Civil Engineering Reference
In-Depth Information
states representing the complex behaviour that have influenced the observed
actions. Nonetheless, the motivations for the actions are often difficult to access or
see, for example, someone's personal values, economics or even emotional con-
dition. So when building the behavioural model, the non-observable (hidden)
states model these factors.
An HMM is a probabilistic model composed by observable and hidden states,
as represented in Fig.
3
a. The observable states represent the symbols that the
model can output associated with the actions performed by the occupants, e.g.,
turning on or off the air-conditioning; or switching the lighting system, Fig.
3
b.
3.2 Building Markov Model Approach
Each building has its own use patterns according to the building purpose and its
occupants. As a result, each building will lead to a different model. Uncertainty is
always present when modelling, predicting and simulating real systems and
buildings are no exception. Many factors affect building performance that cannot
be known with certainty when the building is planned, designed, built, managed
and operated. But uncertainty also arrives due to the complex and unpredictable
nature of human behaviour. The ECM here proposed accounts for this stochastic
nature. The issue being on developing the ECM.
In an abstract way, and with no loss of generality, let us consider the building
illustrated in Fig.
4
. The building is considered as a set of spaces. Consequently,
the building energy consumption is the sum the energy consumption of all spaces.
Likewise,
the
building
configuration
is
the
combination
of
each
space's
configuration.
The spaces are strongly influenced by the stochastic (not deterministic) nature
of occupant behaviour. Therefore, it can be assumed that the state of these spaces
will also have a stochastic nature. Stochastic modelling can take several forms.
Some can be simple functions that produce the probability of an action given a set
of environmental conditions as inputs, while others like Markov chains (MC) can
Fig. 4 Decomposition of the building in a set of spaces in a lighting example. A white circle
represents the lighting system turned off and a shaded circle represents it turned on
