Friends, the long northern hemisphere nights give one a chance to reflect on the wonder of technologies which can ‘bottle the Sun’. Each dark day I dream idly of the Sun’s return in the coming spring, but in more practical moments, I have been collating the statistics describing another year of PV generation and battery storage.
The System
For those readers who have not been paying attention, our solar PV system consists of:
- 6 × 340 W-peak roof panels facing 20° south of West.
- 6 × 340 W-peak roof panels facing 20° south of West.
- 5 × 390 W-peak roof panels facing 20° north of East.
- 3 × 390 W-peak roof panels on a flat roof ’tilted’ to the South East.
Click on image for a larger version.
System#1 (comprising the first 12 panels) was installed in November 2020 and so this is its third complete year of service, and System #2 (comprising the last 8 panels) was installed just one year ago in November 2022. The 5 roof-mounted panels in System #2 face 20° north of east, and so only generate significantly in summer: recall that in the UK the Sun rises 40° north of east at midsummer. The panels are connected via two Solis 3.6 kW inverters to a 13.5 kWh capacity Tesla Powerwall 2 which was installed in March 2021.
Powerwall Battery
The first graph shows the average amount of charging and discharging from the battery through the months of the year – the data are averaged monthly, but expressed as average kWh per day. It’s clear that the battery is worked harder in winter where it is basically fully-charged and fully-discharged once on each winter day.
Click on image for a larger version. Graph showing the amount of electricity stored in and then discharged from the Tesla Powerwall 2 each month since March 2021.
Over a year the battery is used to store and discharge around 3,500 kWh of energy. If we consider the battery cost (about £10,500 in March 2021) and imagine a lifetime of about 10 years, then we can imagine that every unit of electricity drawn from the battery – even apparently “free” electricity from the solar PV system – actually costs about 30 p/kWh in terms of battery “wear and tear”. If we more realistically assume that only 30% of the battery capacity will be lost after 10 years (see below) and that the battery as a whole will remain otherwise viable, then the cost per kWh falls to about 10 p/kWh.
It’s also interesting to note that the amount of energy discharged from the battery is always less than the amount of energy with which it is charged. The ratio of these two quantities is known as the “round trip efficiency”. Inefficiencies arise because the AC electricity in the mains must be converted to DC electricity for storage in batteries, and the reverse process takes place when the battery discharges. Also, the basic physical processes of charging and discharging are never 100% efficient.
The Powerwall specifications suggest that the round trip efficiency is 90%, and if we evaluate this for the 30 months for which we have measurements, we see that in the best cases, 90% is just exceeded. It’s interesting to notice that the round trip efficiency has a seasonal variation – being close to 90% in winter, but falling to about 85% in summer.
Click on image for a larger version. Round Trip Efficiency. Graph showing the ratio of the energy discharged from the Tesla Powerwall 2 to the amount of energy with which it was charged. The data are evaluated each monthly since March 2021. The Round Trip Efficiency appears to be between 85% and 90%.
I think this seasonal variation occurs because – as shown in the figure below – in winter, the battery does not hold the charge for long: almost as soon as the battery has been charged overnight, it is immediately discharged. In contrast, in summer the battery is often fully-charged by midday, and the charge is stored for perhaps 8 hours before the discharge process begins. As I have written previously, this storage actually consumes energy, and I think this is showing up as an apparent loss in round trip efficiency.
Click on image for a larger version. Screen shots from the Tesla App for a typical winter day (left) and a typical summer day (right). In winter the battery charges at night and discharges through the day, running flat late in the evening. In summer the battery charges when the sun shines and the household runs on solar energy most of the day, only drawing energy from the battery at night.
Finally we should look again at the inevitable degradation of battery capacity from which all batteries suffer. I assess battery capacity by recording the amount of energy discharged on days when (a) there is very little or no solar charging of the battery and (b) the battery fully discharges from “full” to “empty”: these conditions occur mainly in December and January.
Click on image for a larger version. Estimates of the practical capacity of Tesla Powerwall 2 battery over the last three winters. The decline in capacity from 21/22 to 22/23 was 3.5%, but the decline in capacity from year from 22/23 to 23/24 has been only 1.9% (so far).
The decline in capacity from 21/22 to 22/23 was 3.5%, but the decline in capacity from 22/23 to 23/24 has been only 1.9% (so far). Extrapolating – and assuming linear trends! – this suggests that battery capacity may fall to under 11 kWh in 2030. If the other parts of the battery system continued to work in 2030, this would still represent a useful battery capacity.
Click on image for a larger version. The same data as in the previous graph but plotted on a larger scale. Linearly extrapolating the last two years of data suggests that battery capacity may be less than 11 kWh in 2030.
