Civil Engineering Reference
In-Depth Information
Fig. 5.9
Classical thermal comfort predictive control system architecture
The first control law defined by Eqs.
5.22
and
5.23
tries to find the best future
control sequence,
u
, which is able to minimise the PMV within the horizon (in this
case the PMV reference in Fig.
5.9
is zero, which is the desired value). Notice that,
at the same time, the cost associated to the control action is penalised for taking into
account associated costs. Remember that, the first part of the cost function, Eq.
5.23
,
can be related to the control action through the static nonlinear relation between
which relate indoor temperature and fancoil power, see Eq.
5.16
. Another fact that
is important to mention is that Eq.
5.23
is the same as Eq.
5.19
but in this case the
changes in the control variable are directly applied whereas in Eq.
5.19
the theoretical
indoor air temperature changes in relation to these control variable changes, which
can be obtained with the closed loop of the lower layer, were calculated and sent as
references to this layer.
In the second control law defined by Eqs.
5.22
and
5.24
, a two-stage optimisa-
tion algorithm is used. In a first stage, an indoor temperature reference,
w
T
(
,
is estimated by solving a nonlinear optimisation problem based on the PMV index
variables, the problem is solved looking for a PMV value equal to zero. Then, with
this value of temperature,
w
T
(
k
+
j
)
, an indoor temperature reference array of length
N
,
w
T
, is built with all its elements equal to
w
T
(
k
+
j
)
. Thus, in the second stage,
the MPC strategy based on Eq.
5.24
tries to minimise the future temperature track-
ing error, which is defined as the difference between the future predicted indoor
k
+
j
)
