Methodology

Energy cost

Annual energy is mileage divided by 100, times consumption. The price per kWh is a weighted blend across the charging channels you set, so rapid charging is priced as rapid charging rather than averaged away.

Why charging losses are not added

The official WLTP consumption figure for an electric car is measured as the energy drawn from the mains to recharge the battery (UN GTR 15, applied in the EU by Regulation 2017/1151), so it already includes AC charging losses. Adding a loss on top would count them twice, which is why no charging loss is added.

Real-world consumption

Both drivetrains use more than their test figure in practice, but not by the same amount, and the evidence for each is different. Petrol and diesel consumption is raised 20%: the European Commission's analysis of the fuel meters fitted to cars on the road found real use around 20% above WLTP, and the ICCT puts it at 18 to 19%. Electric consumption is raised 15%: there is no fleet-wide meter data yet, and published comparisons range from about 11% for the car alone to more once charging losses at the wall are counted. Until September 2026 both sides used 15%, which understated what the combustion car costs to run.

Break-even mileage

Total cost is linear in annual mileage, so rather than searching for the crossover we solve for it directly. The result is the annual mileage at which both cars cost the same over your holding period. It does not depend on the mileage you entered, which is deliberate: the question is how far you would have to drive.

Emissions

Electric emissions use the carbon intensity of the national grid, taken from Ember for the most recent year available. It is an annual average: a real grid is much cleaner on a windy night than at the evening peak, so this cannot capture when you charge. Ember's figure is a life-cycle one: it counts fuel supply chains and building power stations as well as burning fuel in them. So the combustion side is counted the same way, well to wheel: 3,008 g CO₂eq per litre of petrol and 3,384 g for diesel, from the EU's fossil fuel comparator of 94 g per MJ (Renewable Energy Directive, Annex V) and the energy content of each fuel (Annex III). Until September 2026 this used the tailpipe figures of 2,310 g and 2,680 g, which left the fuel's extraction and refining out of one side only. Neither side counts building the car or its battery, so this compares running emissions, not whole lives.

What the car totals leave out

Servicing is a figure you enter, but insurance and road tax are excluded on both sides. Insurance is usually higher for an electric car and road tax lower, so leaving both out is not neutral. The totals are therefore running and depreciation costs rather than the full cost of keeping a car.

Consumption bands

Eurostat publishes household electricity prices by annual consumption band. Charging a car usually moves a household from the 2,500 to 5,000 kWh band into the 5,000 to 15,000 one. The two bands are priced differently, so the EV calculator uses the larger band and the solar calculator the smaller one.

Production

Yield per kWp comes from PVGIS, which accounts for irradiation, temperature and a 14% system loss. If you enter your roof pitch and direction, PVGIS is queried for that exact orientation rather than approximating the loss from optimal with a formula. Where a stored figure is used instead, the page says so.

PVGIS is asked for panels mounted on a building rather than free standing. Free-standing assumes air flowing freely behind the panels, which keeps them cooler and gives 4 to 5% more. Panels on a roof rack run much closer to the building case (in Sandia's temperature model a close roof mount is nearer an insulated back than an open rack), so this is the cautious choice for a rooftop.

Output falls 0.5% a year, the median of nearly 2,000 measured degradation rates (Jordan and Kurtz, NREL, 2013).

Self-consumption Estimate

This is the weakest link in any solar calculator, including this one. Knowing how much of your production you use rather than export needs hour-by-hour metering that no public dataset provides. The quick estimate, the default, works through a year day by day. A month is not thirty identical days: a battery sized for a sunny day's surplus sits half empty through a dull week, and a big one carries a bright day's surplus into the dull day after. So each month is sixty days drawn in order from the spread of dull and bright days for a month that clear, and the battery's charge is carried from each day to the next.

Where the days come from. How clear a month is comes from its own output, compared with what a perfectly clear sky would deliver at that latitude and panel angle. The spread of days for a month that clear comes from the distribution published by Bendt, Collares-Pereira and Rabl (Solar Energy, 1981), corrected to PVGIS's own daily output for 16 EU cities, which is narrower in bright months and wider in dark, snowy ones. And one day resembles the last as much as PVGIS's weather does: a correlation of 0.28 from one day to the next.

