ThinkHome: Improved energy efficiency based on artificial intelligence in future homes

Status

completed

Summary

Starting point / motivation

Current automated buildings are equipped with a great amount of sensing, actuating and controlling equipment that should guarantee comfort for the users. Although automation technology aids in controlling building functions, the user is in practice still confronted with manually having to adapt the system to changes in usage and comfort desires so that an optimized operation is still achieved. However, the user is often over challenged with exploiting all the different possibilities not last because of an excess of different possible configurations. For that reason even technically versed users resort to standard profiles, and thus are not able to exploit all energy savings potentials in the building. Additionally, current automation systems often do not reach an energy-optimized operation of the building, as many influence factors like the current state of the building hull or external influences like weather remain unconsidered. Therefore, it is in many cases not possible to exploit all energy saving potentials of the building in a comprehensive way.

Contents and Objectives

Main goal of the ThinkHome approach is to develop a system for residential homes which supports residents in keeping a comfortable environment while operating the building in an energy-efficient way. ThinkHome shall be able to use information on its inhabitants, occupancy, the building structure and hull and many more parameters to optimize different building services, mainly the energy-intensive subsystems of the heating/ventilation/air-conditioning domain and lighting/shading. The available information shall be considered in novel control approaches and applications for the daily operation of a building. Furthermore, the integration of renewable energy sources shall be promoted. ThinkHome shall also be able to detect and learn user and usage patterns, and use this information to autonomously execute routine tasks on behalf of the users.

This autonomous control system shall be designed as software system that exploits mechanisms from the artificial intelligence domain and that integrates all available information on the building and its users in the control strategies. For this purpose the information shall be represented in a comprehensive knowledge base. The knowledge base shall incorporate a preferably complete model of the building and its users so that this information (building structure, occupancy, user activities, …) can be used by the control system for autonomous and energy-efficient building control. Above this autonomous control, the user shall also be provided with feedback on energy consumption and further reduction possibilities.

Methods

After a thorough analysis and specification of a complete system concept, selected parts of this comprehensive system shall be prototypically realized. For this purpose, it is first important to specify (novel) use cases and to characterize the necessary system components. In parallel, related work and previous approaches shall be analyzed for possible starting points. Herein focus is put on a complete description of parameters that are necessary for an optimized operation of a building. In a next step, all parameters are specified that influence the control of building services. These parameters are subsequently represented as knowledge base which makes knowledge accessible by intelligent software agents.

The intelligent and energy-efficient control of the building is realized by a multi-agent system. It is designed based on the use cases and developed following a tailored agent design methodology that ensures a later implementation with established software frameworks.

Furthermore, intelligent control strategies shall be developed that are capable of exploiting unused savings potentials. Hereby, focus is put on the detection of recurrent patterns that can be used by the system’s control strategies to predict future situations and to act in a predictive way according to them. To describe these patterns in the system and enable logical inference, methods from the artificial intelligence domain will be evaluated and adapted for the use in ThinkHome.

Results

In the course of the ThinkHome project, a complete system concept including use cases, scenarios and a detailed system architecture was developed. For an integration of building information into the intelligent control strategies of the system, an integration process was specified which allows the transformation of information from a building information model (gbXML) into the ThinkHome knowledge base. A software tool allows the automatic inclusion of knowledge about the building from architectural software such as AutoCAD or Autodesk into the ThinkHome system where the information can be used for an optimized control.

Therefore, a building model was created in Autodesk, which has been extended with energy-relevant data from building physics in Ecotect. Subsequently, the model was realized as MATLAB/Simulink simulation, in order to support the evaluation of the system. At the same time, the knowledge base for the ThinkHome system was designed and the most important parts were implemented in detail. Special focus was put on realizing a comprehensive representation of energy information, like energy producers and consumers in the building as well as exterior energy providers.

Likewise, the control system was specified as multi-agent system and parts were prototypically implemented. In course of this task also the interfaces to external and internal systems were specified as well as the agent architecture and the agent framework were defined after a thorough analysis.

Starting from the use cases also novel control strategies based on habit profiles were realized and prototypically implemented. Their prototypical implementation showed that with the application of predictive methods in a residential home, energy can be saved without loss of user comfort.

Project management

Univ.Prof. Dr. Wolfgang Kastner
Vienna University of Technology, Institute of Computer-Aided Automation Automation Systems Group

Technische Universität Wien, Rechnergestützte Automation
Univ. Prof. Dr. Wolfgang Kastner
E183/1, Treitlstr. 1-3
A-1040 Vienna
Tel.: +43 (1) 58801-18320
Fax: +43 (1) 58801-18391
E-Mail: k@auto.tuwien.ac.at