Current research projects

Performance tests of condensing units

Ionocaloric cooling

Laseroptical measurement

Innovative small helium liquefier

Development of test methods and test rigs for stationary integrated refrigeration units

Measurement of insulated packaging

Influenced melting point of water by magnetic field

In-Situ-Swelling Behaviour of Polymer Materials in Flammable Fluids

Calibration of Low Temperature Sensors

In-situ investigation concerning the swelling behaviour of polymer materials under elevated pressures and temperatures

Hybrid- Fluid for CO2-Sublimation Cycle

Low Temperature Tribology

Investigation of material-dependent parameters

Micro fluidic expansion valve

Pulse-Tube-Refrigerator with sealed compressor

Characterisation of Superconductors in Hydrogen Atmosphere

You are here: / Home

Optimizing HVAC operation with machine learning

BMWi Euronorm Innokom

01/2019–05/2021

Dr.-Ing. Thomas Oppelt

+49-351-4081-5321

send mail

in progress

Intelligent control of HVAC systems – high comfort with low energy demand

Motivation

During operation, the energy efficiency of many HVAC systems remains considerably below the value predicted when planning. One reason is that especially complex systems with multiple generators, storages and consumer locations frequently are not operated optimally.

Aim of the project

Development of a tool for optimizing the operation of HVAC systems which uses machine learning (ML) methods and data from the digital building model (Building Information Model, BIM):

Optimization goal: high energy efficiency with at the same time high comfort for users
Saving operating costs, energy and carbon dioxide emissions due to increased efficiency
Continuous autonomous improvement of the ML algorithm by learning from new measured data with auto-adaptive reaction to changing conditions (building, system, use, smart meter for real time billing of energy and media, etc.)

Approach

Reproduction of the real system’s thermal-energetic behaviour in the machine learning system, training with BIM data, measured data and a digital twin of the real system

Application of ML methods for load forecasting (weather, usage patterns)
Automatic classification of utilisation scenarios, fault detection
Integration of available tools for efficient simulation of indoor air flows and for calculating energy demands
Co-Validation of optimization tool, experimental studies and digital twin