Each day. 43% of the overlap of production and demand is used as it is generated, since panels peak at midday while households use most in the morning and evening. A battery charges from the rest, after its losses, up to the room it has, and serves up to 80% of the demand the panels missed, since some of that is a peak beyond its inverter or comes before it has charged.

These figures are fitted, not guessed. HTW Berlin publishes self-sufficiency figures simulated with 74 measured household load profiles at one-second resolution (the Unabhängigkeitsrechner, by Weniger, Quaschning and colleagues). The 43% is fitted to its figures without a battery; the 80% to the hour-by-hour model in eight cities and to HTW together, then checked on eight other cities it had not seen, with batteries up to 5 kWh per MWh of use. There it lands within 5 points of the hour-by-hour model everywhere and 2.4 on average, and within 2.6 points of HTW. A battery's benefit keeps rising with its size as HTW's does: for 5 kWp and 3,500 kWh a year in Berlin, 74% of demand covered with 10 kWh and 77% with 30, against HTW's 75% and 78%.

Mistakes found and fixed in September 2026. The factor used to jump to 0.85 as soon as any battery was added, so a 2 kWh battery looked as good as a 10 kWh one and was credited with nearly three times what it could deliver cycling fully every day. The battery's losses were counted as exported rather than lost. Every day of a month was treated as its average day, so a large battery looked as if it filled every gap. And then, once days were told apart, the battery was emptied every day, so nothing bigger than a day's need added anything. Each is now covered by a test that checks the answer against physical limits and published figures, not only against the code.

How a battery behaves Estimate

Efficiency. 90% of what goes in comes back out, the loss split between charging and discharging. HTW Berlin's Stromspeicher-Inspektion 2026 measured battery efficiencies from about 88% to 97%, before inverter losses, so this is the efficient end.

Standby. A battery system draws power around the clock just to stay ready: 4 W for the best systems in HTW's 2026 test, 64 W for the worst. The calculator assumes 10 W, about 88 kWh a year, met first from any surplus solar, then from the battery's own charge, and bought from the grid when neither has any. For a small battery that is a large share of what it delivers.

Ageing. Home batteries lose about two to three percentage points of usable capacity a year (Figgener and colleagues, Nature Energy, 2024, eight years of field data from 21 systems). The calculator takes 2.5 points a year, so a battery replaced after 10 years ends at 77.5% of its capacity. Every year of the cashflow uses the capacity the battery has that year; a replacement starts again at full.

Power and charging. It charges only from surplus solar, never from the grid. Working hour by hour, it also charges and discharges at most half its capacity per hour (5 kW for 10 kWh); the quick estimate works in whole days, so that limit does not arise there.

Monthly consumption profile

In the quick estimate, household consumption is spread evenly across the year. Real households use more in winter, so a flat profile slightly overstates summer self-consumption and understates winter imports. The hour-by-hour model uses the seasonal shape of the profile above instead.

Export arrangements

Under net metering, exports are credited at the retail price up to your annual consumption, with any surplus beyond that paid at the export rate. Under a feed-in tariff, only the electricity you use on site avoids the retail price and everything exported earns the export rate. With no compensation, exported electricity is worth nothing.

Monthly netting sits between the last two, and is the arrangement many households actually have. Nothing is paid for the surplus and nothing is charged for taking it back, but the balance only holds within a calendar month: electricity you export can offset what you import in that same month, and whatever you have not drawn back by the end of it expires. So it is worked out month by month, comparing what each month exported against what that month still had to buy, and taking the smaller of the two.

The monthly reset matters more than it sounds. Under annual net metering a July surplus can pay for a January bill; here it cannot, so a roof that produces far more in summer than the house can use loses that difference.

One consequence is worth stating because it is the opposite of what you might expect: under monthly netting a battery adds almost nothing. The grid is already holding your surplus for a month at no charge and handing it back free, which is exactly the job a battery does. For a 5 kWp system and a household using 3,500 kWh in Berlin, monthly netting with no battery covers 88% of the year's use (84% worked out hour by hour); with no compensation, even a 30 kWh battery given away free reaches only 77% (76%). Where there is no compensation, a battery is worth a great deal; under monthly netting it is worth almost nothing.

