency and no loss of capacity.
In 2011 there was a fire at the Mitsubishi Materials’ Tsukuba plant, which had been operating since 2009. 10 of the 40 50kW modules caught on fire and took more than 2 weeks to be extinguished. This didn’t kill the technology outright, but seems to have limited its investment and deployment scope.
By the early 2020s the technology sat entirely with NGK, who have deployed more than 200 projects and 4GWh of total storage. Compare that to about 13GWh of batteries in Australian households now.
Overall sodium-sulphur is still a good battery technology, and seems to have some useful technical niches, it’s just not as good as the alternatives, which really took off
Lithium-ion batteries - did take off
In 2006 I was sceptical about the prospects for lithium in grid storage applications.
“While Li-ion batteries took over 50% of small portable market in a few years, there are some challenges for making large-scale Li-ion batteries. The main hurdle is the high cost (above $600/kWh) due to special packaging and internal overcharge protection circuits.”
That was all correct then, but the rise of mobile computing, followed by increased production of electric vehicle batteries have meant that lithium has become a lot cheaper now than I could have imagined in 2006.
Our World in Data has a cool graph with a slider that allows you to select the start and finish years for lithium prices. Their inflation adjusted price per kWh for 2006 was $687 USD, down to $78 in 2024. I’ve seen fully engineered packs for mobile applications under that price in 2026.
It would have been hard to predict in 2006, but there were some signs this might happen. In the early 90s lithium cells were closer to $10,000, so they’d already come down >90%. In parallel we had long cost curves for solar and computer chips showing how manufacturing scale brings down costs.
And there was a fundamental physical attribute that pointed the way: lithium has the highest energy density, and because of its chemical attributes, lithium was likely to be the choice for mobile applications. That was predictable.
Tony Seba had the earliest forecast I know of that predicted this, in his excellent book Clean Disruption of Energy and Transport, but even someone as prescient as Tony didn’t forecast this until 2014. With that in mind, this was the right decision at the time, but it’s another example of “the cost now won’t be the cost always”.
Flow-Batteries - the technology looked promising, but that wasn’t the product
I was cagey about flow batteries in my report, I was only learning about them for the first time while I was writing, but in the period afterwards I thought they might be the solution that gets widespread deployment. I liked the idea that per project it was possible to size the storage tanks for the right amount of energy needed, and the inverter can be sized for the peak power. Seems simple? But that requirement for bespoke engineering per project is probably the reason it hasn’t taken off.There have been some attempts to “productise” flow battery technology, in particular Red Flow’s residential scale ZBr, but in some ways that project seemed to highlight the shortcomings of the technology. In deployment the Red Flow battery was great in hot weather and challenging environments, perfect for telecoms and mining applications in Australia. But it came with this operational limitation of a “deplating” cycle; one hour out of every 24 the battery needs to run a maintenance cycle. So if the deployment is a critical power problem, the site needs the calculated number of batteries, plus one so there’s always a spare to cover a deplating cycle. This effectively pushed the price up by 25% or more.Then for the Red Flow business the batteries never got cheaper. I suspect that’s because the components; pumps, tanks and electrical control gear, are all mature and don't have much scope for cost improvements. Red Flow closed permanently in late 2024.
Power quality - more important than I realised then
I mentioned power quality when assessing the technologies, but at the time didn’t fully understand what it meant. I had a vague notion that it was related to handling voltage movements and maybe frequency, but that was about it. I’ve learned in great detail since then what it means and how vitally important it is.
Most energy storage projects are managing against the risk of running out of electricity. What I didn’t appreciate back then was that you can “run out of electricity” at millisecond scale as well as hours and weeks, and the millisecond scale still results in the lights going out.
In a typical AC system the devices connected have some protection or sensing and will turn off if there’s too much voltage movement, or if the frequency drops, both of which can happen if there’s a large load disturbance on the grid. A battery’s ability to manage those changes in load speaks directly to “power quality”, and the long term system stability.
