Prediction MODELS
Measuring building variables gives us crucial information on building operation. Sensor data captures relevant fluctuations which are often missed by simulation models.
Using artificial neural networks on this data, we train prediction models for energy consumption with an objective to get as close to the real values as possible. The trained models could then be used to predict future energy consumption.
The box plot above shows four variables that are measured for three single-family residential buildings in Switzerland. Although the buildings are very similar in age, construction, orientation etc., we can see slight differences in their performance.
Disaggregating electricity
We use density-based clustering do disaggregate electricity consumption into major end uses. This is very effective in deriving the heating energy consumption for residences with electrical resistance heaters. By knowing the heating energy from the total electricity profile we determine the potential for energy improvement and retrofitting.
