I bought a WiFi electriQ DESD8LW desiccant dehumidifier for the garage to stop rust. I can set it to achieve 50% humidity and it will happily turn off when it gets there using 610W.
The thing is, condensation on tools occurs when their temperature is below air dew point which means if avoiding rust is the only goal here then we’re unnecessarily paying for dry air.
I solved that, 100% for free, albeit my employer paid for the ChatGPT credits.
Basically I take the temperature and humidity data from nearby airports, extrapolate that to where I live and use the garage temp and humidity from the dehumidifier to predict when condensation is likely to form and turn it on to prevent that.
At first I thought I should use forecast rather than actual weather data but because my garage is fairly well air-tight/insulated it lags far enough behind that extrapolated actuals are a better measure.
I’ve also got a way to tell it more accurate readings manually using Telegram e.g. the temp of my lathe. The model uses these to tune its predictions.
Here’s the (free) Grafana dashboard that I’ll improve a bit next week.
I use a free for life Oracle VM, PostgreSQL as the DB. Open Meteo and Metar for the forecast and actual weather data and Tuya APIs to interact with the dehumidifier. Telegram is used to send me alerts when it’s turning it on and also to suggest good times to take manual readings to tune the model that I can reply to.

