Experimental Demonstration of Model Predictive Control in a - PowerPoint PPT Presentation

3 rd International High Performance Buildings Conference at Purdue, West Lafayette, IN. Experimental Demonstration of Model Predictive Control in a Medium-Sized Commercial Building Pengfei Li, Dapeng Li, Draguna Vrabie, Sorin Bengea, Stevo Mijanovic Presented by Pengfei Li United Technologies Research Center, USA July 14 th , 2014

Advanced Building HVAC Controls Demonstration Outline  Motivation  Overview of Building 101 HVAC system  Control-oriented HVAC and zone model development  Optimization-based controls design and problem formulation  Advanced control deployment toolchain  MPC demonstration results and analysis  Summary 2

Advanced Building HVAC Controls Demonstration Motivation Building Specification Installation Startup Commissioning occupation System diagram • More than 70% cost of implementation of advanced Engineering & Commissioning controls is involved in installation & commissioning 20% Project Mgmt • Our goals: & Installation Other Material 50% • Reduce deployment and commissioning time & Warranty 18% (reduce cost) DDC Controls • Save energy 12% • Maintain occupant’s comfort 3

Advanced Building HVAC Controls Demonstration Overview of Building 101 HVAC system AHU3 schematic Primary system: Air-cooled chiller Secondary system: AHU w. DX evaporator Terminal system: VAVs w. reheat coil Zones served by AHU3 Individual zone temp., RH and CO2 sensors (AHU3) 4

Advanced Building HVAC Controls Demonstration Model Predictive Control for building HVAC supervisory control • Supervisory (centralized): optimal control MPC architecture Basic idea : receding prediction horizon • Local: set point tracking past future Sequence of steps for on-line implementation Predicted outputs Sensor Initialize Manipulated u(t+k) measurements thermal states Inputs u*(t) t+N t t+1 t+m Weather Generate load Receding forecasts forecasts horizon Generate optimized set points u*(t+1) Communicate set points to t+1 t+2 t+1+m t+1+N field controllers Source: https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxtcGNsYWJvcmF0 b3J5fGd4OjYwMWNhMWE4OTBiOTYxYjI 5

Advanced Building HVAC Controls Demonstration System identification for dynamic thermal zone model � � � 1 � �� � � �� � � � � ���� � � ��� � �� � � ����� Validation data vs. model (low-order state-space model) Zone 5 Zone 6 82 79 Data Data Model Model 78 80 77 78 76 Zone Temp. ( °F) Zone Temp. (°F) 76 75 74 74 73 72 72 71 70 70 68 100 200 300 400 500 600 700 800 900 100 200 300 400 500 600 700 800 900 Samples (3 min.) Samples (3 min.) 6

Advanced Building HVAC Controls Demonstration Data-driven control-oriented HVAC model ��� � � � ��� � � � � ��� � � � AHU mixing box � ��� � ���� �� � �1 � ����� �� DX compressor � �� � � � �� � �� �� � ��� �� � �� � � �� � �� � , � � �� � �� � � � ��� � � �� power � ��� � � � �� � � � � � � AHU airflow � ��� ��� � � � � � � � � � � � ��� � � � � � �� � � � AHU fan � �� � �� ��� � � � � �� � �� � � � ���,��� � � �� � � � ��� � �� � VAV reheat coil � �� � � � � � �� � �� � ��� � � � � �� �� ������ Fan power validation Air flow validation Fan Power (Validation Data) 18 16 14 Measured Fan Power (kW) 12 10 8 6 4 2 0 0 5 10 15 20 7 Calculated Fan Power (kW)

Advanced Building HVAC Controls Demonstration Problem formulation Objective: minimize energy + satisfy comfort Energy  n P ( m , T ) 1  VAV    cooling DA DA   P aP ( m ) b c P ( m , T )  total fan sa reheating sa sai COP  i 1 DX boiler • Soft constraints on zone temperature lower and upper bound • Constraints on control authority   T T s z , i z , max k   Comfort T T s z , min z , i k  s 0 k Slack variables 8

Advanced Building HVAC Controls Demonstration Control architecture and state estimation   T k ( ) ˆ ˆ x k ˆ( ) T k ( ) ... T k ( N 1) z z z * ( ) u k Kalman Filter for state estimation at each time step A-priori state estimate : � � �|��� � � � � ���|��� � �� � A-priori error covariance : � �|��� � ���� � 1�� � +Q � � � � A-priori output estimation error: � � � � � � � �|��� � �|��� � � � � Residual covariance: � � � �� �� �|��� � � � � Optimal Kalman gain: � � � � � � � A-posteriori state estimation: � ���� � � � �|��� � � � � A-posteriori estimation error covariance : ���� � �� � � � ��� �|��� 9

Advanced Building HVAC Controls Demonstration Optimization-based control development & deployment toolchain Rapid advanced control deployment toolchain Optimization solver: IPOPT WebCTRL 10

Advanced Building HVAC Controls Demonstration Similar OAT patterns MPC demonstrates improved thermal comfort 85 Baseline (06/13/2013) Outdoor Air Temp. (ºF) Zone Temp. (Heuristic-Based Baseline) 80 78 Cooling Setpoint 75 MPC (09/13/2013) 76 70 Heuristic-based 74 65 ºF baseline 0 50 100 150 200 Samples (3 min.) 72 Heating Setpoint 70 09:00 12:00 15:00 Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Z9 Z10 Zone Temp. (MPC) 78 Cooling Setpoint 76 MPC ºF 74 72 Heating Setpoint 70 09:00 12:00 15:00 Time (occupied hours) 11

Advanced Building HVAC Controls Demonstration MPC reduced compressor power with penalty of more fan power Less compressor power Compressor Power (kW) 100 50 0 09:00 12:00 15:00 Fan power (kW) Baseline More fan power 20 MPC 10 0 09:00 12:00 15:00 Total power is reduced! Compressor + Fan power (kW) 100 50 0 09:00 12:00 15:00 Time (occupied hours) 12

Advanced Building HVAC Controls Demonstration MPC performance analysis based on similar OAT patterns Similar OAT patterns Energy Consumption Reductions from MPC (%) 85 Baseline (06/13/2013) Outdoor Air Temp. (ºF) 70 80 60 75 MPC (09/13/2013) Energy Consumption Reductions (%) 70 50 65 0 50 100 150 200 40 ~33% Samples (3 min.) 30 20 10 0 -10 -20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Test Days 13

Advanced Building HVAC Controls Demonstration Summary  MPC implemented on a medium-size building HVAC system with chiller, AHU and VAV boxes and demonstrated its benefits in energy saving and thermal comfort.  Data-driven control-oriented HVAC model and low-order state space model suitable for MPC execution and their effectiveness demonstrated  Future work towards statistics-based performance analysis method 14

Acknowledgement Acknowledgement This work is funded by Consortium for Building Energy Innovation (formally known as Energy Efficient Buildings Hub), sponsored by the Department of Energy under Award Number DE- EE0004261. The authors are grateful to Hayden Reeve for managing the demonstration activities as well as his technical insights and valuable inputs to improve the paper and Timothy Wagner for his project management and supervision. We also thank Ken Kozma from Radius Systems for his technical support on WebCTRL during our advanced controls demonstration. The authors also thank our collaborators from Purdue University and Virginia Tech within the subtask 4.2.  2 15