I have a Nest "smart home" thermostat because we are living in the future, and my comfort should approach the perpetual zen of Wall-E's La-Z-Boys with no more mental effort than it takes to believe in the power of Brawndo.
Instead, I wrote about 300 lines of Python to coerce the damn thing into achieving its life goals.
Nest has a habit learning function that seems only to learn to annoy with nonsensical, poorly timed automatic settings. It has an economy mode whose only job is to make you slightly yet perpetually uncomfortable.
It’s capable of providing MONTHS of historical data, explaining exactly why energy usage was better or worse during a given timeframe based on local weather conditions. One might expect when purchasing overly expensive HAL 9000 doppelgänger wall art that it could use that same local weather data to enhance real time performance...but no-can-do, Dave.
One might attempt using IFTTT to achieve a modicum of un-dumber thermostat-ing, but in 2019, Google decided to withdraw from IFTTT without rational thought, reasonable explanation, or viable alternative. Isn't the future brilliant?!
But at least you'll probably pull only some of your hair out while creating the Google Device Access Console application that is required for the aforementioned 300 lines of code to unf*ck your little blue/red wall pill. Fret not - there's about 50 lines of that code dedicated to keeping a few populated follicles on your skull during setup.
Initially, I borrowed heavily from Danielyaa5's therm-nest-command-line NodeJS project. It was super helpful to see how someone else navigated the twisted tangle of Google's device API, and I recommend his project if you simply need CLI access to your intellect lacking thermostat.
Before you begin, set your Nest to Heat/Cool mode, and turn Eco mode off. If the script detects a mode other than Heat/Cool or that Eco mode is enabled, it will exit without making changes (assuming that you prefer the Nest's own dumb behavior). If Heat/Cool mode is not available, the script won't work as written.
I wrote this script for use on a Raspberry Pi, but it should run equally well on Windows or Mac. The instructions below are somewhat geared toward RPi use.
To use the script, first install the requests library with pip:
pip install requests
or, if you like:
sudo pip install requests
Place the script into a directory that makes sense (perhaps /usr/local/bin). Run it with python undumb.py and follow the on screen prompts.
If you intend to run the script on an automated schedule (as it's intended), you may want to do the initial setup with sudo to give it access to the /etc directory. You could also create a script user on your system if you don't want to run the script with root privileges, but that's beyond the scope of this documentation.
There are a lot of setup steps if you haven't used the Google Cloud Platform before, and it can be frustrating and confusing. This link is slightly outdated but may help clarify some of the steps. Google is going to charge you a one-time fee of $5 as well, so there's that...
Once the script has successfully discovered your devices and created secrets.json, you can set up a systemd service (or Windows Task) to run it a few times per day, or even hourly like I do. Running it too often could re-dumb your thermostat, so stick to once per hour or so.
This file contains all API keys and settings required for the script to undumb your Nest. It isn't very secret. Sorry. Maybe I should rename it Victoria.json - for as well as she keeps a secret...
If you have not published the project with Google, and it is still sandboxed, the authorization code will expire weekly. Run the script with --refresh to update secrets.json with a new authorization code when it expires.
All units are unapologetically Fahrenheit. 'Murica.
Most of the keys are self-explanatory. Here's some additional explanation for those that aren't quite as clear by their names alone:
day_temp:your preferred set temperature during daylight hoursnight_temp:your preferred set temperature during night hourshysteresis:the gap between the heat and cool set points - do not use less than 3.0day_offset:the number of minutes before/after sunrise thatday_tempshould be setnight_offset:the number of minutes before/after sunset thatnight_tempshould be set
The remainder of the device settings have to do with your house's behavior with respect to outside temperature and radiant heat from the sun. The script will always use HEATCOOL mode to ensure that the indoor temperature never gets too far away from the desired set point (more accurately, it will just exit if HEATCOOL mode is not already set).
When the outside temperature is cold, it will make the heat set point the desired indoor temperature and raise the cool set point to hysteresis degrees above that. On the contrary, on a hot day, it will make the cool set point the desired indoor temperature and decrease the heat set point to hysteresis degrees below it. This way, your thermostat is undumb, and the indoor temperature always remains close to the actual desired temperature (the script even accounts for indoor relative humidity and modifies the set point to a "feels like" value!).
To accurately determine the threshold at which the set point should be biased toward heating or cooling, you'll need to enter some information about your home's behavior with respect to outside temperature. Most houses will require a bit of cooling even when the outdoor temperature is below the desired indoor temperature due to radiant heating from the sun as well as heat from indoor appliances or even people inside.
To over-simplify, set mid_uv_offset to the number of degrees below your house's indoor temperature where you expect to switch from heating to cooling. Set max_uv_offset 5-10 degrees larger than mid_uv_offset, and set min_uv_offset 5-10 degrees less. Set morning_offset to 60 minutes or so, and set evening_offset to around 120 minutes.
Keep reading for a more technical explanation (or at least more words on the subject).
The script uses current temperature and UV index from OpenWeatherMap, along with the *_uv_offset values to determine the heat/cool bias threshold. When the UV index is high, max_uv_offset will be subtracted from the set point to arrive at the threshold value. When the UV index is low, min_uv_offset will be used the same way. mid_uv_index is used to further modify the threshold in an attempt to account for latent heat in the evening and cooler temperatures in the morning. Evening and morning in this case are defined by morning_offset and evening_offset (in minutes). The script will delay applying a large UV offset in the morning and likewise continue applying a larger UV offset in the evening.
If that doesn't make sense, just use the default values and keep an eye on your house's behavior when the outside temperature swings above and below the desired indoor temperature (like Spring and Fall in most locales). If the script isn't being smart enough, modify the UV offset values and/or morning/evening offset values to suit.
Long story short, I guess you have to be smart enough to undumb your thermostat, lest this script just make it dumber.