Power and the Glory

Part 3 — Sunshine, more or less

Let’s start with some numbers: on a normal day with no clouds, the surface of the earth is hit by roughly 1000W of power per square metre (W/m2) depending on latitude and day of the year.

Where we are — Scandinavia — the maximum is 995W/m2 in the middle of June. Wouldn’t it be neat if we could, y’know, siphon off a little of that?

Enter the world of photovoltaic (PV), or more commonly: solar power.

It’s a rum business, this. The first demonstration of it was 1839 (yes, 18, not 19) by a Frenchman from a family quite possibly more known for a different discovery: Alexandre–Edmond Becquerel1

By way of such relative unknowns as Heinrich Hertz and Albert Einstein, the first practical cell was shown in 1954. ’nuff wikipedia stuff2

Some Terms

A solar cell is an electronic device which convert sunlight into electricity by way of light–sensitive semi–conducting photodiodes.3

A solar panel is a collection of solar cells. Panels are in turn connected to each other, and to an inverter.

Solar cells, more often than not based on single–junction crystalline silicon (c–Si)4, produce direct current (DC). In other to use this in a house, we need to invert it into alternating current (AC).5

The inverter — typically a large box with loads of electronics and heaps of capacitors — feed AC to the house via the regular fuse–box. It may also be connected to a battery on either the DC or AC side.

Simple enough.


Or so we thought.

As mentioned above, at our latitude we’ll have a maximum of 995W/m2 at the best of times. The practical output from that depends on a number of factors:

… and, of course, the construction and size of the panel, the size of the roof or other location to put’em, and — to our surprise — the ambient temperature. The lower the better!

Let’s start slow: we have, on our property, two buildings — the house and an adjacent garage. Some measurements:

The average, ambient temperature is around 8.1°C (2021 — bad for the environment, actually, but good for us). We have little to no shading on any of the roofs.

Photovoltaic Panel Technology

Not too fast now: In very rough terms, there are three6 different types of PV panels on the market today, listed by amount of usage: polycristalline, monocrystalline, and thin–film.

In the first two the cells (called “wafers”) are based on c–Si — see above — and are either poly–Si (composed of a number of smaller crystals) or mono–Si (composed of one single crystal). The latter is a–Si (non–crystalline silicon).

mono–Si has the highest efficiency of these — laboratory experiments has reached 26.7% (2021) or, in less technical terms, almost 27% of the sun that hit the cells are converted to actual electricity. In the real world the numbers float around 20%7

We can further divide panels into two types: glass–polymer (GP) and glass–glass (GG). In both cases the overall design is the same — an aluminium frame into which is set, first, a front layer of glass, then the silicon wafers, followed by a back layer and cabling.

GG refer to the panel where the front and the back is glass; GP has only a front panel and use a flexible polymer (plastic) sheet for the back.

The most popular solution has been GP; it’s also the cheapest. GG weigh more, but can withstand more stress. GP has a normal manufacturing warranty of between 10 and 20 years; GG is usually guaranteed for 30–35 years.

And, finally, the glass–glass panels can be either mono– or bi–facial. The latter produce energy from both sides of the panel, taking advantage of the “albedo” (aka reflected light).

Each panel is rated in Wp or “watt peak”, the potential amount of effect it can produce. A “400Wp” panel can theoretically yield 400W; a bi–facial version mounted at an angle on a white surface could produce some 500W due to reflections.



I hate the stuff8 … fine, fine. Let’s do it. Given:

One single glass–glass mono–Si bi–facial panel produce a theoretical maximum of 400W. An inverter typically sport two “strings” — or inputs for panels — on which you can connect ca. 20 panels.

This results in 8,000W (8kW) per string theoretical power output – then you need to consider such things as sun hours, latitude, array tilt, day of the year, and other reasons which will reduce the actual effect.

One of the reasons is also one of the more complex issues.


First, let’s dispense with a myth: “solar panels only produce energy in direct sunlight”.

