On December 14, 2024 – three weeks after I published the last exciting installment in this series of posts – our new Redflow ZCell battery, which replaced the original one which had developed a leak in the electrode stack, itself failed due to a leak in the electrode stack. With Redflow in liquidation there was obviously no way I was getting a warranty replacement this time around. Happily, Aidan Moore from QuantumNRG put me in touch with Jason Litchfield from GrazAg, who had obtained a number of Redflow’s post-liquidation stock of batteries. With the Christmas holidays coming up, the timing wasn’t great, but we were ultimately able to get the failed unit replaced with a new ZBM3.
At this point the obvious question from anyone who’s been following the Redflow saga is probably going to be: why persevere, especially in light of this article from the ABC which speaks of ongoing reliability issues and disturbingly high failure rates for these batteries. That’s a good question, and like many good questions it has a long and complicated answer.
The technical path of least resistance would have been to migrate to a small rack of Pylontech batteries, as these apparently Just WorkTM with our existing Victron inverter/charger gear. The downside is they’re lithium, so a non-zero fire risk, and our installation is currently in the crawl space under the dining room. If we switched to lithium batteries, we’d need to arrange a separate outdoor steel enclosure of some kind with appropriate venting and fans, probably on the other side of the driveway, and get wiring to and from that. My extremely hand-wavey guess at the time was that it’d easily have cost us at least $20K to do that properly, with maybe half of that being the batteries.
The thing is, I remain convinced that flow batteries are in general a better idea for long-term stationary energy storage than lithium. This article from the Guardian provides a quick high-level summary of what makes flow batteries different. What I really want to be able to do – given Redflow is gone – is migrate to another flow battery, ideally one that actually lives up to the promise of multi-decade longevity. Maybe someone will finally come up with a residential scale vanadium flow battery. Maybe someone will buy Redflow’s IP, carry on their work and fix some of their reliability issues (the latest update from the liquidators at the time of writing says that they have “entered an exclusive negotiation period with a party for the acquisition of Redflow Group’s intellectual property (IP) and certain specific assets”). Maybe we’ll even see a viable open source flow battery – I would love for this to happen, not least because if it failed I’d probably be able to figure out how to fix the damn thing myself!
Leaving our current system in place, and swapping in a new ZBM3 meant we could kick the migration can down the road a ways. It bought us more time to see what other technologies develop, and it cost a lot less in the short term than migrating to lithium would have: $2,750 including GST for a post-demise-of-Redflow 10kWh ZBM3 (although shipping was interesting – more on that later). The real trick going forwards is seeing exactly how far down the road we’ll be able to kick that can. How can we ensure the greatest possible longevity of the new battery?
Why do ZCells Fail?
The ABC article puts it down to manufacturing problems, notably a dependence on repurposed third-party components. While I can see that dependence causing all sorts of extremely irritating manufacturing and design issues, I’m not entirely convinced this is the whole story. I will freely admit that my personal sample is very small, but my two batteries both failing due to electrode stack leaks? If a hose split or a pump had died, or some random doohikey let the magic smoke out, then OK, cool, I get it, those I can see being repurposed third-party components. But these failures were apparently in the electrode stack, and I’m struggling to see how that could be a repurposed third-party component. If nothing else the stack (and the tanks) are surely the pieces that Redflow manufactured themselves. This is their core technology. What could be causing stack leaks? Are they just poorly manufactured, or is there some sort of chemical failure at runtime which physically splits the stack? Or something else? Bear in mind that this is all speculation on my part – I’m neither a chemist nor a battery manufacturer – but I know what I’ve seen, and I know what I’ve heard about leaks in other peoples’ batteries.
On the chemistry front, I found a paper from 2023 entitled Scientific issues of zinc-bromine flow batteries and mitigation strategies. This was authored by a bunch of researchers from the University of Queensland and the former CTO of Redflow, and highlights hydrogen evolution, zinc corrosion and zinc dendrite formation as the fundamental issues with zinc bromine flow batteries. I sincerely hope the authors will forgive me for condensing their fascinating ~9,000 word paper into the following 95 word paragraph:
When the battery is being charged, zinc is plated onto the electrodes. During discharge, the zinc is removed. Dendrites (little tree like structures) can grow due to uneven zinc deposition, or due to hydrogen gas evolution. Left unchecked, dendrites can puncture the separator between the electrodes and lead to short circuits. Additionally, hydrogen gas generated by the battery can raise the electrolyte pH. If the pH is too high, solid zinc can clog a membrane in the stack. If the pH is too low, it can cause zinc corrosion which can make the battery self-discharge.
