predictive building control - opticontrol project · building" meteo" service"...
TRANSCRIPT
Fachveranstaltung «Gesamtheitliche vorausschauende Gebäudeautomation» Allschwil, 20. September 2012
Predictive Building ControlInsights from the Research Perspective
Roy Smith"ETH Zürich,"Physikstrasse 3"8092 Zürich"[email protected]"
2
Institut für Automatik
The Automatic Control Laboratory / Institut für Automatik (IfA) is located within the Department of Information Technology and Electrical Engineering (D-ITET) at the ETH in Zürich."
John Lygeros (Head of the Laboratory)"Heinz Köppl "Manfred Morari"Roy Smith"
We also have: "
We have 4 professors:"
Overview:
• 3 senior researchers;"• 7 post-doctoral researchers;"• 34 Ph.D. students."
3
Institut für Automatik
Research Activities:
Optimization!Optimal control!
Model predictive!control!
Process!control!
Hybrid systems!
Robotic systems!
Formation!control! Kite control &!
aerodynamics!
Control theory!Networked and!cooperative systems!
Building!control!
dx(t)
dt= Ax(t) + B u(t)
y(t) = C x(t) + Du(t)
u(t) = K y(t)
4
Overview
OptiControl II: Model predictive control
• An outline of Model Predictive Control (MPC)"
• Control relevant description of the Actelion demonstrator building"
• Building modeling for MPC"
• Commissioning the MPC system"
• MPC operational experience"
• EnergyPlus simulation comparisons with Rule-Based Control (RBC)"
• Discussion"
5
Model predictive control (MPC)
Principle of operation
Building"Room temps."Stored energy"
Energy use"
Predictive"controller"
TABS, ventilation, blinds"
Building"
Meteo"service"
Comfort constraints"Energy costs"
Occupancy predictions"
Measurements:"
Actuation:"
Disturbances:"
Temperature, radiation forecasts"
Minimize the predicted energy cost"
Actuation within limits"Predicted temperatures within limits"Predicted dynamics of the building"
u(t)
y(t)
w(t)
Predicted Cost = minimizeu(t)
Expected
�t+N�
t
energy cost(t)
�
subject to u(t) ∈ Ux(t) ∈ Xx(t+ 1) = f(x(t), u(t), w(t))
6
Model predictive control (MPC)
Main advantages of MPC
Actuation constraints can be satisfied:" • TABS heating and cooling limits"• Blind operation limits."• Air handling flow and temperature limits"
State variable constraints can be satisfied:" • Room temperatures kept within limits"
Predictions of the disturbances (weather, occupancy) can be used for an optimal strategy."
Constraints can change with time and predicted weather."
Constraints and limits can be changed in a straight-forward way."
7
Model predictive control (MPC)
Principle of operation
TABS/ventilation/blinds"Building"
Room temps."Stored energy" Solar radiation / ambient temp."
Wind"Energy use"
Building model"
TABS/ventilation/blinds"Weather forecasts"Room temps."
Stored energy"Energy use"
Predictions "
Optimize "
Objectives (energy, comfort)"Constraints" }
Time steps"
Optimal future actuation"
At each time step: measure, calculate future optimal actuation setpoints, apply first actuation setpoints"
x(t+ 1) = f(x(t), u(t), w(t))
w(t)
y(t)
u(t)
8
MPC controller operation
Measure:"
Estimate states:"
Calculate control:"
Implement control:"
Predict disturbances:"
Weather forecast: 72 hours, updated every 12 hours"
Prediction horizon: 60 hours (240 time steps ahead)"
Optimal future actuation"
Measure:"
Estimate states:"
Calculate control:"
Implement control:"
Predict disturbances:"
Model predictive control (MPC)
t = 0 t = 15 min.
y(t) y(t)
x(t) x(t)
w(t) w(t)
u(t)
u(t)
u(t)
u(t)
t = 15 min. t = 60 hours
9
Actelion demonstrator building
Instrumentation for MPC
• Room temperatures (2nd floor),"• Outside air temperature,"• Solar radiation, façade illuminance."