Yeah, no. Not so much. A solar panel is active as long as it is hit by photons. It might not be enough of’em to be much use, but there will be generation of electricity as long as there is light.

That said, shading and PV systems is a nasty chapter.

First, since solar cells rarely operate on their own, they are connected in series on each panel; say about 60 per. Each panel has a number of sections (usually 3); the panels are connected (again, in series) into strings, the string to the inverter. How many panels per string depends on your inverter; how many strings … you get the idea.

Secondly, there’s this idea floating around that if you get a little bit of shade on a panel it’ll stop working and even take out the entire string! Shock Awe! Bypass diodes!

As my self–drawn little picture show, each section of wafers are connected in series but via a bypass diode to the other sections. The diode ensure that if shading occur, only the section in question cases to work. In the picture the panel will deliver 66% of its theoretical maximum with the shading indicated.

You’d be hard pressed to find modern–day solar panels without these diodes.

That said, there will be losses from shading. Enter DC power optimisers — basically small electronic components connected to each panel in order to adjust the voltage by way of maximum power point tracking (MPPT). This means your panels have optimal performance, regardless of shade or damage.

It also means that in a 50–panel setup you have 50 power optimisers layin’ about on the roof. Points of failure, anyone? Even so it is a very popular solution, pioneered by the company SolarEdge. Actually, it’s more or less only SolarEdge who maketh these which means you’ll also have to get a SolarEdge inverter. Make of it what you will.

An alternative is that offered by — more often than not — Enphase: micro–inverters. In this scenario you’ll not have 50 panels feeding one inverter with DC, but 50 inverters feeding the house with AC, one per panel. See notes above on the topic of points–of–failure.

And, finally, there’s Fronius — an Austrian company — who instead of using hardware for the MPPT has settled on software. Their products are all string inverters, with about 20 panels/string (and two strings per inverter), and are fitted with the Dynamic Peak Manager. Basically voodoo, the software will adjust the MPP continuously in such a way that the detrimental effects of shading is eliminated to the same extent as with optimisers.

I’ve included a link to a Danish study by Assc. Prof. Dr. W.–Toke Franke which seem to confirm this theory: The Impact of Optimizers for PV-Modules; a number of fairly instructive – and funny — videos on the topic can be found on youtube. That particular battle is ongoing.

The software approach would appear to gain ground, with several manufacturers such as SMA going down that route.

Some Observations on Fuses

First a touch of ye olde elementary school or so: the ampere (A) — so named for Frenchman André–Marie Ampère — is the SI unit for electric current. This is what us normal people know to use for fuses.

The watt (W) describe the rate of energy transfer, somewhat simplified. It’s named from James Watt, obviously.

The volt (V) describe electric potential … let’s not go there. Named for Alessandro Volta.

Now: A regular house in Sweden has a 20A fuse, and run on a 3–phase (3φ) 230V system. The formulae P = U×I×φ gives us the watts per phase.

In our case: 230V (electric potential — volts) times 20A (fuse) times 3 = 13,800W

This is the maximum load we can pull from the grid before the fuse goes kaboom9 — either way. We could, theoretically, pour 13.8kW of power back into the grid. More than that, and … kaboom. There’s some leeway, luckily.

All of this is important for dimensioning. First, you want as much solar power as you can possibly get — for obvious reasons. However, you don’t want too much since …

Briefly: when you dimension the system, you want more panels than inverter effect to cover up for shady days, but not too many so that you blow the thing in winter. You want as much inverter effect as you can get, but only up to P = U×I×φ so you don’t do in the fuse.

We’ve put up a total of 17kWp panels, feeding two inverters which can manage 15kWp into a house which has a maximum of 13.8kW. So far so good.


Onwards. If you install a PV system which can produce, say, 15,000 kWh per year you ain’t gonna use all of it.

Unless you are on the equator, the amount of sun hours during a year will not only dictate the total kWh possible, but also the self consumption of same. Let’s illustrate.

