Modeling and control in air conditioning sysytems

The model-basedpredictive control and the state-space feedback control are introduced to theair-conditioning systems for a better local control, and the air-side synergic controlscheme a

Energy and Environment Research in China

Ye Yao

Yuebin Yu

Modeling and Control in Air- conditioning Systems

Tai ngay!!! Ban co the xoa dong chu nay!!!

More information about this series at http://www.springer.com/series/11888

Ye Yao Yuebin Yu

Modeling and Control

in Air-conditioning Systems

123

Energy and Environment Research in China

DOI 10.1007/978-3-662-53313-0

Jointly published with Shanghai Jiao Tong University Press, Shanghai, China

Library of Congress Control Number: 2016948282

© Shanghai Jiao Tong University Press and Springer-Verlag GmbH Germany 2017

This work is subject to copyright All rights are reserved by the Publishers, whether the whole or part

of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on micro films or in any other physical way, and transmission

or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a speci fic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.

The publishers, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publishers nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer-Verlag GmbH Germany

The registered company address is: Heidelberger Platz 3, 14197 Berlin, Germany

With the global warming and the rapid improvement of people’s living standards,energy consumption by air-conditioning (AC) systems in buildings is on the rise.According to the US Energy Information Administration (EIA) and the USDepartment of Energy, the consumption of electrical energy by HVAC (heating,ventilation, and air-conditioning) systems in the residential, commercial, andindustrial sectors corresponds to 18.62 %, 16.20 %, and 2.34 % of the total elec-trical energy consumed in the USA, respectively (totalizing 37.16 %) In China,building sector accounted for 23.4 % and 28 % of total energy use in 2011 and

2012, respectively, and about half of total building energy is consumed by HVACsystems Thus, energy conservation in HVAC systems will play an important role

in search of solutions to meet the growing global energy demand Any

effective models based on which the high-performance HVAC systems and optimal

This book mainly concerns about modeling and control in air-conditioningsystems Some advanced modeling methods including state-space method,graph-theory method, and structure-matrix method, as well as combined forecastingmethod, are employed for the modeling of air-conditioning systems The virtualsensor calibration and virtual sensing methods (which will be very useful for thereal system control) are illustrated together with the case study The model-basedpredictive control and the state-space feedback control are introduced to theair-conditioning systems for a better local control, and the air-side synergic controlscheme and the global optimization strategy with the decomposition-coordinationmethod are developed aiming at energy conservation of the entire system Lastly,control strategies for VAV systems including the total air volume control and thetrim-and-response static pressure control are investigated with practice The bookcomprises ten chapters that are summarized as below:

of the topic related to this book, gives a literature overview about modeling

v

Chapter2(written by Dr Ye Yao) illustrates in detail the modeling process forHVAC components and system with the state-space modeling method.

responses of HVAC components with the state-space models under different turbations and initial conditions

of graph-theory approach for modeling HVAC components and system, andintroduces the structure-matrix analysis method to study control characteristics ofHVAC state-space models

sensor calibration and virtual sensing methods

based on the state-space model

air-conditioning load prediction The two original forecasting models based on thecombined principle are introduced

com-ponents based on which the energy analysis program is developed and used for theenergy analysis on variable-air-volume (VAV) air-conditioning systems

of HVAC system aiming at energy conservation

modular modeling, control strategies, and sequences as well as test script for VAVsystem

Acknowledgement

National Nature Science Foundations (No 50708057; No 51110105012)