What if Redflow just never completely solved or mitigated the above issues? Could a dendrite puncture not just the separator, but actually split the stack and result in it leaking? Could clogged membranes combined with hydrogen gas create enough pressure to do the same?
We know that ZBMs have a maintenance cycle which runs at least every 72 hours to first discharge the battery then (theoretically) completely strip the zinc from the electrodes over a subsequent two hour period. We also know that ZBMs have a carbon sock which sits inside the zinc electrolyte tank and helps to keep electrolyte pH in the correct operating range. This needs to be replaced annually.
What if 72 hours is still too long between maintenance cycles? If you search back far enough you’ll find that the maximum maintenance period was originally 96 hours, and I assume that was later revised down to 72 hours after experience in the field. I’ve had subsequent correspondence which says that even more frequent maintenance (24-48 hours) can be better for the batteries. I’ve also encountered a curious intermittent fault with the ZBM3 where occasionally the Strip Pump Run Timer in the battery operates at half speed. If that happens and you don’t notice and reset the battery, the maintenance cycle will actually occur every 144 hours, which is way too long.
In the past I’ve observed frequent high charge current warnings in the Battery Management System (BMS) logs. This is actually normal, as by default the charge voltage is configured to be 57.5V, and there’s a separate high current voltage reduction setting of 1V. The idea is that this will try to make the battery charge as quickly as possible, and if the current gets too high, it will drop the charge voltage dynamically by 1V, which results in current reduction. Is it possible this variable (i.e. potentially uneven) charge current results in uneven zinc deposition?
I’ve also noticed that the battery State of Charge (SoC) calculations get sketchier the longer it’s been between maintenance cycles. If I have maintenance set to 72 hours, then at the end of the maintenance cycle, the battery fairly reliably still reports about 7% SoC. With a 48 hour maintenance period, it reports about 3% SoC at end of maintenance, and with a 24 hour maintenance period, it’s more like 1%. Once maintenance completes the SoC is reset to 0% automatically (because the battery really is empty at that point), but this got me thinking… If the SoC calculation is off, is there any way the battery could inadvertently allow itself to overcharge? Given the numbers above are all obviously overestimates I hope it’s more likely that the battery undercharges, but still, I had to wonder.
Tweaks to Optimise Battery Lifespan
Aidan suggested three configuration tweaks which Redflow had told him to try to potentially help optimise battery lifespan:
- Set the maximum SoC to 90% (rather than 100%)
- Set the charge voltage lower, to ~55.5V
- Perform maintenance as often as possible
These are all done via the BMS. The maximum SoC and maintenance time limit are set on the Battery Maintenance screen under Capacity Limiting and Maintenance Timing respectively. I went with 90% SoC as above and 48 hour maintenance. The charge voltage is on the EMS Integration screen. I’ve used the following settings:
- Normal Charge Voltage: 55.5V
- Charge-Blocked Voltage: 52.5V
- Discharge/Maintenance Cycle Voltage: 55V
In my case, the Normal Charge Voltage was originally 57V, and as I dropped it by 1.5V to get to 55.5V, I dropped the Charge-Blocked and Discharge/Maintenance Cycle voltages by the same amount to arrive at the above figures.
Dropping the maximum SoC means that the battery can’t get completely full and stay there for a long time. This must reduce the total amount of zinc plated on the electrodes, which I hope helps reduce dendrite formation. I also found when reading the paper mentioned earlier that “H2 evolution occurs mostly near the top of charge with mossy or spongy like zinc being plated”, which looks like another good reason to avoid fully charging the battery.
Dropping the charge voltage necessarily reduces the charge current and I assume keeps it much more even than it would be otherwise. I have not seen any high charge current warning since making this change. On the other hand, it does mean the battery charges slower than it would otherwise. I did a little experiment to test this, just watching the figures for amperage and kW the BMS gave me when I tweaked charge voltages:
- 55.5V charges at 30A or about 1.7kW
- 56.0V charges at 35A or about 2.0kW
- 56.5V charges at 40A or about 2.3kW
This means I’m not using the battery as effectively as I could be with a higher charge voltage/current, but if this serves to extend the battery life, I think it’s worth it under the circumstances.
It’s important to keep an eye on is the Strip Pump Run Timer, which went weird on me a couple of times. I really should write a little script to automatically warn me if it starts running at half speed, but I’ve been habitually looking at the BMS briefly almost every day since the system was installed, so I noticed when this problem occurred because the maintenance timing was off. To reset a battery that gets in this state, go to Tools: ZBM Modbus Tool and write the value 0x80 to register 0x2053. This will appear to fail because it immediately resets the unit which thus never reports a successful write, but it does the trick.
