This is a short introduction to the Hamburg startup Vilisto and some technical details about their current product.
My motivation for writing this text
I think their idea to reduce energy use via a self-learning system that requires no user input is great. It makes temperature regulation one thing less to think about as an end user and one thing less to misconfigure. While some companies apparently try to maximize the amount of time and attention end users spend on their devices, Vilisto takes the opposite approach and focuses on what matters from the customer’s point of view.
I wrote down what I currently know about their concept, because I find it interesting to get an idea of the kind of system they are building on a technical level.
The company’s thermostat solution aims to save energy by only reaching the desired room temperature when someone is present. This is done by detecting relevant changes in the thermostat’s surroundings, learning each room’s thermal properties, and trying to predict people’s routines.
Their solution consists of two types of devices: A TRV called ovis and a central internet connected unit called shepard. The thermostats collect data about each room, people’s presence as well as when windows are opened. The central unit aggregates this data, processes it, and controls the thermostats on a per room basis.
Goal: Learning each room’s thermal parameters, predicting disturbance variables (like open windows), and controlling a room’s temperature by setting the valves positions for each room individually while taking the desired room temperature into account as well as the predicted times of presence.
Setup: TRVs are transmitting all collected data to an internet connected base station that uses this information to feed the mathematical prediction models and control the valves’ positions. It also fetches current weather and weather predictions to feed the models.
Technology used by Vilisto
- ZigBee (802.15.4)
- MATLAB/Simulink with code generation
Sensors in the TRVs
Other inputs provided by the central unit (via meteorological services)
- Current outside temperature
- Current intensity with which the sun’s heat hits the area the room is located in
- Predictions of above parameters
Vilisto uses mathematical models for
- Detecting presence (sound, motion)
- Detecting disturbances like open windows (temperature, humidity, sound?)
- Learning room parameters, e.g. room size (air volume), thermal insulation (speed with which the room heats up when it is sunny and cools down once the sun does not shine and/or the outside temperature drops, e.g. during the night)
… and uses above data as inputs for the mathematical prediction models:
- Predicting presence, i.e. the need meet the desired room temperature at a certain time in the future.
- Predicting disturbances, e.g. open windows
These learned external room factors, detected changes as well as predicted presences and disturbances are used to calculate the optimal valves’ positions individually for each room while taking the respective desired room temperature into account.
- TUHH Spektrum November 2016, TUHH, Nov. 2016, accessed: 2019-03-23
- Smarter Einheizer: Ein Start-up aus Hamburg will mit intelligenten Thermostaten die Heizung automatisieren., brand eins, 2019, accessed: 2019-03-23
- Vilisto: Funktionsweise, accessed: 2019-03-23
- Vilisto: Jobs, accessed: 2019-03-23