In the graph you’ll find the average hours of sun per month (blue), the consumption per month (red), as well as the predicted production of solar energy (green).

January, February, November and December are out. The consumption will, due to the few hours of sunlight, outpace production. In March and October they balance (roughly).

April through September we’ll produce far more than we can use with this setup.

And then, of course, the sun goes down occasionally, even here in Scandinavia. Let’s take an exceptional month and go with it: June.

We’ll have a surplus of 1,048 kWh per month. Neat! Except, of course, that we can’t use’em. They will, in practical terms, be wasted as heat.

This is were grid feed–in comes to play, whereupon you become a producer of energy, and sell the surplus back to the electricity company (The Grid) who will pay you a certain amount per kWh10

The feed–in tariff reduces your energy bill; in very real terms you are using the power distribution net as a bloody big battery. You put power in during the day, and pull power out during the night.

And in June a night is six hours long. Of course, there’s that pesky difference between what you pay for a kWh, and what they pay (you) for a kWh.11

Note, carefully, that the grid feed–in — even the existence of such – depends heavily on your country, your laws and the regulations of your power company. Example: in our location, the feed–in is generous but is limited to how much you consume. So: if we pull 3kW from the grid, we may put 3kW back and get paid for it.

Enter solar batteries. In principle large–scale versions of the AAA cells you use around the house :), these — often based on lithium iron phosphate (LiFePO4) chemistry for safety, can hold 10 — 20 kWh of power. The best ones are modular, so that you can start off with 2–3kWh and add more later. The battery connect to the inverter, either on the DC or AC side, and store surplus energy from day to night. This is one method of increasing your self–consumption.

Not all inverters are equal, mind. You need a specialised unit (hybrid) in order to handle batteries or pick a system design that split out the battery management from the inverter — SMA or Victron are worth a look.


Batteries like these are expensive. Some countries, like Sweden, have green deals — you can deduct up to 50% from your taxes for the cost of batteries, hybrid inverters and installation cost.

They are heavy, a concrete floor is a given. I’d not recommend these in an apartment, or upstairs in a house.

Some are connected to the AC side of things. They also come with a power–loss of 5–10% at absolutely no extra cost … ’cause you have to do a DC ↠ AC (panels to AC), a AC ↠ DC (AC to battery) and a DC ↠ (battery to AC) triple–conversion. DC–connected batteries has only one conversion: DC ↠ AC.

They will only help “move” power from day to night; not from week to week or, [afterlife of your choice] forbid season to season. To “store” solar power from summer to winter, grid feed–in is the only realistic option for the vast majority of us.12

Coup de grâce: under normal circumstances your battery will not work as a gigantic UPS that can power your house during a blackout. To achieve that, you need to work a little bit more in order to achieve an “off–grid” system (“ö–drift” for the swedishly inclined). All of what I’ve described above are “on–grid” designs.13


Remember, in part 1, that I expressed a desire to “shop locally”? There are various reasons for this — one a wish to have a reasonably short time to travel in order to kick gluteus maximus if things go wrong, another the somewhat more rational thought that I do not want to buy a couple of metric tonnes of glass, aluminium and plastic and so forth just to have it shipped around the world at N tonnes of bunker fuel per hour; yet another that I require guarantees of proper environmental considerations in factories, proper treatment of employees, payment of taxes … in short rational, modern thinking from a manufacturer.

The number of PV system manufacturers in Sweden is … low. Europe is a different matter. So onwards and upwards: we begun limiting our choices:

Functional requirements:

Non–functional requirements

Out goes the tenders. This was, for a while, more complicated than one would have hoped as the installers are all in–sane–e–ly busy. We finally managed four bids.

Prior to evaluating these, we played a little with a design tool made available as a web app by SMA Solar Technology AG: “Sunny Design”. With it we could estimate the effect of various panels. Fronius has a similar software — the Solar Configuration tool. Both free.