Actuation for MPC
• TABS system: (slow actuation)"• Heating mode"• Cooling mode uses tower (effective only at night)"
• Air handling/ventilation system: (fast actuation but low power due to low air-flow)"
• Blinds (limited operation while building is occupied; limited number of positions)"
10
Building modeling for MPC
Models: EnergyPlus: Physics based; construction/material specifications;" 2nd floor (725 sq.m.) divided into 24 zones;" Simulation studies (via BCVTB) possible;" Unsuitable for MPC implementation."
Large-scale RC model: Semi-automated derivation from EP;" 2nd floor thermal model has 294 states;" Too large for MPC implementation;"
Convenient test environment for MPC."
Reduced-order RC model: Hankel-norm reduction of thermal model;" Thermal model: 15 states (with 0.1% error);" Actuation adds 10 more states;" Time constants: 20, 8 and 5 hours. Majority < 30 minutes."
(Gruner)
(ETH)
(ETH)
11
Building modeling for MPC
Modular components
Building geometry,"location and orientation"Construction material data sheets" Actuation components"
Thermal model"
TABS heating model"
Solar radiation"
TABS cooling model"
Ventilation model"
Blinds model"
External temp."
Building"temperatures"
Energy usage"
Heat"fluxes"
12
Commissioning the MPC system
EnergyPlus"model"Room temps."
Stored energy" Solar radiation / ambient temp."(simulated)"
Energy use"
Simplified RC"building model"
TABS/ventilation/blinds"
Weather forecasts"(simulated)"Room temps."
(averaged)"
Stored energy"Energy use"
Optimization"
Random disturbances"
MPC controller commissioning (initial verification)
Matlab environment (periodic operation)"
BCVTB interface"
Energy Plus environment"
Run-time"data storage"
13
MPC controller run in shadow mode to confirm operation."
Sensor failure and bad data tests conducted."
Matlab failure test conducted."
MPC controller implementation:"
- TABs heating implemented: 21 April 2012."
- TABs cooling & blind control implemented: 10 May 2012" - AHU temperature control implemented: 10 June 2012."
MPC controller commissioning (building tests)
Commissioning the MPC system
14
Building"Room temps."Stored energy" Solar radiation / ambient temp."
Energy use"
Simplified RC"building model"
TABS/ventilation/blinds"
Weather forecasts"(MeteoSwiss)"Room temps."
(averaged)"
Stored energy"Energy use"
Optimization"
Occupancy/internal gains"
Matlab environment (periodic operation)"
OPC interface"
Run-time"data storage"
Model Predictive Control implementation
Building"
15
MPC implementation: Kalman filter based
Kalman"filter"
Building "measurements"
Prediction:"low order
model"
Weather "predictions"
TABS/ventilation/blinds"
Room temps."(averaged)"
Stored energy"Energy use"
Constrained"Energy
Optimization"
Estimated states"Actuation"
Kalman"filter"
y(t)
x(t)
w(t)
u(t)
16
Weather prediction filtering Filter dynamics are based on OptiControl I analyses.
Model Predictive Control operation
!"#$%&'%%(%% !"#$)%'%%(%% !"#$))'%%(%% !"#$)*'%%(%% !"#$)+'%%(%%)*
),
)-
).
*%
**
*,/"01234'526'748946:0"64';34<6441'=>
'
'
?2@04643'A6432B02C#
D4:1"6484#0
EF:GH'A6432B02C#
="664#0'0284
!"#$%&'(()((!"#$%&'%*)((
!"#$%+'(()((!"#$%+'%*)((
!"#$*('(()((!"#$*('%*)((
!"#$*%'(()((!"#$*%'%*)((
!"#$**'(()((!"#$**'%*)((
!"#$*,'(()(((
-(
%((
%-(
*((
*-(
,((
'
'
!"#$%&'(()(( !"#$%+'(()(( !"#$*('(()(( !"#$*%'(()(( !"#$**'(()(( !"#$*,'(()((%(
%-
*(
*-
,(
,-./01/234"2/'56/789:;<32'=/34><"?'5@A0*9=BC4;2BD3< E2/6BD4B;#
F3?B0"0':;<32'=/34><"?''G:HI'5@A0*9
J0KB/#4'./01/234"2/'56/789
F3?B0"0':;<32'=/34><"?''GL@I'5@A0*9
17
Disturbances: measured and filtered predictions
17
Model Predictive Control operation
w(t)
18
Actuation: TABS cooling and blinds Blind actuation uncertainty: - based on predicted solar radiation; - limited to only four possible positions; - occupants may over-ride blind settings.
