How Many Solar Panels Do You Actually Need? Three Methods Compared

Type how many solar panels do I need UK into a search bar and you’ll get three answers dressed up as one. An installer’s calculator says ten. A roof-layout tool says fourteen. Your own smart meter data, if you actually interrogate it, says eight. All three are arithmetically defensible. Only one of them is a sensible place to start.

This post runs all three methods against a single real dataset, shows the working, and then explains why the gap between the smallest and largest answer is the most useful thing on the page.

The house and the data

Semi-detached 1930s house near Bristol. Gas boiler, no EV, two adults, one of whom works from home three days a week. Twelve months of half-hourly electricity readings pulled from the Octopus Energy REST API (the /v1/electricity-meter-points/{mpan}/meters/{serial}/consumption/ endpoint, page_size=25000, period 1 September 2025 to 31 August 2026). If you’re not with Octopus, n3rgy Data gives you the same half-hourly feed free through the DCC, and Hildebrand’s Glowmarkt API is the other reliable route.

17,520 intervals. Total consumption: 3,812 kWh. Before touching a single panel calculation, here’s what the profile looks like once you group it:

Thirty-one percent of this household’s electricity is used in the seven hours when a UK roof does most of its generating. That single number does more work than any rule of thumb, and we’ll come back to it.

Roof: rear slope, 8.6 m along the eaves, 4.4 m up the rafter, 35° pitch, azimuth 160° (roughly SSE). PVGIS-SARAH3 with 14% system losses gives 930 kWh per kWp per year at that orientation. The MCS Solar Energy Calculator, using its postcode zone tables, lands within 4% of that, which is close enough that arguing about it is a waste of an evening.

Method 1: the rule of thumb

The rules you’ll find on UK installer sites are mostly variants of three:

• A three-bedroom house needs 3.5 kWp, “about ten panels”
• Divide annual consumption by 1,000 to get kWp: 3,812 kWh → 3.8 kWp
• One kWp per occupant plus one: 3.0 kWp

Run those and you get a spread of 3.0 to 4.4 kWp. At 440 W panels, that’s 7 to 10 panels. Most calculators will nudge you to the top of that range, so call the rule-of-thumb answer 10 panels, 4.4 kWp, 4,092 kWh/year.

Here’s the problem nobody mentions. The “ten panels for a three-bed” rule was written when a domestic panel was 250 W. Ten panels meant 2.5 kWp. Panel wattage has risen by roughly 75% since then and the panel count in the rule never moved, so the same sentence now quietly recommends a system three-quarters larger than it did in 2015. The rule didn’t get updated; the hardware outgrew it. Any heuristic expressed in panels rather than kWp has this bug baked in.

Rules of thumb also have no idea what time you use electricity. Two households on 3,812 kWh, one of them a shift worker who’s out all day and one of them working from home, get the identical recommendation. That’s not a rounding error, it’s the whole question.

Method 2: fill the roof

Now measure instead of guessing. Usable roof after 300 mm setbacks at both verges, the eaves and the ridge: 8.0 m × 3.8 m.

A 440 W panel (Aiko Neostar, Longi Hi-MO 6 and most of the current crop) is 1722 × 1134 mm. Two orientations to test:

Portrait:  rows = floor(3.80 / 1.722) = 2
           cols = floor(8.00 / 1.134) = 7
           total = 14 panels  →  6.16 kWp

Landscape: rows = floor(3.80 / 1.134) = 3
           cols = floor(8.00 / 1.722) = 4
           total = 12 panels  →  5.28 kWp

Portrait wins by two panels, which is not obvious until you do it, and which Easy-PV, OpenSolar and SolarEdge Designer will all work out for you in about ninety seconds if you’d rather trace the roof than measure it. Roof-fill answer: 14 panels, 6.16 kWp, 5,729 kWh/year.

One constraint bites here. A system whose inverter exceeds 3.68 kW per phase can’t go in under a G98 notification; it needs a G99 application to your DNO before commissioning, which takes weeks and can come back with conditions or an export limit. At 6.16 kWp you’re running a 5 kW inverter and you’re firmly in G99 territory. That’s not a reason not to do it, but it is a reason the roof-fill number needs a phone call attached to it.

Method 3: match the consumption data

This is the one that requires actual work, and it’s the one that gives you a defensible starting number.