Solar PV
Below are several graphs showing solar PV generation through the year. Of particular interest this year is the extra generation from the System#2 panels installed in November 2022. The conventional month-by-month chart shows the expected summer boost to generation. As I remarked back in April, March 2023 was exceptionally dull – with generation down on March 2022 despite the addition of 8 extra panels! December 2023 likewise appears to have been similarly grey. Happily, June 2023 was exceptionally sunny.
Click on image for a larger version. Solar PV generation averaged monthly and expressed as average kWh/day. See text for details.
From a performance point of view, I find it more useful to plot the cumulative generation through the year. The graph below shows cumulative generation throughout each of the last three years. The 2021 and 2022 curves show the typical variability between years – alongside the forecast generation based on the MCS recommended procedure. The 2023 curve shows the boost of 1.8 MWh arising from the extra System#2 panels. If we divide the cost of the system by the anticipated generation (assuming (optimistically) 20 years of flawless performance) the price of the solar electricity comes out at about 8 p/kWh.
Click on image for a larger version. Cumulative solar PV generation throughout each of the last 3 years.
In the graph below I compare this year’s cumulative generation (5.7 MWh) with cumulative household consumption (6.2 MWh) showing that on a whole-year-basis, the solar installation is not quite large enough to produce enough generation to match consumption. If the year had been sunnier, and if since August we had not been charging an EV at home (800 miles at 4 kWh/mile ~ 0.2 MWh) we might just have broken even.
But of course the timing of consumption and generation are quite out of phase. The graph below shows that for the middle portion of the year we consume no grid electricity, and were net exporters from August until late November.
Click on image for a larger version. Cumulative solar PV generation for 2023 compared with cumulative consumption, grid exports and imports.
Summary
Friends, the Solar PV and battery systems are performing as well as I might reasonably have hoped. So my current plan is to just monitor the performance of the system as they and I grow old together.
There are options, but without really shocking the neighbours, I can’t see a way to generate any more electricity locally. I might consider investing in more batteries, but our annual bills are only around £400, and so there is not much more money that we can save. But if prices were low enough, having extra battery capacity might allow us to avoid emptying the batteries in winter, and this would probably prolong their life. Some more entrepreneurial people might begin to try to export electricity for profit, and I wish them well – but that’s not my goal.
Click on image for a larger version. Estimated monthly costs for heating, cooking, and electricity for the last two years.
From April 2023 the Ripple wind farm at Kirk Hill should begin generating and our share of that generation amounts to about 3.5 MWh/year which should cover all our grid consumption with low-carbon generation, and be more in-phase with our consumption than our solar generation. I am not sure quite how I will account for the generation, but it’s good have some problems to look forward to in the New Year.
Protons for Breakfast 
December 30, 2023 at 1:06 pm |
I have been thinking maybe micro-nodes of houses might be useful. Shared battery and solar resource among 4-8 houses to average out usage and take advantage of more surface area for solar and to get more people actively involved in energy considerations.
December 30, 2023 at 1:26 pm |
Bruce, Good Afternoon.
I think that’s a great idea – and I would don’t be surprised if such neighbourhood batteries become ‘a thing’. They would be good for the network too.
M
December 30, 2023 at 11:42 pm |
Indeed, neighbourhood-scale BESS (“battery energy storage systems”) seem quite promising … but only in the longer term, unless you’re in an eco-neighbourhood that is willing (and able) to install an additional set of power lines, or unless you’re part of a pilot-study by your lines company (which is willing and able to “smarten” its local distribution system to the point that it can control the possibly-massive swings-and-roundabouts of energy supply & demand from a ‘hood with mongo rooftop PV *and* mongo BESS).
Another possibility comes to mind… a well-heeled and techo-savvy neighbourhood association might set up a battery-backed DC fastcharger installation (with its own PV and/or wind turbine) in some mutually-accessible location in their ‘hood. Peak efficiency would come if a BEV is fastcharging whenever there’s significant local generation. BESS are still pretty expensive per-kWh even when you’re building at neighbourhood scale; but you can definitely save quite a bit per kWh over domestic-scale BESS, especially in regions where active thermal management is required to keep its batteries “happy”.
IMHO the only net-societal advantage of smaller-scale BESS (in comparison to putting the same $$ into utility-scale BESS) is when small-scale BESS is the most cost-effective way to improve power quality on a local distribution line. Then again, it’s the rare country in this century whose regulators are powerful enough to do *anything* about optimising net-societal benefit, except when it’ll occur as a pleasant side-effect of laissez-faire investment by the folks who have capital to invest. (End of political sermon, I’ll climb off my soapbox now, thanks for reading 😉
December 30, 2023 at 9:48 pm |
Wow, another great set of graphics, thanks for posting!
