Comparison of MPC Strategies for Building Control
J. Drgoňa and M. Kvasnica
Slovak University of Technology in Bratislava, Slovakia
June 20, 2013
IAM
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 1 / 18
Motivation
Problem: Almost 40% of final energy use in the world goestowards comfort control in buildings - heating, ventilation and air
conditioning (HVAC).*
In European countries it is over
76% !!!*
Goal: Increasement of building’s energy efficiency
Solution: Building control
*International Energy Agency ‘Energy efficiency requirements in building codes, energy efficiency policies for newbuildings’ 2013 OECD/IEA.
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 2 / 18
Motivation
Problem: Almost 40% of final energy use in the world goestowards comfort control in buildings - heating, ventilation and air
conditioning (HVAC).*
In European countries it is over
76% !!!*
Goal: Increasement of building’s energy efficiency
Solution: Building control
*International Energy Agency ‘Energy efficiency requirements in building codes, energy efficiency policies for newbuildings’ 2013 OECD/IEA.
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 2 / 18
Motivation
Problem: Almost 40% of final energy use in the world goestowards comfort control in buildings - heating, ventilation and air
conditioning (HVAC).*
In European countries it is over
76% !!!*
Goal: Increasement of building’s energy efficiency
Solution: Building control
*International Energy Agency ‘Energy efficiency requirements in building codes, energy efficiency policies for newbuildings’ 2013 OECD/IEA.
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 2 / 18
Motivation
Problem: Almost 40% of final energy use in the world goestowards comfort control in buildings - heating, ventilation and air
conditioning (HVAC).*
In European countries it is over
76% !!!*
Goal: Increasement of building’s energy efficiency
Solution: Building control
*International Energy Agency ‘Energy efficiency requirements in building codes, energy efficiency policies for newbuildings’ 2013 OECD/IEA.
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 2 / 18
Presentation Outline
1 Building Modeling and Control
2 Model Predictive Control
3 Case Study
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 3 / 18
Presentation Outline
1 Building Modeling and Control
2 Model Predictive Control
3 Case Study
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 4 / 18
Building Thermal Control Scheme
����
��
Building
Measurements
Energy minimization
Comfort criteria
OptimizationHeating/Cooling
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 5 / 18
Single Zone Building Model
State Variables
x1− floor temperaturex2− internal facade temperaturex3− external facade temperaturex4− internal temperature
Disturbances
d1− external temperatured2− occupancyd3− solar radiation
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 6 / 18
Single Zone Building Model
State Variables
x1− floor temperaturex2− internal facade temperaturex3− external facade temperaturex4− internal temperature
Disturbances
d1− external temperatured2− occupancyd3− solar radiation
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 6 / 18
Control Objectives
Thermal Comfort
Minimization of Energy Consumption
qx
qu
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 7 / 18
Control Objectives
Thermal Comfort
Minimization of Energy Consumption
qx
qu
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 7 / 18
Control Objectives
Thermal Comfort
Minimization of Energy Consumption
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 7 / 18
Presentation Outline
1 Building Modeling and Control
2 Model Predictive Control
3 Case Study
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 8 / 18
Model Predictive Control
Objective function
minu0,...,uN−1
N−1∑
k=0
ℓ(xk , uk)
Constraints
xk+1 = Axk + Buk + Edk ,
x ≤ xk ≤ x ,
u ≤ uk ≤ u,
x0 = x(t)
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 9 / 18
Model Predictive Control
Objective function
minu0,...,uN−1
N−1∑
k=0
ℓ(xk , uk)
Constraints
xk+1 = Axk + Buk + Edk ,
x ≤ xk ≤ x ,
u ≤ uk ≤ u,
x0 = x(t)
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 9 / 18
Different Formulations of the Objective Function
Reference Tracking + Energy Minimization (Basic)
ℓ(xk , uk) = qx(Cxk − r)2 + quu
2k
r
Comfort Zone Tracking + Energy Minimization (CZT)
ℓ(sk , uk) = qss2k+ quu
2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk
+ϵ
-ϵr
Minimization of Zone Violations + Energy Minimization (Hybrid)
ℓ(δk , uk) = qδδk + quu2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk(sk > 0) =⇒ (δk = 1)
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 10 / 18
Different Formulations of the Objective Function
Reference Tracking + Energy Minimization (Basic)
ℓ(xk , uk) = qx(Cxk − r)2 + quu
2k
r
Comfort Zone Tracking + Energy Minimization (CZT)
ℓ(sk , uk) = qss2k+ quu
2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk
+ϵ
-ϵr
Minimization of Zone Violations + Energy Minimization (Hybrid)
ℓ(δk , uk) = qδδk + quu2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk(sk > 0) =⇒ (δk = 1)
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 10 / 18
Different Formulations of the Objective Function
Reference Tracking + Energy Minimization (Basic)
ℓ(xk , uk) = qx(Cxk − r)2 + quu
2k
r
Comfort Zone Tracking + Energy Minimization (CZT)
ℓ(sk , uk) = qss2k+ quu
