Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in...

17
Tony Kelman MPC in Buildings January 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering Email: [email protected]

Transcript of Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in...

Page 1: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 1

Overview of Model PredictiveControl in Buildings

Tony KelmanMPC Lab, Berkeley Mechanical Engineering

Email: [email protected]

Page 2: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 2

OutlineModel predictive control

Basic idea and elementsAdvantages, disadvantages

Modeling and MPC in buildingsWhat works, what doesn’tGenerating models using historical dataAdvanced control behavior

Experimental projectsSuccess storiesIncreased scope and capabilities over time

Page 3: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 3

System model – state evolution vs inputs and disturbancesConstraints on inputs or states – requirements, actuator limitsCost function – reference tracking, energy, comfortForecast trajectories of future disturbance inputs – weather, occupancy, utility ratesOptimization algorithm – fast enough to solve in real time

Use predictive knowledge for controlBasic components

System modelConstraints on inputs or statesCost functionForecast trajectories of future disturbance inputs

Optimization algorithm

Advantages: multivariable, model based, nonlinear, constraint satisfaction, incorporates predictionsDisadvantages: computational complexity, design effort of accurate modeling

Model Predictive Control

Page 4: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 4

Model Predictive Control

Initialize with current measurements at time tPredict response over horizon of p stepsSolve for best input sequence, apply first element u*(t)Repeat at time t+1 with new measurements (feedback)

Page 5: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 5

Optimization FormulationPredicted states xk, inputs uk, disturbances wk

At each time step, solve:

Constrained finite time optimal control problemOptimization much faster if explicit structure of J, f, g (and derivatives) can be provided

Page 6: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 6

OutlineModel predictive control

Basic idea and elementsAdvantages, disadvantages

Modeling and MPC in buildingsWhat works, what doesn’tGenerating models using historical dataAdvanced control behavior

Experimental projectsSuccess storiesIncreased scope and capabilities over time

Page 7: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 7

Modeling for Building Energy SystemsCommon practice is black-box simulation

DOE2, EnergyPlus, TRNSYS, etcUseful for design, very difficult to use for controlDerivative-free optimization not very efficient or scalable

Need model structure for optimization and controlSimpler approach: reduced order modeling

Physics based model structureData driven parameter identificationCan adjust accuracy vs complexity tradeoffLarge scale real time optimization tractable

Page 8: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 8

HVAC good target for energy savings by better controlCommon configuration for commercial buildings:VAV with reheat

Control inputs: supply fan, cooling coil, heating coils, zone dampers, air handling unit dampersStates: zone temperatures

HVAC Example System

Page 9: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 9

Thermal Zone Model

Page 10: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 10

Network of bilinear systems

A(simple extension to multiple statesper zone, RC network analog)

Simplest Useful Model Abstraction

u1_Q1

_Q2u2

un_Qn

Thermal zone model

Static nonlinearities Equipment performance maps (chillers, cooling towers, pumps, fans, coils)

Equality and inequality constraints Comfort range Dynamic coupling: thermal zones, supply air & return air

Uncertain load predictions Human: occupancy, thermal comfort, … Environment: ambient temperature, solar radiation, …

Page 11: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 11

How to Generate Reduced ModelsSeveral options to create model data

Direct physics based lumped parametersModel reduction from high fidelity design toolsUse historical data for model identification

Identification results vs measured data, Bancroft library

Page 12: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 12

Using Data to Quantify Uncertainty

Ambient temperature Load

SMPCPrediction model

Historical loadrealization

Page 13: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 13

Advanced Control Behavior

MPC is able to incorporate time-varying energy price and reduce peak power consumption

Time-varying price Penalize peak power

A. Kelman, Y. Ma, A. Daly, F. Borrelli, Predictive Control for Energy Efficient Buildings with Thermal Storage: Modeling, Stimulation, and Experiments, IEEE Control System Magazine, 32(1), page 44-64, February 2012.

Page 14: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 14

OutlineModel predictive control

Basic idea and elementsAdvantages, disadvantages

Modeling and MPC in buildingsWhat works, what doesn’tGenerating models using historical dataAdvanced control behavior

Experimental projectsSuccess storiesIncreased scope and capabilities over time

Page 15: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 15

UC Merced –, Merced, CA 4% Improvement

LBNL+UTRC- Storage, Chiller OptimizationHorizon 24hrs, Sampling 30minProblem Size: ~300 variables , ~1440 constraints

CERL Engineering Research Laboratory, Champaign, IL 15% improvement. UTRC- HVAC distribution – 5 zones

Horizon 4hrs, Sampling 20 min,Problem Size: ~1600 variables , ~1400 constraints

Naval Station Great Lakes, North Chicago, IllinoisUTRC- Conversion + Storage – 250 zonesProblem Size: ~~20k variables , ~?? constraints

CITRIS Building (UC Berkeley) – Major issues Siemens - Generation + HVAC distribution -135 Zones

Horizon 4hrs, Sampling 20 min, Problem Size: ~~10k variables , ~?? constraints

Brower Center (Slab Radiant), Berkeley, CA Architecture Department Models based on step tests experiments

White Oak, Silver Spring, MD Honeywell

Microgrid Optimization

Experimental Projects

Simplified models and BLOM tool critical for real-

time implementation of large MPC experiments

Page 16: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 16

Distributed Implementation

Page 17: Tony KelmanMPC in BuildingsJanuary 11 th – Slide 1 Overview of Model Predictive Control in Buildings Tony Kelman MPC Lab, Berkeley Mechanical Engineering.

Tony Kelman MPC in Buildings January 11th – Slide 17

Distributed Implementation

Supply FanCooling coil

damper

Supply FanCooling coil

damper

Heating coilVAV damperHeating coilVAV damper

Zone temperature

Zone temperature

CoordinatorDual variablesCoordinator

Dual variables