We get power outages quite often in the Sofia suburbs where I live. I will spare you my rant about how a Balkan grid operator treats its customers, but in short, the grid goes down for no stated reason way too often. In July, a short grid glitch left the car charger refusing to charge until I restarted it. When our Tesla died in its sleep, I first suspected the grid (it wasn’t). And a heat pump cut mid-cycle is a recipe for trouble. Something had to be done.
Over the last three years, the PV inverter’s logs show that the grid dropped more than 140 times; that’s about once a week. Most cuts lasted a few minutes or less, 22 lasted an hour or more, and we spent about 60 hours off the grid in total.
At least the planned outages are announced. ERM Zapad (our grid operator) publishes a PDF every day listing the areas it will cut the next day. Last November I built an n8n workflow to read that PDF and look for our area, but n8n’s free tier had limitations, so I switched to a native Home Assistant custom integration. Every evening it fetches the next day’s PDF and, if our area is in it, pushes the outage window to our phones.
Curious about the history, I downloaded a year’s worth of these PDFs and compared them with what Home Assistant recorded at our end:
- 15 outages announced for “part of” our area, most of them for a full working day
- 4 of them cut our power for hours, but all were shorter than announced
- 6 more were a cut of a few minutes
- 5 never came
- 36 of the 46 cuts we did get were in no PDF at all
The notifications have been half-useful so far; I switch the sensitive things off before a planned window. The next step is an automation that stops the heat pump and the Tesla charging before the window and brings both back once the grid has been stable for a while. By the numbers above, most of the time it will be switching things off for hours to dodge a cut of a few minutes, or none at all.