2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk
+ϵ
-ϵr
Minimization of Zone Violations + Energy Minimization (Hybrid)
ℓ(δk , uk) = qδδk + quu2k
s.t. r − ǫ− sk ≤ Cxk ≤ r + ǫ+ sk(sk > 0) =⇒ (δk = 1)
+ϵ
-ϵr
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 10 / 18
Presentation Outline
1 Building Modeling and Control
2 Model Predictive Control
3 Case Study
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 11 / 18
Simulation Study
Closed Loop SimulationParameters
Prediction horizon:N = 10
Sampling time:Ts = 444 sec
Simulation time:Tsim = 31 days
Initial indoor temperature:x4 = 10
◦C
No weather predictions
Evolution of ExternalTemperature
0 5 10 15 20 25 300
5
10
15
20
Time [days]
Tem
pera
ture
[◦C
]
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 12 / 18
Simulation Study
Closed Loop SimulationParameters
Prediction horizon:N = 10
Sampling time:Ts = 444 sec
Simulation time:Tsim = 31 days
Initial indoor temperature:x4 = 10
◦C
No weather predictions
Evolution of ExternalTemperature
0 5 10 15 20 25 300
5
10
15
20
Time [days]
Tem
pera
ture
[◦C
]
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 12 / 18
PI Controller
0 5 10 15 20 25 3014
15
16
17
18
19
20
21
22
referenceinternal temperaturecomfort boundaries
Time [days]
Tem
pera
ture
[◦C
]
0 5 10 15 20 25 30−1000
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
PI
Time [days]Energ
yflow
[W]
Control strategy PI
Thermal comfort [%] 87.5Energy consumption [kWh] 753.0Energy savings [%] -
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 13 / 18
Reference Tracking (Basic)
0 5 10 15 20 25 3014
15
16
17
18
19
20
21
22
referenceinternal temperaturecomfort boundaries
Time [days]
Tem
pera
ture
[◦C
]
0 5 10 15 20 25 30−1000
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
PIBasic
Time [days]Energ
yflow
[W]
Control strategy PI Basic
Thermal comfort [%] 87.5 89.2Energy consumption [kWh] 753.0 722.7Energy savings [%] - 4.0
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 14 / 18
Comfort Zone Tracking (CZT)
0 5 10 15 20 25 3014
15
16
17
18
19
20
21
22
referenceinternal temperaturecomfort boundaries
Time [days]
Tem
pera
ture
[◦C
]
0 5 10 15 20 25 30−1000
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
PICZT
Time [days]Energ
yflow
[W]
Control strategy PI Basic CZT
Thermal comfort [%] 87.5 89.2 84.1Energy consumption [kWh] 753.0 722.7 684.0Energy savings [%] - 4.0 9.1
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 15 / 18
Minimization of Comfort Zone Violations (Hybrid)
0 5 10 15 20 25 3014
15
16
17
18
19
20
21
22
referenceinternal temperaturecomfort boundaries
Time [days]
Tem
pera
ture
[◦C
]
0 5 10 15 20 25 30−1000
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
PIHybrid
Time [days]Energ
yflow
[W]
Control strategy PI Basic CZT Hybrid
Thermal comfort [%] 87.5 89.2 84.1 88.2Energy consumption [kWh] 753.0 722.7 684.0 640.1Energy savings [%] - 4.0 9.1 15.0
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 16 / 18
Conclusions - Comfort vs Energy Consumption
84 85 86 87 88 89 90
640
660
680
700
720
740
760
Thermal comfort [%]
Ene
rgy
cons
umpt
ion
[kW
h]
PI
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 17 / 18
Conclusions - Comfort vs Energy Consumption
84 85 86 87 88 89 90
640
660
680
700
720
740
760
Thermal comfort [%]
Ene
rgy
cons
umpt
ion
[kW
h]
PIBasic
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 17 / 18
Conclusions - Comfort vs Energy Consumption
84 85 86 87 88 89 90
640
660
680
700
720
740
760
Thermal comfort [%]
Ene
rgy
cons
umpt
ion
[kW
h]
PIBasicCZT
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 17 / 18
Conclusions - Comfort vs Energy Consumption
84 85 86 87 88 89 90
640
660
680
700
720
740
760
Thermal comfort [%]
Ene
rgy
cons
umpt
ion
[kW
h]
PIBasicCZTHybrid
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 17 / 18
Conclusions
Design and evaluation of different MPC strategies
Best case: hybrid formulation with energy savings up to 15%
Drawback: higher computational complexity
No weather predictions
Suitable for application in ”intelligent“ building
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 18 / 18
Conclusions
Design and evaluation of different MPC strategies
Best case: hybrid formulation with energy savings up to 15%
Drawback: higher computational complexity
No weather predictions
Suitable for application in ”intelligent“ building
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 18 / 18
Conclusions
Design and evaluation of different MPC strategies
Best case: hybrid formulation with energy savings up to 15%
Drawback: higher computational complexity
No weather predictions
Suitable for application in ”intelligent“ building
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 18 / 18
Conclusions
Design and evaluation of different MPC strategies
Best case: hybrid formulation with energy savings up to 15%
Drawback: higher computational complexity
No weather predictions
Suitable for application in ”intelligent“ building
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 18 / 18
Conclusions
Design and evaluation of different MPC strategies
Best case: hybrid formulation with energy savings up to 15%
Drawback: higher computational complexity
No weather predictions
Suitable for application in ”intelligent“ building
J. Drgoňa (STU Bratislava) Building MPC June 20, 2013 18 / 18
Building Modeling and ControlModel Predictive ControlCase Study