I’m intrigued that you’re seeing a seasonal variation in the RTE of your PowerWalls that’s in the opposite direction to the one I had expected. My hypothesis (now pretty throughly discredited) had been that the PowerWall’s active thermal management system will kick in on the cold winter days, diverting some high-quality electrical joules into low-quality thermal energy (self-heating) rather than being released for use in your household.
Your current hypothesis is that the RTE is lower in summer months because of the longer dwelltime of its stored energy. Hmmm… I doubt the self-discharge rate of the lithium-ion batteries in your PowerWall is above a couple of percent per month, so I can’t see how this could explain more than a small fraction of the seasonal swing in RTE (which is roughly 5%).
I have a couple of new hypotheses.
1. If it’s warm enough on your porch during summer that — in combination with the heat thrown off by the battery-charging circuitry of the PowerWall — the active temperature management will kick in, probably mostly with a fan or two but perhaps some heatpump action as well. If the fan(s) and occasional heatpump were consuming 100 Watts on average for 6h, that’d be 0.6kWh — and that’s about the size of the seasonal swing in RTE.
2. On a typical winter day, your PowerWall’s inverter may be running pretty-much full-bore for an hour or two, and then enter a low-power “idle” state. On a typical summer day, your PowerWall’s inverter (and associated control systems) may never go “idle”, but may always be consuming some standby power — perhaps 100W?
To test these hypotheses… perhaps you’ll want to get a couple of Tapo T310 thermosensors (maybe £15 apiece) and a Tapo hub (the entry-level H100 will set you back maybe another £20)? If you can open up the PowerWall without voiding its warranty, there maybe someplace near its power-electronics where you could embed one thermosensor? Alternatively just tape a thermosensor somewhere on the PowerWall’s casing that gets noticeably warmer when it has been rapidly discharging its battery. The other thermosensor would somewhere on your porch that’s a bit distant from the PowerWall but would have a similar ambient temperature and exposure to the sun. The hub will store the temp (and humidity) readings at 1 minute intervals, and Tapo’s cloud-services will email a CSV of these measurements to you upon request.
(You might also want to tape a T310 thermosensor to the back of one of the PV panels in each of your differently-facing rooftop sheds. That’ll generate a dataset which *might* be helpful in some future analysis, when you’re trying to estimate how much PV efficiency your panels have lost. The thermal coefficient of their efficiency is a confounder in such analyses if you can’t accurately estimate their temperature during each hour of operation. See https://www.bostonsolar.us/solar-blog-resource-center/blog/how-do-temperature-and-shade-affect-solar-panel-efficiency/)
December 31, 2023 at 1:31 pm |
Clark, Good Afternoon,
Wikipedia say:
If that’s right, then I guess the change in apparent RTE is really a measure of the energy losses in teh battery management system to cope with faster charging and discharging – and environmental management.\
So it may well be an overhead of the battery thermal management – which includes a heat pump. The porch rises to perhaps 30 °C in the summer which is not a bad temperature for a battery, but there may also be some small amount of direct insolation. However, I’m not going to take the Powerwall apart – it’s just too precious!
Anyway, I found the data interesting and surprisingly – given the number of such systems installed – it’s hard to see examples of real world data.
Best wishes for the New Year – which I guess for you has already begun!
M
December 31, 2023 at 11:09 pm
Indeed it’s already 2024 here! I hope you are having a good celebration tonight with other similarly-behind-the-times folks 😉
I’m aware of *no* other careful measurements of the efficiency of a Powerwall 2 installation. The Wikipedia paragraph you cite is relevant only to the Powerwall 1, which is a DC-coupled system (which can run more efficiently when harvesting power from a PV array than the AC-coupled Powerwall 2, because there’s no DC-AC conversion as occurs in your solar inverter).
If you purchase your Powerwall 2 after Feb 2020, then the relevant datasheet will probably be https://www.tesla.com/sites/default/files/pdfs/powerwall/Powerwall2ACDatasheet_EN_UK_feb2020.pdf
The warranted RTE is 90% when it’s brand-new, is at 25 degrees C, and when it is running almost full-bore (3.3kW charge & discharge, i.e. not quite at its max rate of 3.68kW). This efficiency is measured from its AC input to its AC output, as you have done. Your measurements tell me that you’d have no hope of a warranty claim, as your Powerwall 2 is occasionally running above 90% efficiency even though it’s a few years old.
The important part of your dataset is what it reveals about the Powerwall 2 efficiency in an installation where the ambient temperature is not always 25 degrees C, and where the charges and discharges are not always at 3.3kW.
Slower charges and discharges will cause the battery-charging electronics to run less efficiently. Ambient temperatures lower than 25 degrees will cause the batteries to work less efficiently; and if the Powerwall’s fan(s) or heatpump ever kick in, that’ll be a further degradation of efficiency.