Take the half-hourly consumption series. Pull the matching hourly generation profile from PVGIS (the hourly radiation tool exports 8,760 rows of P in watts for your exact lat/long, tilt and azimuth). Resample one to match the other, scale the PV series linearly by system size, then compute min(load, pv) for every interval. Sum it. That’s self-consumption. Everything else is export.

Twenty lines of pandas, and a decent prompt to Claude or ChatGPT with both CSVs attached will write it correctly on the first try if you tell it explicitly that the consumption column is kWh-per-interval and the PVGIS column is instantaneous watts. That unit mismatch is the single most common way this analysis goes wrong: people sum watts as if they were kWh and end up with a system that appears to generate 40,000 kWh a year.

Output for this house:

panels   kWp   gen_kWh  self_kWh  self_%  export_kWh  Δself/panel
   6    2.64     2455     1570     64%        885         262
   8    3.52     3274     1810     55%       1464         120
   9    3.96     3683     1910     52%       1773         100
  10    4.40     4092     2000     49%       2092          90
  12    5.28     4910     2150     44%       2760          75
  14    6.16     5729     2270     40%       3459          60

Read the right-hand column. The first six panels each deliver about 262 kWh of self-consumed electricity a year. Panels seven and eight deliver 120 each. By panel fourteen you’re down to 60, a fifth of what the first panels gave you, because the roof is producing surplus at midday in June that this household has no way to absorb.

The knee is at eight panels. Past that, each additional panel is a bet on the export price rather than a saving on the import price. Consumption-matched answer: 8 panels, 3.52 kWp.

If you want the full treatment of pulling, cleaning and aligning these two series, including the daylight-saving trap that silently shifts your PV curve by an hour for seven months of the year, that’s covered in Sizing From Your Own Consumption Data.

Three answers, one roof

Now price them. Import at 26.4p. Export at two rates, because this is where the whole decision lives: Octopus Outgoing Fixed at 15p, and a legacy SEG deal at 4.1p.

Eight panels (quoted £5,400 installed): self-consumption saves £478. Export earns £220 at 15p, or £60 at 4.1p. Annual benefit £698 or £538.

Fourteen panels (quoted £7,900): self-consumption saves £599. Export earns £519 at 15p, or £142 at 4.1p. Annual benefit £1,118 or £741.

The marginal six panels cost £2,500 and return £420 a year on a 15p export tariff. That’s a 5.9 year payback on the extra hardware, better than most of what else you could do with £2,500. On 4.1p they return £203 a year, a 12.3 year payback, and now you’re weighing it against the inverter’s own warranty life.

That asymmetry is the argument. Eight panels is the number that’s right regardless of export tariff, regardless of whether Outgoing Fixed still exists in 2031, regardless of whether you move house. It’s the portion of the system that pays for itself out of electricity you were going to buy anyway. Panels nine through fourteen are a separate investment with a separate risk profile, priced entirely by a tariff you don’t control.

So: start at the matched number, stop at the roof-fill number, and treat everything between as a tariff bet you’re consciously placing. Anyone who hands you a single figure has collapsed that decision on your behalf.

What a battery does to the slider

Add a 5.2 kWh battery (GivEnergy, Fox ESS, whatever your installer stocks) and re-run the same simulation with a simple charge-discharge rule. Self-consumption at eight panels rises from 1,810 to about 2,480 kWh. At fourteen panels it rises to roughly 3,050 kWh, and now the big system’s surplus has somewhere to go.

The battery is what makes roof-fill sensible on a poor export rate, because it converts midday surplus from 4.1p export into 26.4p avoided import. It also moves the bottleneck: with a battery in the system, the marginal panel is worth having for longer, and the knee in that Δself/panel column shifts right by three or four panels.

Worth knowing before you get excited: on Intelligent Octopus Go at 7p overnight, a battery earns most of its money from tariff arbitrage, not from solar, particularly between November and February when a 6 kWp array in Bristol produces under 200 kWh a month. Model the battery against the tariff and the panels separately or you’ll credit the solar with savings the cheap-rate import actually delivered.

Run it on yours

Pull twelve months of half-hourly data, group it into the five windows in the first table, and look at your 09:00–16:00 share. If it’s above 35%, the matched number will come out close to your roof-fill number and the decision is easy. If it’s under 25%, no panel count fixes that, and the honest next question isn’t how many panels you need but whether a battery should come first.