June 2016

1 Introduction 1

1.1 Background 1

1.2 Modeling Approaches in HVAC Field 2

1.2.1 Physics-Based Modeling Approach 2

1.2.2 Data-Driven Modeling Approach 7

1.2.3 Hybrid Modeling Approach 10

1.3 Proposed Methods 11

1.3.1 State-Space Modeling 11

1.3.2 Graph-Theory Modeling 11

1.3.3 Combined Forecasting Modeling 12

1.3.4 Decomposition–Coordination Algorithm for Global Optimization Model 13

1.3.5 Virtual Calibration for HVAC Sensors 14

1.3.6 Model-Based Predictive Control (MPC) 17

1.4 Organization of This Book 18

References 21

2 Component Modeling with State-Space Method 29

2.1 Basic Knowledge About State-Space Modeling Method 29

2.2 Modeling for HVAC Components 30

2.2.1 Water-to-Air Heat Exchanger 30

2.2.2 Chiller 44

2.2.3 Cooling Tower 60

2.2.4 Duct (Pipe) and Fan (Pump) 71

2.2.5 Air-Conditioned Room Modeling 85

2.3 Modeling for HVAC System 96

2.3.1 Component Model Connection 96

2.3.2 State-Space Representation for HVAC System 100

2.3.3 Case Study 103

References 108

vii

3 Dynamic Simulations with State-Space Models 109

3.1 On Water-to-Air Surface Heat Exchanger 109

3.1.1 Subjected to Different Perturbations 109

3.1.2 For Different Initial Conditions 113

3.2 On Chiller 124

3.2.1 Subjected to Different Perturbations 124

3.2.2 For Different Initial Conditions 128

3.3 On Cooling Tower 141

3.3.1 Subjected to Different Perturbations 141

3.3.2 For Different Initial Conditions 143

3.4 On Duct and Pipe 147

3.4.1 On Straight-Through Duct 147

3.4.2 On Straight-Through Pipe 150

3.5 On Air-Conditioned Room 152

3.5.1 Basic Conditions 152

3.5.2 Subjected to Different Perturbations 152

4 Graph-Theory Modeling and Structure-Matrix Analysis 159

4.1 Graph-Theory Modeling for HVAC Component State-Space Models 159

4.1.1 Fundamental Rules 159

4.1.2 Case Study 160

4.2 Graph-Theory Modeling for HVAC System 172

4.2.1 Basic Method 172

4.2.2 Case Study 173

4.3 Structure-Matrix Analysis Approach 176

4.3.1 Model Structural Matrix 176

4.3.2 Reachability Analysis of Model Input–Output 176

4.3.3 Controllability/Observability Analysis of Model 178

4.3.4 Case Study 180

References 188

5 Virtual Measurement Modeling 189

5.1 Virtual Calibration 189

5.1.1 Conventional Calibration 189

5.1.2 Methodology of Virtual In Situ Calibration 192

5.1.3 Case Study 200

5.2 Virtual Sensing 203

5.2.1 Development Methodology for Virtual Sensing 204

5.2.2 Case Study 207

5.2.3 Model Development 210

References 218

6 Control Design Based on State-Space Model 221

6.1 Model-Based Predictive Control (MPC) 221

6.1.1 Introduction of MPC 221

6.1.2 MPC in Broad Definition 222

6.2 Applications of MPC in HVAC Field 229

6.2.1 Control of a Hybrid Ventilation Unit 229

6.2.2 Control of Space Thermal Conditioning 266

6.3 State-Space Feedback Control System Design 285

6.3.1 Basic Principle 285

6.3.2 Control System Design for Water-to-Air Heat Exchanger 287

6.3.3 MATLAB Simulation of the Control System 289

6.3.4 Control System Design for Refrigeration System 291

References 295

7 Combined Forecasting Models for Air-Conditioning Load Prediction 297

7.1 Typical Methods 297

7.1.1 MLR Modeling 297

7.1.2 ARIMA Modeling 299

7.1.3 GM Modeling 301

7.1.4 ANN Modeling 302

7.2 Combined Forecasting Model Based on Analytic Hierarchy Process (AHP) 304

7.2.1 Principles of the Combined Forecasting Method 304

7.2.2 Determining Weights by Analytic Hierarchy Process (AHP) 305

7.2.3 Combined Forecasting Model for Hourly Cooling Load Prediction Using AHP 308

7.3 Forecasting Model Based on Neural Network and Combined Residual Error Correction 316

7.3.1 Model Development 316

7.3.2 Case Study 323

References 327

8 Energy Analysis Model for HVAC System 329

8.1 Energy Models for HVAC Components 329

8.1.1 Chiller 329

8.1.2 Boiler 331

8.1.3 Pump and Fan 332

8.1.4 Cooling Tower 332

8.1.5 Water-to-Air Heat Exchanger 333

8.2 Energy-Saving Analysis on VAV Air-Conditioning System 335

8.2.1 Evaluation Program for Energy Saving

of VAV System 336

8.2.2 Case Study 339

8.3 Energy Analysis on VAV Air-Conditioning System with Different Air-Side Economizers 346

8.3.1 Scheme for Air Economizer Cycle [27] 347

8.3.2 Case Study 351

References 356

9 Optimal Control of HVAC System Aiming at Energy Conservation 359

9.1 Air-Side Synergic Control 359

9.1.1 Background and Basic Idea 359

