Advanced Building Simulation

Self-organizing models for sentient buildings 181
time-step. Detailed lighting simulation runs are computationally intensive and
require considerable time. Obviously performing this number of simulations within a
time-step (of, say, 15 min) is not possible. To reduce the size of the segment of the
control state space to be explored, one could either couple devices (e.g. by dimming
the four space luminaires in terms of two coupled pairs) or reduce the number of permissible
device positions. An example for the latter would be to consider, at each
time-step, only three dimming states for each luminaire, namely the status quo, the
immediate higher state, and the immediate lower state. In the present case, this would
mean that D � 2 and P � 3, resulting in 9 electrical lighting options. Considering
4 candidate louver positions, the total number of required simulations would be
reduced to the manageable number of 36.
The concurrent simulation-based assessment of daylight and electrical light options
allows for the real-time incorporation of changes in room and aperture configuration,
as well as flexibility in the definition of the relevant parameter for performance variables
(such as the position of observer, etc.). However, the limitation of possible
dimming options at each time-step to the immediate adjacent positions may result in
the inability of the search process to transcend local minima and/or maxima. This
problem can be handled to a certain degree by considering additional randomly
selected control state options to be simulated and evaluated in addition to the default
“greedy” search option in the control state space (cp. Section 7.4.5.2).
The second approach to the generation and evaluation of alternative control
options involves a sequential procedure. In this case, first, the preferable louver position
is derived based on the methodology described earlier. The result is then combined
with a preprocessed matrix of various luminaire power levels. This matrix (or
look-up table) can be computed ahead of the real-time control operation based on the
assumption that the incident electrically generated light at any point in the space may
be calculated by the addition of individual contributions of each luminaire. The
matrix needs only to be regenerated if there is a change either in the configuration of
interior space or in the number, type, or position of the luminaires. The advantage of
this approach is the possibility to reduce computational load and extend the search
area in the control state space. The typical time interval between two actuation events
(e.g. change of louver position and/or change of the dimming level of a luminaire)
would then be generally sufficient to allow for the simulation of an increased number
of louver positions. Combining the results of the selected louver settings with the
matrix of electrical lighting states does not require real-time simulation and is thus
efficient computationally. As a result, a larger number of dimming options may be
considered and evaluated toward the selection of the preferable combined daylighting
and electrical lighting settings.
The following steps illustrate this process for a generic time-step as experimentally
implemented in IW:
1 Outdoor light conditions, the current louver position, luminaire power levels,
and the current time were identified (Table 7.5).
2 Simulations were performed for each of the eight candidate louver positions
based on the input data. Calculated performance indices for each louver position
were further processed to generate the utility value (UF) based on the preference
indices and corresponding weights (Table 7.6). Subsequently, the louver position
that maximizes utility was selected (105�).