Money over time

The cashflow runs 25 years, with output falling 0.5% a year.

Two of the figures reported measure the same system in different ways, and the difference matters enough to spell out. The net gain over 25 years simply adds the money up: a euro saved in 2051 counts the same as a euro spent today. The value in today's money, which finance calls net present value, does not. Money you get back later is worth less than money in your hand now, because money in hand could be earning something meanwhile and because prices rise, so each future year is reduced before being added. How much it is reduced by starts at 3% a year and is adjustable. Despite the usual name for that rate, a discount rate, none of this has anything to do with a discount in the retail sense; nothing is being marked down.

Because of that reduction, the two can point at different system sizes: the plain total favours a larger array that returns more in the end, present value favours a smaller one that ties up less money now. Neither is wrong, and the solar calculator says so when they diverge. Payback is likewise given both ways, and it is the point after which the running total stays positive: a battery replacement can pull a system that has already paid for itself back into the red, and the calculator says when it does. Alongside them sit the internal rate of return and the levelised cost of the electricity you actually get the benefit of, which is the figure worth comparing against your grid price.

Rising electricity prices Estimate

The solar model assumes the retail electricity price rises 2% a year by default, applied to each year after the first, so 24 increases across the horizon. This is one of the most influential assumptions on the page, because the saving is the price of the electricity you no longer buy: a higher rate shortens the payback and pushes up the net present value. Nobody can forecast 25 years of European electricity prices, so the rate is adjustable and setting it to zero shows the result at today's price held flat.

The electric car model deliberately does the opposite and holds both electricity and fuel at today's prices. There, the answer depends on the gap between two escalation rates rather than on one rate, and guessing which of the two rises faster would dominate the comparison. A snapshot at known prices is the more honest output.

Upkeep and replacements Estimate

Solar running costs are taken as 1% of the total installed cost a year, battery included, covering cleaning, inspection and the odd repair. On a 5 kWp system with no battery that is about €60 a year, deducted from every year of the cashflow and included in the levelised cost.

Batteries are assumed to last 10 years by default, which is roughly where manufacturer warranties sit, and are bought again every time they reach that age. On the default a 25-year analysis therefore buys one on day one and replaces it twice, in years 10 and 20. Change the lifetime in the calculator and the number of replacements changes with it. This is usually what decides whether a battery pays for itself.

A battery whose life ends exactly as the analysis does is not replaced, since nobody would buy one on the day we stop counting. One bought near the end is charged in full, with no credit for the years of life left in it, which makes the figure cautious rather than flattering.

No inverter replacement is charged, even though inverters commonly need replacing once in 25 years, so solar payback figures remain at the optimistic end of the range.

Why bigger is not better

Each extra panel produces electricity at the same cost, but a household can only absorb so much of it at the moment it is generated. Once production covers the demand it overlaps with, the remainder has to be exported for less than it saves you, or under no export arrangement for nothing at all. The value of an extra panel therefore falls while its cost does not, and past some size another panel subtracts value.

So there is a best size, and it depends on the household rather than the roof. The solar calculator works it out and says so when your chosen size is well away from it; the reasoning is set out beside the charts there. It does not size the battery for you.

What a car is worth at the end

Depreciation, not the purchase price, is what the car comparison charges: the price paid less what the car is still worth when you sell it. It is usually the largest single number in the comparison, which makes the resale assumption more consequential than the fuel prices. The defaults are 40% of the price retained for the electric car and 45% for petrol or diesel after five years, both adjustable, and both genuinely uncertain given how quickly used electric car prices have moved.

Currency

Everything is shown in euro, including for the six member states that use their own currency, because the source figures are published in euro. Converting would introduce an exchange rate we would then have to source and refresh.

Population and coordinates

From GeoNames, which offers its data without warranty of accuracy or timeliness. Some population figures are visibly out of date, but population only decides which cities are listed. Coordinates do feed the figures: every irradiation lookup is made at the coordinates held for the city, so a misplaced city would give solar numbers for the wrong spot.