9.1.2 Mathematic Deduction of Synergic Control Model 361

9.1.3 Control Logic Details 373

9.1.4 Case Study 376

9.2 Global Optimization Control 387

9.2.1 Model Development 387

9.2.2 Decomposition–Coordination Algorithm for Model Solution 393

9.2.3 Case Study 399

Appendix 413

References 420

10 Modeling and Control Strategies for VAV Systems 423

10.1 Background and Research Status 423

10.2 Modular Modeling with Simulink Tool 429

10.3 Model Library for Components of VAV System 432

10.3.1 VAV Terminal Unit 432

10.3.2 Variable Speed Fan 434

10.3.3 Air Ducts 436

10.3.4 Other Local Resistance Components 444

10.3.5 Application of Component Model Library: Case Study 445

10.4 Control Strategies for VAV System 449

10.4.1 Constant Static Pressure Method 450

10.4.2 Total Air Volume Method 453

10.4.3 Variable Static Pressure Method Based on Trim-and-Respond Logic 458

10.5 Control Sequences for VAV System with Different Terminal Units 464

10.5.1 For Cooling-Only Terminal Unit 464

10.5.2 For Reheat Terminal Unit 465

10.5.3 For Series Fan-Powered Terminal Unit 467

10.6 Test Script for VAV Control Study 46810.6.1 Preparation 468

System 46910.6.3 Trend Data Review 473References 477

About the Authors

Shanghai Jiao Tong University, China He received his Ph.D from Shanghai JiaoTong University (SJTU), China He was promoted as Associate Professor of SJTU

in December 2008 From September 1, 2009 to September 1, 2010, he performedhis research work in Ray W Herrick Lab at Purdue University (PU), USA He wasawarded as Excellent Reserve Youth Talent and SMC Excellent Young Faculty bySJTU, respectively, in the year 2009 and 2015, and got Shanghai Pujiang ScholarsTalent Program in the year 2012 His current research interests mainly include(a) HVAC modeling and optimal control for energy conservation; (b) Heat andmass transfer enhancement assisted by ultrasound He has successfully published

30 Chinese patents He is now the peer reviewer of many international academic

‘Building and Environment’, ‘Energy and Buildings’, and ‘Applied Energy’

Engineering and Construction at University of Nebraska-Lincoln, USA He receivedhis Ph.D degree in Building Performance and Diagnostics from Carnegie Mellon

including (a) smart building technology, including automated continuous sioning and advanced controls, automated fault detection and diagnosis, virtualsensing and virtual calibration; (b) active utilization of renewable energy for heating,ventilation and air-conditioning, including low-grade energy, solar and geothermalthermal energy, active phase change material, bionic building enclosure; and(c) built environment modeling and evaluation At UNL, he maintains astate-of-the-art laboratory with well-instrumented facilities and advanced web-basedAFDD platform for smart buildings and advanced building envelope studies He is

commis-an active commis-and voting member in the Technical Committee TC7.5 for Smart Building

xiii

Systems and serves as the sub-committee chair of Fault Detection and Diagnostics inASHRAE He participated in the revision of ASHRAE Handbooks on FaultDetection and Diagnostics and Energy Estimating and Modeling Methods He haspublished about 50 academic publications.

xv

PMV Predicted mean vote

(m3/s) (or m3/h)

xvii

q Latent heat of condensation (J/kg); or irreversible loss rate (W); or solar

radiation (W/m2); or heatflux (W/m2)

deviation

Greek Letters

/m Power factor of the motor

r2

Subscripts

env Environment

wall; or interior region

rew Outer wall of room

air-conditioning systems are increasingly popular for the improvement of thermalenvironment indoors, which results in ever-increasing energy consumption ofbuildings The need for rational energy use is a global concern, and challenges are

ventilation and air-conditioning) systems in the residential, commercial, andindustrial sectors corresponds to 18.62, 16.20, and 2.34 % of the total electricalenergy consumed in the USA, respectively (totalizing 37.16 %) In China, buildingsector accounted for 23.4 and 28 % of total energy use in 2011 and 2012, respec-tively, and about half of the total building energy is consumed by HVAC systems

search for solutions to meet the growing global energy demand

con-sumption reduction require effective models based on which the relevant software