Some time in the next six months I’m going to need to beg, borrow, steal or figure out how to manufacture carbon socks. The good news is that this time the replacement procedure is going to be really easy, because unlike the ZBM2 (where you had to mess with some pipe work) and my previous ZBM3 (where there was a cap on the side which in my case would have been completely inaccessible due to proximity to a wall), this one has an easy access screw cap on the front of the electrolyte tank.
Accessing the BMS without the Redflow Cloud
The Redflow cloud went offline in late October 2024. This allowed remote access to the BMS, and I understand that some Redflow customers were unaware that it’s possible to access the BMS locally without the cloud. The Redflow cloud allowed firmware updates, and also let Redflow staff monitor batteries and configure them remotely, but it is not actually a hard requirement that this system exist in order for the batteries to continue to operate.
One way to access the BMS locally is via the wifi network on the BMS itself. If this is turned on, and you search for wifi access points you should find one named something like “zcell-bms-XXXX”. The password should be “zcellzcell”. Once you’re connected, open a web browser and go to http://zcell:3000. If that doesn’t work, try http://172.16.29.241:3000. This should let you see the BMS status. If you try to make any configuration changes it will ask you to log in. The default username and password are “admin” and “admin”. These can be changed under Configuration: Users.
The other way of accessing the BMS is to connect to whatever the IP address of the BMS is on your local network. The trick in this case is figuring out what the IP address is. I know what mine is because I logged into my router and looked at its list of attached devices.
Given the Redflow cloud is down and Redflow is out of business, I would actually suggest going into the BMS Site Configuration screen and unchecking the “Enable BMS cloud connection” and “Allow Redflow access to system for service intervention” boxes. There are two reasons for this:
- There’s no point having the cloud connection enabled if the cloud is gone. If this is left on, a process on the BMS will continually try to connect back to cloud.zcell.com which is never going to work, and you’ll perpetually have this irritating little “Cloud connection has been lost” indicator.
- Possible subsequent risk of hacking. I assume the following is pretty unlikely, but it’s not impossible. Imagine once the zcell.com domain name expires if some malicious actor were to re-register that domain name, then set up their own cloud.zcell.com service. BMSes with the cloud connection enabled would blindly try to connect to it, which could allow a sufficiently sophisticated attacker to hack back into the BMS and mess with it, or potentially use the BMS as a jumping off point further into a site’s local network.
Personally I hope whoever buys the Redflow IP will turn the cloud back on, in which case the above advice will no longer apply.
Shipping ZBMs
Individuals such as myself can’t just ring up a random courier and say “Hey, can you please go to New South Wales, pick up a 278kg crate with hazchem stickers that say ‘corrosion’ and have pictures of dead fish, and bring it to me here in Tasmania?” The courier will say “Hell no”, unless you have an account with them. Accordingly I would like to thank Stuart Thomas from Alive Technologies through whom I was able to arrange shipping, because his company does have an account with a courier, and he was also after some batteries so we were able to do a combined shipment. If anyone else is looking to move these batteries around, the courier in this instance was Imagine Cargo. I understand Redflow in the past used Mainfreight and Chemcouriers. In all cases, the courier will need to know the exact dimensions and weight which are in the manual, and will want a safety data sheet. Here they are:
Further thanks to Stuart and Gus (whose flatbed truck almost didn’t make it up our driveway) for last mile delivery, swapping the new ZBM3 into the old enclosure, and getting the damn thing in under our house.
Final Thoughts
It’s disappointing on many levels that Redflow went under, but like I said earlier, I remain convinced that flow batteries are in general a better idea for long-term stationary energy storage than lithium. I find it interesting that the sale of Redflow’s IP includes “specific assets and shares in Redflow (Thailand) Limited”. Given that’s where the manufacturing was done, could that indicate that the buyer is interested in potentially carrying on further development or manufacturing work? The identity of the buyer remains confidential right now, and final settlement is still a year away, so I guess we’ll just have to wait and see.
Our new ZBM3 was commissioned on March 18, and has been running well ever since. I’ve done everything I know to do to try to ensure it has a long and happy life, and will continue to keep a very close eye on it. There will be followup posts if and when anything else interesting happens.
Some time rather earlier in this journey, I found an easter egg in the BMS, which I didn’t mention in any of my previous posts. I think that might be a nice note to finish on here.