TABS cooling uncertainty: - based on predicted overnight temperatures.
18
Model Predictive Control operation
u(t)
!"#$%&'(()(( !"#$%*'(()(( !"#$+('(()(( !"#$+%'(()(( !"#$++'(()(( !"#$+,'(()((
(
-(
%((
%-(
+((
+-(
,((
!"#$%&'(()((
!"#$%&'%+)((
!"#$%*'(()((
!"#$%*'%+)((
!"#$+('(()((
!"#$+('%+)((
!"#$+%'(()((
!"#$+%'%+)((
!"#$++'(()((!"#$++'%+)((
!"#$+,'(()((
(
-
%(
%-
./012'341560"7'89:;+<3=>5/2=?10
@//0=#A'B/C42'8D9<B24E=?5=/#
./012'341560"7F.GH'89:;+<
./012'341560"7'FI9H'89:;+<
JKL.'@//0=#AB/C42'8D9<
19
Controlled variables: room temperatures
Model Predictive Control operation
y(t)
18
20
22
24
26
28
30
18
20
22
24
26
28
30
Historical Prediction
Average Room Temperature (SE)Average Room Temperature (NW)
Temperature [deg C]
Jun.18 00:00Jun.18 12:00
Jun.19 00:00Jun.19 12:00
Jun.20 00:00Jun.20 12:00
Jun.21 00:00Jun.21 12:00
Jun.22 00:00Jun.22 12:00
Jun.23 00:00
Temperature [deg C]
Comfort constraint
Comfort constraint
20
Performance: room temperatures (50 days) TABS heating was required on 18 May.
Model Predictive Control operation
!"#$%& !"#$%' !"#$&& !"#$&' !"#$(& !"#$(' !"#$)& *+,$%- *+,$&% *+,$&- *+,$(%(%
(&
((
()
(.
(-
('
(/
(0
123425"6+52$7829:;
<=25"92$!2">+528$?@@3$123425"6+52
!"AB3+3$",8$!B,B3+3$!2">+528$?@@3$123425"6+52
1"5926$:@3C@56$?",92
?@@3$123425"6+52>$7829522>$:;
21
MPC controller operation: 21 April 2012 to 31 July 2012."
Model Predictive Control operation
Cooling operations:"
TABS" Ventilation" TABS &"Ventilation"
Heating operations:"
Blinds"
(limited)"
Heating operations to be tuned and tested in November and December 2012.
Operational modes demonstrated:
22
MPC simulation comparisons using EnergyPlus
MPC simulation studies
EnergyPlus"model"Room temperatures"
Radiation measurements" Solar radiation / ambient temp."(simulated with 2010 data)"
Energy use"
MPC "controller"
Random disturbances"
BCVTB interface"
TABS/ventilation/blinds"
Energy Plus environment"
Matlab environment"
23
MPC simulation comparisons using EnergyPlus
MPC simulation studies
MPC controller uses too"little TABS heating and too much TABS cooling."
(work in progress)
RBC 0 RBC 1 RBC 2 MPC0
20
40
60
80
100
120
140 Annual energy use [MWh]
TABS HeatingStatic HeatingVentilation HeatingTABS CoolingVentilation Cooling
Electrical pumps
Electrical Equipment Offices
Electrical Lighting Offices
Electrical Lighting/Equipment
24
MPC simulation comparisons using EnergyPlus
MPC simulation studies
MPC controller requires"tuning to increase"temperatures."
(work in progress)
This will be done on the building in November & December 2012."
RBC 0 RBC 1 RBC 2 MPC0
50
100
150
200
250
300
350
400
450
Too Cool
Too Warm(Low Outside Air Temp.)Too Warm
(High Outside Air Temp.)
Annual comfort violations [Kelvin-hours]
25
Discussion
MPC controller has been successfully demonstrated on the building during"Spring/Summer period."
Initial models can be obtained from EnergyPlus models or building construction data sheets."
Models were refined by simulation-based regression analysis using EnergyPlus."
Tuning of the objectives and constraints is simple and straightforward."
