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

Abstract

From a sustainable development perspective, the newly developed automatic controllers for building services are very promising in that they increase energy efficiency and reduce commissioning and maintenance costs. But a major problem has appeared as the automatic building control systems have been implemented: the user rejection of this kind of system is quite high. This is mainly due to a lack of user considerations in the controllers. An integrated blind, electric lighting and heating control system that adapts to user wishes on a long-term basis has been developed in this work to deal with this issue. The adaptation of the control system to user wishes was achieved by means of Genetic Algorithms. They have been seen to be the most appropriate optimization method for this task. They ensure a 100% convergence whereas standard search methods such as Gauss-Newton and Nelder-Mead converge in less than 25% of the time and Simulated Annealing method converges in about 75% of the time. In addition, simulations with a consistent virtual user have shown that the user adaptive controller is capable of anticipation. Nine months of experimental tests were carried out in 14 office rooms of the LESO building with a total of 23 users concerned. Three controllers were compared: a manual control system, an automatic controller without user adaptation and an automatic controller with user adaptation. Tests were conducted in a similar fashion as clinical randomized trials are carried out: control systems are randomly attributed to rooms and users do not know which system they have (single-blind study). Results show that the automatic control rejection percentage is greatly reduced with the user adaptive system. Indeed, after four weeks with an automatic control, 25% of the users with the non-adaptive system reject the automatic control, whereas only 5% of the users with the user adaptive system reject it. These percentages depend neither on age or gender of the user, nor on the number of occupants in a room. Moreover, the energy savings due to automatic control (26% compared to a manual system) are not reduced by the user adaptation. These large energy savings are mainly due to the predictive feature of the heating controller and to the efficient control of electric lighting. In addition, indoor comfort is slightly improved by the automatic controllers for both thermal and visual aspects. The indoor comfort is even slightly more improved by the user adaptive control compared to the non-adaptive one. The user adaptation has not converged properly in the mechanical workshop, a space used by several persons and also considered in the experiments. It has been concluded that user adaptive systems are probably not appropriate for places with irregular users, such as workshops, libraries, corridors and all public spaces.