Advanced Building Simulation

exploratory implementations. The first case study addresses the daylight-based
dimming of the electrical lighting system in a test space (Section 7.4.6.2). The second
case study is concerned with the thermal control of a test space (Section 7.4.6.3).
7.4.6.2 Daylight-based dimming of the electrical light in a test space
We introduced the simulation-based control method using an illustrative case, which
involved the selection of a preferable louver position toward improving the daylight
availability and distribution in a test space (see Section 7.4.4). In this section, we consider
the problem of daylight-based dimming of the electrical lights in the same test
space (Mahdavi 2001a). The objective of this control strategy is to arrive at a configuration
of daylighting and electrical lighting settings that would accommodate the
desired value of one or more performance variables. The present scenario involves a
five-dimensional control state space. As indicated before, the daylighting dimension
is expressed in terms of the position of the external light redirection louvers. For the
purpose of this case study, eight possible louver positions are considered. The electrical
lighting dimensions encompass the dimming level of the four (independently controllable)
luminaires in the space. It is assumed that each of the four luminaires in the
test space can be at 1 of 10 possible power level states.
An attractive feature of a model-based control strategy is the diversity of the performance
indicators that can be derived from simulation and thus be considered for
control decision-making purposes. Furthermore, these performance indicators need not
be limited to strictly visual criteria such as illuminance levels, but can also address other
performance criteria such as energy use and thermal comfort. The lighting simulation
application LUMINA can predict the values of the following performance indicators:
average illuminance (Em) on any actual or virtual plane in the space, uniformity of illuminance
distribution on any plane in the space (U, cp. Mahdavi and Pal 1999), Glare
due to daylight (DGI, cp. Hopkinson 1971), Glare due to electrical light (CGI, cp.
Einhorn 1979), solar gain (Q), and electrical power consumption (C). The glare on the
CRT (GCRT) is also considered and is taken as the ratio of the luminance of the screen
to the background luminance. User’s preference for the desired attributes of such performance
variables may be communicated to the control system. Illustrative examples
of preference functions for the performance variables are given in Figure 7.10.
These preference functions provide the basis for the derivation of objective functions
toward the evaluation of control options. An objective function may be based on a single
performance indicator, or on a weighted aggregate of two or more performance indicators.
An example of such an aggregate function (UF) is given in Equation 7.6.
UF � w Em · P Em � w U · P U � w DGI · P DGI � w CGI · P CGI
� w GCRT · P GCRT � w Q · P Q � w C · P C
Self-organizing models for sentient buildings 179
(7.6)
In this equation, w stands for weight, P for preference index, Em for average
illuminance, U for uniformity, DGI for glare due to daylight, CGI for glare due
to electrical light, GCRT for glare on CRT, Q for solar gain, and C for power
consumption.
Needless to say, such weightings involve subjective and contextual considerations
and may not be standardized. Rather, preference functions and the weighting
mechanism could provide the user of the system with an explorative environment for