Today’s smart thermostats are significantly more advanced than their counterparts from even five or six years ago. They are light years ahead of the old programmable thermostats of the 1980s and 90s. A modern smart thermostat can actually learn and modify its own programming.
Learning is artificial, of course. Smart thermostats are not equipped with biological brains. But they can artificially learn by gathering and analyzing data. They can also modify their own programming based on what is learned.
For the record, Vivint Home Security published a blog post several years ago discussing how smart thermostats work. Their post was mainly about automating programming and making adjustments remotely. I mention it to say that not all smart thermostats can artificially learn and self-modify. Some of them are limited strictly to automation and remote access.
Initial Programming
Artificially learning begins with analyzing initial programming. A smart thermostat will gather and store data about the first program a consumer enters. It will also record data generated with every manual temperature adjustment. Between these two datasets, the thermostat begins building its database for learning.
As a rudimentary example, if data demonstrates that a homeowner manually turns down the temperature several nights and a row, the thermostat will take note of that in an attempt to make the adjustment automatically.
Temperature Preferences
Smart thermostats also gather data on temperature settings throughout the day. The data serves to help software recognize any patterns in user preferences. Recognized patterns become the basis for automated adjustments.
This portion of the learning process takes quite a bit of time. Software needs to distinguish between regular preferences and one-off temperature changes so that programming is not skewed. But over several weeks and months, programming accuracy tends to improve quite a bit.
Sensing Occupancy
Some smart thermostats have built in sensors designed to detect occupancy. They offer limited function all by themselves. But when a smart thermostat is part of a more comprehensive system that includes smart lights, motion sensors, cameras, etc., sensing occupancy is a lot more accurate.
With this function, a smart thermostat can begin building a database of information around when a home is normally occupied versus when it is empty. Occupancy sensing can also account for when occupants are normally awake versus when they are sleeping. Occupancy data can be combined with other datasets to modify programming.
External Data Gathering
Even external data can be helpful to a smart thermostat designed to maximize efficiency. Given that smart thermostats connect to the internet, they can go out and fetch information pertaining to current weather, temperature, and so forth. All the data can be plugged into programming algorithms.
Smart thermostat software can also account for the current season. If it is summer, programming will focus on cooling. Heating will be the priority during the winter. All the external data the system gathers provides more information through which the thermostat can learn.
The Consumer Is Still in Control
It is important to remember that a smart thermostat cannot truly think on its own. It also cannot truly learn. All it can do is gather data and analyze it. When all is said and done, the consumer is in control of the device and all heating and cooling decisions. Anything a smart thermostat does can be overridden by the homeowner.
Now you know how a smart thermostat artificially learns and adjusts its own programming. I do not have one yet in my home, but it is on my home automation to-do list. I am looking forward to seeing just how smart modern smart thermostats are.