The result was a reasonable expectation of what we could expect. It’s a bit complicated ’cause of the two roofs, but, in the end, we landed on:

Expected returns are notoriously difficult. The included calculation shows the peak — the Leibniz–effect15 We don’t quite believe it, of course. Our installer landed on a rather more likely, if conservative, number of 14,000kWh per year.

With a planned expansion of the battery storage to 22.1kWh we ought land at a required grid–usage of 2,300kWh per annum. Numbers have, and will continue to, change over the course of the projects.

And, yes, before you ask: we need two inverters. One, the GEN24, is there to handle the 29 panels on the house as well as manage the battery stack. The other is there to handle the 18 panels on the garage; it’ll primarily handle in–house consumers and grid feed–in. They’ll communicate with each other, Home Assistant, and, in the future, a BEV charger.

A final “thank you” to the most excellent folks over att Energy Effective Solutions AB who won the bid, organised the work, did the job, and delivered on time, on budget and on quality. Damned fine show.

The installation was inspected by JB EcoTech AB.


By Midsummer’s Eve, the ’cells have been running for 48 hours. Don’t ask about the delay. It’s a sorry story of so–called locked switches, plastic signage and the local power co.

However: on the 24th of June, the system produced 118 kWh which, extrapolated, would yield 3,500 kWh per month for — at a minimum – June, July and August. That’s 10,620 kWh or over half of what we estimate using per year.

Chew on that one for a while …

It seems more and more likely that we will, during March through September


The overly observant might have noticed that the house sports a 20A fuse. This, by way of simple maths (!), tells us that we can pull a maximum of 13,800W (13.8kW) from the grid. It also means we can push 13.8 to the grid.

Keep in mind that the installed total maximum of the panels is 17kW. The maximum of the inverters is 15kW.

On the 28th of June at noon the PV system produced 15.1kW. We are rather in awe and more than a little glad for the built–in margin of error.

Some useful links

1 Yes, that’s the dad of Henri, who with Marie and Pierre invented the glow–in–the–dark field of fashion …

2 Juicy tidbit: Albert got a Nobel in physics. “Ah! Relativity!” I hear you cry. Alas, no. He got it for, hang on now, the photoelectric effect.

3 Which has exactly the consequences for your project which you suspect it has in a world with a shortage of semi–conductors.

4 If you have a desire for more, then look up the Shockley–Queisser limit, multi–junction cells, and perovskite. Be warned that this rabbit hole goes deep. I can, however, warmly recommend https://www.pveducation.org

5 For a fun read, check out War of the Currents. Us humans are weird.

6 Three is a conservative estimate. As of 2022 the work on new cell technologies is proceeding very rapidly. Multi–junction and perovskite, as mentioned.

7 As of June 2022, reports surface on perovskite cells with 30% efficiency. This may sound too good to be true …

8 “I recognise my failing and will be sure to correct it.”

9 “Where’s the kaboom?⁉ There’s supposed to be an Earth–shattering kaboom‼”

10 An amount less than what you pay them for the same kWh, naturally. Ave! Duci Novo, similis duci seneca!

11 Around Stockholm, the price per kWh is determined by the Nord Pool power market, specifically zone S3. See also https://www.nordpoolgroup.com/en/Market-data1/#/nordic/chart

12 For an amusing read, look up “pumped solar” (no, it’s not as dirty as your mind just claimed! Shame on you!) or “Snowy 2.0”

13 To complicate stuff: the hybrid inverter called Fronius GEN24 has a concept they call a “power point”; a single 1–phase plug that, during a blackout, can power a 220V device.

14 “Wait! BYD isn’t very local!” — a very astute observation. The BYD battery systems are produced and shipped from China, and could be considered an excellent example of my hypocrisy. However … the HVM system is also the one recommended by Fronius as the best possible fit with their inverters. Life’s a compromise, yeah?

15 As misquoted by Rifleman Harris: “All is for the best in the best of all possible worlds”. What? Doesn’t everyone learn philosophy from the merry men of Richard Sharpe? Huh.

2009 — 2013 archive (aka "ye olde stuff")