Civil Engineering Reference
In-Depth Information
Fig. 5.20
General architecture of the nonlinear controller for users' thermal comfort
it. This is not a common degree of freedom for control purposes in buildings, but in
this work experimentation using water flow was performed to analyse if this control
action, besides fancoil velocity, can provide improvements to the performance of the
closed loop system. Therefore, the use of a split-range controller is necessary, as it
will be required afterwards.
More specifically, this section presents the control results obtained in Castilla
et al. (
2014
) where a hierarchical control strategy, which ensures thermal comfort
inside a certain environment, and at the same time, tries to reduce energy consumption
is developed. This hierarchical controller is formed by two layers: (i) the upper
layer consists of a nonlinear model predictive control optimizer, which provides
impulse air temperature setpoints for the HVAC system to reach a thermal comfort
situation, that is, to reach a PMV
0. This controller uses a nonlinear model
temperature setpoints by means of linearising the forced response of the system
at each sample time, following a methodology similar to Plucenio et al. (
2007
)
which was presented in Sect.
5.2.3
. (ii) The lower layer, which contains a classical
PI controller combined with a split-range controller, to obtain the desired impulse
air temperature setpoint saving as much energy as possible by regulating the fancoil
unit through both, the fancoil velocity,
V
Fan
signal in %, and the water flow,
Ap
Va l ve
signal in l/min. Notice that, instead of the valve aperture, a desired water flow can be
demanded by implementing a cascade controller acting on the valve aperture. This
is useful under circumstances of changing water flow from the main circuit.
Compared to the work in Ferreira et al. (
2012
), where an MPC based on ANN
is presented, the proposed approach has several differences. The ANN in Ferreira
et al. (
2012
), which are used to estimate the value of the PMV index function as
well as to predict the behaviour of several variables such as air temperature and rela-
tive humidity, is able to predict the PMV value with high accuracy, as a main benefit.
However, using a complete nonlinear model of the process increases the complexity
of the optimisation problem to solve. On the other hand, in the approach presented
here an approximated linear model is calculated in each sample time causing a tiny
loss of accuracy but allows the users to use QP algorithms since the computational
effort to solve the optimisation problem decreases. In addition, the proposed strategy
has been implemented in the characteristic room of the CDdI-CIESOL-ARFRISOL
building, see Fig.
5.21
, in order to obtain real results that validate the proposed con-
trol system. However, the results obtained with this control strategy can be easily
=
