New Trends - Comfort Control in Buildings - page 219

whole day and also as a function of the period of the year, energy saving in thermal
buffers of the building (the thermal tanks, see Sect. 2.2.4 ) can be consumed at these
instants in which it is most expensive to displace the energy demand from high to
low prices. Finally, some suggestions for buildings' technicians are offered.
DMPC deals with the decomposition of the typical optimisation problem, which
appears in the MPC approach, into a set of sub-problems where each sub-problem
is assigned to a different agent or controller. These agents exchange useful informa-
tion among themselves to obtain a suboptimal solution for the optimisation problem.
Although the comfort control problem in several rooms can be dealt within a cen-
tralised fashion as was explained in Sect. 5.6 , DMPC becomes attractive when the
process has a geographically distributed structure, for instance when the rooms are
spread out in several buildings, or when difficulties are encountered in the application
of centralised control, such as the centralisation of communications and computa-
tional power. The distributed controller can be used when the set or rooms shares
a common energy source, such of a renewable power source or a conventional one,
with a limited power capacity to fulfil the energy demand of the set of rooms. Some
works have dealt with users' comfort through the use of a DMPC controller (Ma et
al. 2011 ;Morosan et al. 2010a ; Scherer et al. 2014 ). The work described in Scherer
et al. ( 2014 ) is discussed in detail and some of its results are shown in the following
section, whereas in this section the other works are summarised.
In Ma et al. ( 2011 ) the control objective is to keep zone temperatures within
a comfort range while consuming the least energy by using predictive knowledge
of weather and occupancy. The HVAC system architecture is made up of an air
handling unit and a fan serving multiple variable air volume boxes which control air
temperature and flows in a network of thermal zones (the rooms). The air handling
unit uses a mixture of outside air and return air to generate cool air by using a cooling
coil. The cool air is then distributed by a fan to variable air volume boxes connected
to each room. A simplified model of the global HVAC system is made through a
resistor capacitor network. Then, the model is discretised using the Euler method
and linearised to use it with the DMPC algorithm. Moreover, thermal loads through
people and thermal conditions are predicted using historical data. The optimisation
problemwhichmust be solved by theDMPC algorithm, has a quadratic cost subject to
nonlinear constraints. In order to reduce the time to solve the nonlinear optimisation
problem, a sequential quadratic programming and dual decomposition is applied.
Another work by the same authors can be found in Koehler and Borrelli ( 2013 ).
A DMPC algorithm with one information exchange per step is proposed in
Morosan et al. ( 2010a ) to regulate the indoor temperature of a building. First, the
control strategy is proposed for a single zone building (a room), the key idea is to use
the future occupation profile of the room and to obtain a certain degree of thermal
comfort while the room is occupied. When the room is empty, any particular tem-
perature setpoint is imposed to the controller to save energy. Second, this approach
is extended to a multi-zone building (several rooms or environments). In the multi-
zone approach the DMPC is compared to both, a centralised and a decentralised
MPC algorithm, the centralised MPC consists of a single controller for all zones
whereas in the decentralised MPC each zone has its own controller but there is no