AdControl

using-genetic-algorithms-to-take-into-account-user-wishes-in-an-advanced-building-control-system

Project leader: Nicolas Morel
Project participants (EPFL) : Antoine Guillemin (now Neurobat AG), Prof. Dario Floreano

Duration and funding: January 2002 to December 2003, funding by EPFL

Summary

The AdControl project is an important part of the work carried out at LESO-PB for elaborating bio-mimetic control algorithms for building services (heating, cooling, ventilation, blinds, electric lighting), with the goals:

  1. to improve the indoor environment quality, i.e. the thermal, visual and air quality comfort (the first aim of building services is to improve the indoor comfort);
  2. to save energy used by the building services and reduce their environmental impact (in a perspective of sustainable development);
  3. to improve the acceptance of control algorithms by the users (a control algorithm allowing a large energy saving but irritating the user and/or not providing indoor comfort will certainly be disconnected by the users).

The AdControl project was focusing on the last but essential aspect: allowing the adaptation to the user’s preferences of advanced control algorithms for heating, blinds and electric lighting already available from other projects aiming toward biomimetic control.

The control algorithms already elaborated by our laboratory through several other projects are using artificial neural networks and fuzzy logic, which are convenient for describing the control rules and the various models (building, building services) to be considered. The first part of the project was devoted to the elaboration of adaptive rules, which would be adapted to the individual user preferences by the way of Genetic Algorithms.

In a second phase, measurements have been done on the LESO experimental building, involving 14 occupied office rooms (mostly with one or two persons in each room), during 9 months. The monitoring results have proved the interest of the new user-adaptive algorithms. These results can be summarized by the table below.

Controller type Energy savings Thermal comfort satisfaction Visual comfort satisfaction Rejection rate after 4 weeks
manual 84% 86%
automatic, without adaptation to user’spreferences -26% 84% 88% 25%
automatic, with adaptation to user’s preferences -26% 86% 89% 5%

The table shows clearly that the significant energy savings due to the automatic controller were not altered by the introduction of the adaptation to the user’s preferences, and that at the same time the rejection rate after 4 weeks was reduced considerably from 25 % to only 5 %.

Publications and references

A. Guillemin: USING GENETIC ALGORITHMS TO TAKE INTO ACCOUNT USER WISHES IN AN ADVANCED BUILDING CONTROL SYSTEM, thesis # 2778, EPFL, June 2003 (this reference includes the most comprehensive description of algorithms and results of the project)

A. Guillemin, S. Molteni: AN ENERGY-EFFICIENT CONTROLLER FOR SHADING DEVICES SELF-ADAPTING TO THE USER WISHES, Building & Environment 37 (2002)

A. Guillemin, N. Morel: EXPERIMENTAL RESULTS OF A SELF-ADAPTIVE INTEGRATED CONTROL SYSTEM IN BUILDINGS: A PILOT STUDY, Solar Energy 72 (2002)

A. Guillemin, N. Morel: APPLICATION OF GENETIC ALGORITHMS TO ADAPT AN ENERGY EFFICIENT SHADING DEVICE CONTROLLER TO THE USER WISHES, EPIC 2002 Conference, Lyon, France

A. Guillemin, N. Morel: EXPERIMENTAL ASSESSMENT OF THREE AUTOMATIC BUILDING CONTROLLERS OVER A 9-MONTH PERIOD, CISBAT 2003 Conference, EPFL, Lausanne, Switzerland

Antoine Guillemin, Nicolas Morel, Jean-Louis Scartezzini: ENERGY EFFICIENCY AND USER ADAPTATION IN AUTOMATIC BUILDING CONTROL, submitted to Energy & Building in October 2003