| KEYWORD |
Feeding Building forecasting models via hybrid simulated/measured database
keywords BUILDING AUTOMATION, BUILDING SIMULATION, MACHINE LEARNING
Reference persons GIACOMO CHIESA
External reference persons Possibile correlatore industriale
Research Groups ETD
Thesis type RESEARCH / EXPERIMENTAL
Description The application of forecasting algorithms in the building secotr is progressively growing, allowing estimating of building energy needs and temperatures. Nevertheless, the initial lack of historical monitored data is causing industries to struggle to apply their models - especially for free-running usages - missing the feeding data. Nevertheless, new building dynamic simulation platforms, including accurate weather data and the possibility to semi-automatically calibrate models, can produce needed data to feed ML instruments, allowing a few months/weeks of data for the first application. The thesis candidate will develop an approach to progressively substitute simulation data with monitored ones while they are stored. The thesis will be tested on actual buildings, monitored via cloud solutions, for which a sufficient amount of historical data is available to try the approach (2-3 years). The work is part of the EU H2020 projects EDYCE and PRELUDE
Required skills ML algorithm, building monitoring/variables
Deadline 12/03/2025
PROPONI LA TUA CANDIDATURA