Weather. Five years of measured hourly radiation (global and diffuse) and temperature at your point, composed into a typical year the way a TMY file is: for each calendar month, the real days of whichever year sat closest to the five-year average. Day-to-day variety survives; the total stops depending on whether the last twelve months were a freak. That year is then laid over the twelve months ahead, starting today, so the dates and weekdays you see are the ones you are heading into rather than the ones the measurements came from – which also gives the household model a real, unbroken run of weekends. The clock offset is worked out for every day, daylight saving included.

Generation. The archive supplies the diffuse share directly, so each measured hour is separated into beam and diffuse without estimating either, then transposed onto your roof plane in local clock time and derated for cell temperature and system losses. Because the shape of every day is measured rather than assumed, a cloudy morning that clears at two in the afternoon is modelled as exactly that. Panels fade half a percent a year. Assumes an unshaded roof.

Household use. The rest of the house takes a typical daily shape set by who lives there and whether anyone is home on a weekday, with a separate shape for real Saturdays and Sundays. Heating and cooling are built day by day from the measured temperature: cooling for every degree the afternoon tops 25° (plus a little for a warm night), heating for every degree the daily mean sits below 18°, at a heat pump’s or a bar heater’s rate, with cooling landing in the afternoon and evening and heating at breakfast and dusk. Hot water is so much a day per household, more in winter, reheating overnight. A car and a pool sit on top of the typical figure. It is a typical home, not your own meter data.

Appliances. Each is a real load with a power draw and a run length, not a lump of energy: a dishwasher is 1.8 kW for 45 minutes. Every load the home has is booked into the day, in the sun or not, so the “before” evening genuinely contains the dishwasher. Discrete cycles chasing the sun are placed on the sunniest days of each week, against a budget that empties as they are booked; car charging and pool pumps fill the best hours of the best days; pre-cooling brings 40% of a hot day’s cooling into the afternoon. An oven or a dryer in a house that is empty on weekdays can only chase the sun at the weekend. Hot water is never moved, because nobody chooses when a tank reheats. Within a day the runs are laid down biggest first into whatever surplus is left, so they genuinely compete. A run the surplus cannot cover still happens – you buy the shortfall – and nothing is ever moved to 3am. Because nobody follows a plan every day, only 85% of the planned moves are counted.

Money. Imports are priced hourly, flat or time of use, exports at the feed-in rate, and the whole tariff rises together each year. The battery is dispatched hourly with a C/2 inverter limit and fades each year; a payback beyond its fifteen-year life reads as never. Nothing is discounted to today’s dollars.

If the roof is still bare. Answer “not yet” to the solar question and the comparison changes shape: the thing every path is measured against becomes a bill with no solar in it at all, the array’s installed cost sits in front of all four paths, and putting the panels up and changing nothing else becomes a path in its own right – the floor a battery has to beat. The battery is then judged only on what it adds once the array is already earning, so it is never credited with the sunshine. Install prices are pre-filled from a typical 2026 per-kW rate for your market and a per-kWh rate for the battery, after any rebate that is standard there; both are starting points, and typing over either one stops the tool touching it again.

What it leaves out. Supply charges, demand tariffs, controlled-load rates, wholesale pricing, virtual power plant payments, charging the battery from the grid overnight, the value of backup during an outage, finance costs and one-off subsidies, any of which can move the answer. Hourly resolution overstates slightly how much of a spiky load the sun covers. The market and currency are picked from the timezone of the pin you drop, and its cost defaults are broad typical figures rather than quotes; drop the pin where we hold no cost data and the tool says so instead of guessing.

Data sources. Historical weather from Open-Meteo, licensed CC BY 4.0 and derived from the ERA5 reanalysis produced by ECMWF for the Copernicus Climate Change Service. Place search by Open-Meteo and Nominatim, using © OpenStreetMap data. Map tiles © Esri. Simulation and appliance artwork by OnSun.