A stark example: at Chargefox we designed and built a battery to support EV charging at the Goulburn Gateway Petrol Station. Arguably the most technically advanced petrol station in the world for a while there.
The chargers at Goulburn were behind an isolation transformer, designed to separate the neutral of the chargers from the rest of the grid. Transformers like this need to be “energised”, and so when they are first powered on there is an in-rush of current, sometimes way beyond the system design. In this case the transformer demanded 10,000A for about a third of a second, through a 200A connection. This in-rush made power-up of the site very difficult, as it either tripped the site breakers on in-rush, or the charger breakers tripped on voltage drop. It was infuriating.
The solution, provided by Power-Tec was an extreme use of the power quality functions of the battery we had deployed. Another important characteristic of grid batteries is the ability to provide a fault current. Most electrical safety equipment works on the theory that failing equipment will demand too much power. If that happens the circuit breaker senses it and trips. But in some cases, if the power source is weak, and there’s a fault, then the system doesn’t receive enough power to trip the breakers. Instead maybe a 200A breaker gets 250A instead of 2000A, and overheats instead of tripping. Our battery at Goulburn was engineered to provide an 8000A fault current for half a second.
The battery and transformer were both behind the site circuit breaker that kept tripping. When the transformer was first energised the current through the site breaker was reaching 10,000A, but only for a short time. To manage this, we tuned the battery so that if it detected that the transformer was energising, the battery would discharge its full fault current into the circuit, so instead of pulling through the main breaker and tripping, the power would come from the battery. Then once the battery detects that the in-rush is over, it ramps back to normal operation. This all happens within about ⅓ of a second.
So power quality is the ability of your power source to keep everything running. And I know now that’s as much about the chemistry of your batteries as the quality of your power electronics and control, and that this makes as much difference as “running out of electricity”.
Manufacturing scale beats cool engineering
I’ve hinted at this throughout, but the overarching lesson from this work has been that manufacturing scale and deployment is what drives technology uptake, much less than how cool or good the solution seems to be.
Considering the examples of solar panels and lithium ion batteries for stationary storage, I think there are two reasons for this.
The obvious one, manufacturing scale brings down the cost of the items. Solar and battery hardware keeps getting cheaper as more factories to produce them keep getting built. And then once those factories are at capacity new ones get built that produce slightly cheaper panels or cells.
With this simplification through production, it becomes possible for more people to install it, also bringing down the cost of installation in parallel. Thinking about flow batteries above, each system design requires engineers and drawings to get it all right. That adds cost. But solar panels and batteries can be shipped to a local electrician and installed in an afternoon.
This is the Virtuous Cycle Tony Seba and others have been talking about. Manufacturing capacity brings down prices, lower prices lead to more deployments, more deployments create demand for new factories, new factories reduce the manufacturing costs. At this point I don’t see that process stopping.
Lessons for the Future
How can I apply these lessons to my current work? The first practical takeaway is that it’s hard to pick winners, so make sure your choices don’t lock you in.
When we're planning EV charging infrastructure or fleet transitions today, we're making decisions that will last for 10, 15 or 20 years. It's tempting to pick the technology that looks best today and build the whole system around it.
But if I've learned anything from being wrong about batteries, it's that today's best technology isn't necessarily tomorrow's best technology. The technologies that are being manufactured at huge scale are likely to get cheaper and better, while technologies that require bespoke engineering may not.
So I think the answer is to build flexibility into the system wherever you can. Don't unnecessarily lock yourself into a particular charger, battery, vehicle architecture, communications platform or energy supplier just because it looks like the best option today.
The best technology decision might be the one that leaves you with the most options later.
And there’s a related lesson, which informs the creation of Veeworks, our vehicle electrification business; vehicles that are common, like passenger vehicles and last-mile delivery vans will be mass produced by the OEMs, so don’t focus on electrifying those. But it might take years or perhaps decades for the OEMs to turn their attention to the specialist and technical vehicles we use in Australia, so let’s focus on electrifying those first. Watch this space for more announcements soon.