I’d summarise your dataset as showing that, in your installation, the Powerwall 2 runs at about 87% efficiency, with a seasonal variation of +3% in winter and -3% in summer. Now that I have read the Powerwall 2 datasheet, and now that I know the ambient temperature on your porch rarely gets above 30 degrees C, I’m pretty confident that the main factor that controls the seasonal variation of Powerwall 2 efficiency in your installation is the rate of discharge.
And … I think it’s a “good thing” that your Powerwall’s efficiency is highest during the winter months — because that’ll be when electricity is most valuable in the UK. Indeed, sooner or later, I’m confident the PV generation on your local distribution line will rise to the point that exported power in mid-days during summer will be pretty much worthless — unless your electricity retailer is cross-subsidising its export payments to you, from its profits on its other charges to you, and from other customers.
That’s pretty much how things are already in the sunnier and more-affluent parts of the world. See https://en.wikipedia.org/wiki/Duck_curve. All to say that wasting 1 kWh of PV energy in mid-summer isn’t anywhere near as “bad” as wasting it during mid-winter.
December 31, 2023 at 11:46 pm
Clark,
What a clear summary. Thank you.
M
December 31, 2023 at 11:31 pm
Your RTE measurements are now cited as ref 45 of the Powerwall article in Wikipedia 😉
December 31, 2023 at 11:45 pm
Clark – Wow!
I have just turned 64 and I feel I just becoming aware of the secret world that lies beneath the internet. Sometime ago I found I had an entry on IMDB – but to have a reference from Wikipedia – the source of sources – is the ultimate honour! What a way to end 2023!
Warm best wishes to you and yours.
M
December 31, 2023 at 5:51 pm |
Michael
Thanks for this, always good.
I have a modular LiFePO4 system and I think my RTE is more like 75% max than 85%. This is based on the ratio over long time periods (eg a quarter) of total energy inverted (ie used by me) to total energy rectified (ie stored in battery) as reported by my inverter.
You say “the Powerwall specifications suggest” 90%. Have you actually measured it?
Cheers, Graham
December 31, 2023 at 7:32 pm |
Graham, Good Evening.
Yes, I have measured the Round Trip Efficiency – it’s in the article you are commenting on. Using monthly averages, I measured the amount of charge delivered to the battery and the amount it discharged.
I suspect that over the long term, lower round trip efficiencies are caused by self-consumption by the battery management system. I wrote about this here.
All the best for 2024
Michael
January 4, 2024 at 2:55 pm |
Excellent article as usual. It would be interesting to add the impact of low cost electricity overnight to the calculations. Domestic battery storage does not payback for itself even with current high energy costs. However, recharging the battery at night while also charging your EV might be a way to avoid draining the battery in winter and extend its useful life, though.
January 9, 2024 at 2:36 pm |
Giovanni, Good Afternoon,
My current energy bills are around £400/year. It’s hard to say exactly what they would have been without various interventions (PV/battery/heat pump) but a few years ago they were about £1,600 pa and now would probably be in the region of £2,500 pa. So that’s a saving of ~£2,000/year.
In winter we do charge the battery at night (and the EV). I did think about different strategies for extending the battery life – such as reserving the last 1 kWh of storage – and I might try this next year. But I bought the battery as an experiment based on the idea that in 10 years time – 7 years now – batteries would be both cheaper and better.
Anyway, best wishes
Michael
January 9, 2024 at 3:33 pm
Hi Michael, thanks for your reply, it will be easy to reverse calculate what would have been the cost of running the heating system on gas (the bad word) and assess the house electricity consumption without the PV panels. I have extensive data on our EV performance over the last 7 years and I thought to share this data with you to write an article on this together? if it is an interesting idea, please let me know, you should have my email address. We also developed a battery management simulation software I would like to show you! 🙂
January 12, 2024 at 12:57 am |
Nice graphics. I especially liked the comparison of multi-year data to see how stable are the values you get. Congratulations on a good job.
January 12, 2024 at 9:34 am |
Shawn, Thank you for your kind words. All the best for your own adventures. M
February 13, 2024 at 2:37 pm |
Interesting and well written, thanks.
I do think it’s worth investigating exporting electricity for profit, it has both green benefits and significant financial benefits as well. I generate far more than I can consume in summer and after moving to an export tariff last August, received £330 in export payments to the end of 2023. I’m hoping that the summer export payments I receive this year will balance out the winter costs of running the ASHP, so ending on a zero annual bill. We shall see.
You mentioned getting extra battery storage as a potential option, all I can say is do the maths on it. I’ve modelled getting extra batteries for my system and depending on which scenario (e.g. overnight cheap charge to use during the day, capture solar for peak rate export, etc) I get payback periods of 5-22 years! Batteries are still too expensive at the moment