Model development and refinement is the most time-consuming aspect of the control design."
Future work will look at automatic or simplified model refinement based on"building operation data."
26
Acknowledgments
OptiControl I: IfA (ETH): F. Oldewurtel, C. Jones, A. Parisio, F. Ullmann, T. Baltensberger, A. Domahidi, M. Morari TSE (ETH): D. Gyalistras, A. Fischlin EMPA: T. Frank, B. Lehmann, K. Wirth, V. Dorer, S. Carl MeteoSwiss: P. Steiner, F. Schubiger, V. Stauch
Siemens: J. Tödtli, M. Gwerder, B. Illi, C. Gähler Gruner: A. Seerig, C. Sagerschnig
OptiControl II: IfA (ETH): D. Gyalistras, D. Sturzenegger, M. Morari, R. Smith Siemens: D. Habermacher, M. Gwerder, B. Illi, M.Winter, M. Hubacher, S. Bötschi Gruner: C. Sagerschnig, A. Seerig Actelion Pharmaceuticals: A. Gaiser, G. Maltese Funding support:
swisselectric research Siemens Gruner ETH
Actelion Pharmaceuticals.
27
Further information...
F. Oldewurtel, A. Parisio, C.N. Jones, D. Gyalistras, M. Gwerder, V. Stauch, B. Lehmann, M. Morari, Use of Model Predictive Control and Weather Predictions for Energy Efficient Building Climate Control. Energy and Buildings, vol. 45, p. 15-27, Elsevier, 2012.
F. Oldewurtel, C.N. Jones, A. Parisio, M. Morari, Stochastic Model Predictive Control for Energy Efficient Building Climate Control, Submitted for IEEE Transaction on Control Systems Technology, 2012.
J. Siroky, F. Oldewurtel, J. Cigler, S. Privara, Experimental Analysis of Model Predictive Control for an Energy Efficient Building Heating System, Applied Energy, vol. 88, pp. 3079-3087, 2011.
F. Oldewurtel, A. Ulbig, G. Andersson, M. Morari, Building Control and Storage Management with Dynamic Tariffs for Shaping Demand Response in Electricity Grids, Proc. of IEEE PES Conference on Innovative Smart Grid Technologies Europe, Manchester, UK, 2011.
F. Oldewurtel, A. Ulbig, A. Parisio, G. Andersson, M. Morari, Reducing Peak Electricity Demand in Building Climate Control using Real-Time Pricing and Model Predictive Control, Proc. of 49th IEEE Conference on Decision and Control, CDC, Atlanta, USA, 2010.
R. Gondhalekar, F. Oldewurtel, C. N. Jones, Least-Restrictive Robust MPC of Periodic Affine Systems with Application to Building Climate Control, Proc. of 49th IEEE Conference on Decision and Control, CDC, Atlanta, USA, 2010.
F. Oldewurtel, A. Parisio, C.N. Jones, M. Morari, D. Gyalistras, M. Gwerder, V. Stauch, B. Lehmann, K. Wirth, Energy Efficient Building Climate Control using Stochastic Model Predictive Control and Weather Predictions, Proc. of American Control Conference, Baltimore, USA, 2010.
F. Oldewurtel, D. Gyalistras, M. Gwerder, C.N. Jones, A. Parisio, V. Stauch, B. Lehmann, M. Morari, Increasing Energy Efficiency in Building Climate Control using Weather Forecasts and Model Predictive Control, Proc. of Clima - RHEVA World Congress, Antalya, Turkey, 2010.
M. Gwerder, D. Gyalistras, F. Oldewurtel, B. Lehmann, K. Wirth, V. Stauch, J. Tödtli, Potential Assessment of Rule-Based Control for Integrated Room Automation, Proc. of Clima - RHEVA World Congress, Antalya, Turkey, 2010.
F. Oldewurtel, C.N. Jones, M. Morari, A Tractable Approximation of Chance Constrained Stochastic MPC based on Affine Disturbance Feedback, Proc. of 47th IEEE Conference on Decision and Control, CDC, Cancun, Mexico, 2008.
References:
Institut für Automatik:
OptiControl: http://www.opticontrol.ethz.ch/""
http://control.ee.ethz.ch/"