Rooftop PV, four winters later

The rooftop PV has now been through four winters, and I finally sat down with the electricity bills to see what it actually did. In June 2022 I put 6.4 kWp of panels, a Fronius GEN24 inverter and a 7.7 kWh BYD battery on the roof, plus an Ohmpilot to dump surplus into the hot water tank and a Wattpilot charger that waited three years for a car. About 14 800 € all in, and no feed-in – not one kilowatt-hour has ever been sold back to the grid.

I have the invoices going back to 2012 and every month of inverter data since the panels went up, so what follows is measured, not modelled. The baseline is 2016 to 2021, six years when the house bought every kilowatt-hour it used, 9.5 MWh a year on average. And because the parts of the system earn nothing like the same return, each one gets its own verdict.

The array

Production by calendar year: 7.1 MWh in 2023, 7.3 in 2024, 7.4 in 2025, on plan against the 7.6 MWh PVGIS promised, and never once allowed to run flat out. This is a zero-export system, so when the battery is full and the house isn’t drawing, the inverter throttles back and the energy is never made at all. So the production figure measures demand, not the panels. August 2026 was the best month in four years at 1 086 kWh, and nothing changed on the roof, the car was simply pulling 312 kWh out of it.

What it took off the bill

The house bought 5.0, 5.4 and 7.3 MWh in the three full years since the panels, and 8.5 MWh in the twelve months to July 2026. After roughly 14 800 €, I buy 90% of the electricity I used to buy. The house is outgrowing the roof. And that 90% is an average of a summer and a winter that look nothing alike.

In summer the array does its job: June through August, grid draw was a sixth of the old figure in the first two years and a quarter in 2025, because the load and the sunshine happen in the same hours. Winter is the opposite, and it is getting worse: 30% below baseline the first winter, then 26%, then 7%, and last winter the house drew 12% more from the grid than before the panels existed. Six kilowatt-peak on a Sofia roof in January is a rounding error against a heat pump.

Diverter, heat pump, car

Pre-panel night consumption averaged 1 940 kWh a year, nearly all of it resistive water heating on the cheap tariff. In the three full PV years the Ohmpilot has averaged 1 908 kWh a year into the tank from surplus that had nowhere else to go. Thirty kilowatt-hours apart, under 2%. For 886 € it removed a whole category of consumption from the bill.

Then the Ecodan demoted it. A heat pump makes hot water at 2.4 to 3.5 times the efficiency of an immersion element – measured, not nameplate – so that became its job, and a kWh through the diverter is now worth about a third of one off the bill. Since the Tesla arrived in November 2025 the Wattpilot has been eating the diverter’s lunch: January to August, the diverter fell from 1 400 kWh in 2025 to 771 this year while the car took 1 566 kWh. With zero export the surplus is a fixed pot, so the car took the diverter’s share. That is the right way round: the car has no cheaper source, hot water does.

What it saves

Bulgarian household bills have no standing charge, so the marginal price is a real number: 0.15 €/kWh by day now, up 56% since 2014, and night has stayed at roughly 59% of day throughout. The ratio matters because the diverter and the car displace night-tariff kilowatt-hours and the panels displace day ones.

Valuing each block at the tariff it actually displaced, with the generation figures pulled out of Solar.web by a small exporter I wrote:

That is 22% of the system back in a little over four years. The run rate is about 939 €/yr and climbing, though the climb is the tariff rising and the car arriving, not the system doing more work.

What it cost

As invoiced, VAT included. Mounting and cabling were never itemized, so call it 14 800 € rather than the 14 500 I quoted at the time.

The battery is 44% of the column.

The battery

Split the 939 €/yr by what earns it and the panels, inverter, diverter and charger return about 690 €/yr on 8 225 €, paid back in 12 years, comfortably inside the panel warranty. The battery returns about 249 €/yr on 6 575 €, which is 26 years, longer than the pack will live.

And 249 € is the generous reading, the metering flatters it. January 2026: the house bought 1 700 kWh and put 51 kWh into the battery. All month. Forty-four percent of the capital, asleep in the month the bill peaks. Turns out the one thing I had wrong in its favor is timing – 86% of what it discharges lands in day-tariff hours, mornings and evenings, so a stored kilowatt-hour is worth nearly the full day rate. It still does not rescue the arithmetic. Blended, the whole system breaks even in the second half of the 2030s, and the inverter will not see the far side of that date.

But the price I paid was a 2022 price. I bought this on the way out of COVID and in the first months of the war in Ukraine, when the worry was whether the grid would stay up and what electricity would cost by winter. Storage was the point, and that was the going rate for it. The same BYD box sells new for around 3 000 € with VAT today, less than half. At my measured 249 €/yr that is a 12-year payback, the same arithmetic the panels manage. The battery was never a bad idea, it was a bad price. I would buy it again, but as backup power, not as an investment.

What I would tell someone about to spend this

Whether solar is worth it depends on which box on the wall you mean. The panels, inverter and diverter are a sound purchase: 8 225 € returning 690 € a year, rising with the tariff, paid back well inside the panel warranty. The battery is not, at 2022 prices. And know which season you are buying: the array wipes out the summer bill and barely dents January’s.

Efficiency first, generation second, storage when you can justify it on something other than money. That order would have saved me a few thousand euro. It would also have made for a much shorter post.