ModKI

AI applications are currently spreading in many areas of everyday life (chatbots, translators, autonomous driving, etc.). Over the life cycle of residential and non-residential buildings, machine learning (ML), natural language processing (NLP) and reinforcement learning (RL) applications promise great potential for optimising technical building processes. ML, NLP and RL methods are used to learn patterns or behaviour of complex systems that can be used to recognise, predict or influence their behaviour. The research project will transfer this method to building automation (BA) applications. The research project analyses and evaluates fields of application of AI applications for socially, ecologically and economically optimised building processes. In addition to supporting applications for planning processes (e.g. NLP-based automatisms for creating function lists etc.), the potential of recurring applications based on learned building or system dynamics is to be analysed.

An important research goal is interoperable modularization and reusability of the AI applications that are assessed as potentially significant. For this purpose, basic models (e.g. RL agent for the optimization of building setpoints) must be designed, which are trained for the specific use case during the runtime of the application. In addition, solutions for a system architecture are being designed that enable the operation of AI applications on the one hand and further training of these on the other. Such modular solutions are to be integrated into the current BA system architecture (e.g. as standardized function modules). In the research project, prototype AI applications are being developed in cooperation with a manufacturer of automation systems and two municipal users of building automation. These demonstrators are intended to validate modular AI applications in terms of function and efficiency.

Project partner: ---

Funding provider: Zukunft Bau

Duration:  2024- 2027