This document discusses developing intelligent controllers for HVAC systems. It proposes using disturbance observers to estimate disturbances in real-time and compensate for them to minimize energy use while maintaining comfort. An iterative learning controller is also proposed to minimize noise from the HVAC system by adapting to changing conditions. Advanced periodic controllers using a disturbance observer technique are discussed to reject periodic and non-periodic disturbances. Future efforts will focus on developing integrated sensing-controller chips using MEMS technology.
1. Intelligent Control of HVAC Systems
Tarunraj Singh & Nidal Al-Masoud
University at Buffalo – Buffalo, NY
March 25, 2003
INTRODUCTION
The focus of the proposed research is on developing advanced
controllers for micro and meso environments. The desired goal of
any HVAC system is to provide thermal comfort, code specified
IAQ (Indoor Air Quality) and noise free environment. These
controllers will tradeoff between minimizing energy consumption
and maximizing comfort metrics. Disturbance observers are
proposed to be used to estimate disturbances in real-time and
compensate for them in a feedforward manner. Since the
sampling time is significantly smaller than the time constants of
the thermal process, which characterize the dynamics of the micro
and meso environments, one can achieve non-hunting controllers.
To minimize the noise radiated by the HVAC system, an iterative
learning controller is proposed to be used which can adapt to
changing environmental conditions. The potential of incorporating
network of sensors which capture dynamics of the macro scale
systems into control of meso system is enormous. For example
the network of sensors can provide information of upcoming cold
or warm fronts and the meso system controller can schedule
appropriate control action to prevent the controller from high
demands of energy which would be typical of standard feeback
controllers.
Schematic of A/C System
T3ref
Controller
DESCRIPTION OF THE MODEL
f ref
Sr
Modeling of the HVAC process require modeling the
mixing process, the heat exchange process and the
dynamics of the thermal zone (micro or meso
systems). A schematic of a typical HVAC system is
shown.
Exhaust Air
Exhaust Air
T3 (t)
Return Air
qz(t)
ST
Control
Dampers
A mathematical model of the Heating process includes
equations for the mixing, heat exchange and controlling
the flow and the heat input to the system.
Fresh Air
T a(t), f (t)
a
qh (t)
Zone
qc(t) S f
T2 (t)
T1 (t)
Mixing Box
f
T1 (t ) = T3(T ) + [Ta − T3 (t )] a (adiabaticsteady process at the mixing box)
f (t )
ρC pVhT2 (t ) = f (t )ρ C p [T1(t ) − T2 (t ) ]− q h (t ) ( Transient processat the H/C or C/C)
ρC pVzT3(t ) = f (t ) ρCp [T2 (t ) − T3( t )] + q z
( Transient process at the Zone)
Filter
Heating Coil
Cooling Coil
Supply Fan
ØT1: Mixed fresh and return air temperature.
Øq c : Heat removed by the cooling coil
Øf(t): Volumetric flow rate
ØT3: Zone supply air temperature.
ØVz: effective zone volume
Øq z : thermal load
ØTa: Fresh air temperature
ØVh: effective heater volume
Øfa: fresh air volumetric flow rate
Øq h: Heat source input by the heating coil
d
Real Plant
e
u
P
CONTROLLER DESIGN
?
-
Periodic controllers are used to eliminate periodic disturbances. Since noise in HVAC
system can be represented as a periodic disturbance, a technique which can reject such
disturbances in the presence of modeling uncertainties is desirable. A control technique to
desensitize the periodic controller is shown in the schematic below.
For the rejection of non-periodic disturbances such as the effect of a person enetering or
leaving a room, the effect of opening and closing doors, etc., a disturbace observer is
proposed to be used to rapidly estimate the disturbance and compensate for it.
P-1
DOB
Q
Disturbance Observer (DOB) Structure
e
e -ST
p
e -ST
p
e -ST
p
W4
+
FUTURE EFFORTS
W3
+
A major thrust of the controller design in the future will involve the development of controllers
on a chip. With the potential for MEMS technology to provide inexpensive sensors, the
potential of designing and fabricating integrated sensing-controllers system is quite
reasonable and future work will focus on implementing advanced c
ontrollers under the
constraints of limited memory and flops.
W2
+
W1
+
-K(S)
Robust Periodic Controller
z