developed and assist us to design high-performance HVAC systems and optimal

white box) approach which uses detailed physics-based equations to model HVAC

performances The second is data-driven (or black box) approach, viz that the

between the input and output variables using some mathematical techniques, e.g.,

on-site measurements over a period of time In the third type, known as hybrid

© Shanghai Jiao Tong University Press and Springer-Verlag GmbH Germany 2017

Y Yao and Y Yu, Modeling and Control in Air-conditioning Systems,

Energy and Environment Research in China, DOI 10.1007/978-3-662-53313-0_1

1

approach, the basic structure of the model is formed with the physics-based methodsand the model parameters are determined by using the parameter estimation algo-rithms on the measured data of the system In order to build the physics-basedmodels and to determine their parameters, the detailed knowledge about the systemand its working process is needed Physics-based models have good generalizationcapability but suffer from poor accuracy, while the data-driven models have rela-tively high accuracy but suffer from generalization beyond the training domain.Hybrid models combine the advantages of the two, providing good generalizationcapability as compared to the data-driven models and better accuracy as compared tothe physics-based models In the following sections, the state-of-the-art modeling in

Physics-based approach is mostly applied to the modeling of HVAC components.HVAC components include primary and secondary The primary system mainlyincludes chillers, boiler, cooling towers, and liquid distribution system; and thesecondary system includes air-handling equipment, air-distribution system, andliquid distribution system between the primary system and the building interior Thedistribution components are pumps, fans, dampers, valves, ducts, and pipes They

distribution components [7]

A chiller unit basically consists of four individual components, i.e., evaporator,condenser, compressor, and expansion valve, which can be modeled, separately.The chiller works on the basis of vapor compression cycle in which the phasechange of liquid refrigerant in evaporator takes away heat from the air-conditionedspace; then, compressor increases the pressure of gas refrigerant making it super-heated and releases it into the condenser where the gas refrigerant is condensed andthe condensation heat is rejected to the water or air; afterward, the expansion valvereduces the pressure by releasing the refrigerant in the evaporator in a cool state

models including shell-and-tube condenser and evaporator, air-cooled condenser,direct expansion (DX) evaporator, capillary tube, and thermostatic expansion valve

sepa-rating the vapor and liquid phases and coupling the mass and energy balances of the

individual phases through the evaporation or condensation The model was then

performance

refrigeration system based on the component models by using mass and energybalance principles With the dynamic model, the effects of control inputs such ascompressor operational frequency and TEV opening fraction on the output per-formance of the system were investigated

were used for modeling the condenser and evaporator of chiller Meanwhile, the

condenser, and pipes

The moving-boundary (MB) formulation is characterized by phase boundariesthat move with time within the heat exchanger Subject to the same assumptions as

heat exchanger volume into variable control volumes that encompass each phaseregion existing in the heat exchanger [13,14] Nyers and Stoyan [15] built a model

system model of a basic vapor compression refrigeration system for the purpose ofstudying the effect of multivariable feedback control and by other researchers for

approaches applied to shell-and-tube heat exchangers Detailed model formulations

of both the FV and the MB approach for shell-and-tube heat exchanger modelingwere provided, and stability was demonstrated as components and within a com-plete centrifugal chiller system model They concluded that the FV formulationwould be more robust through start-up and all load-change transients, but executeslower, while the MB method could handle all load-change transients, but start-upstability would be more sensitive to compressor and expansion valve formulations

Cooling towers are widely used to remove heat from industrial processes and fromHVAC systems Heat rejection in cooling towers is accomplished by heat and masstransfer between hot water droplets and ambient air Physical models of cooling

Lewis number was assumed to be one in order to simplify the analysis Although

the Merkel’s model has been the basis for most modern cooling tower analysis, itdoes not accurately represent the physics of heat and mass transfer process in the

reliability still depends on the accuracy of geometric information of cooling towers.The physical models of cooling tower are also developed with the numerical

water-cooling tower The model could be used to make a detailed numericalanalysis on the air and water states at any horizontal plane along the tower height

tower The model consisted of two interdependent boundary-value problems Thefirst boundary-value problem described evaporative cooling of water drops in the

mathematical model of a mechanical draft cooling tower The model represented aboundary-value problem for a system of ordinary differential equations, describing

a change in the droplets velocity, its radii, and temperature, and also a change in thetemperature and density of the water vapor in a mist air in a cooling tower Inaddition, the model allowed one to calculate contributions of various physicalparameters to the processes of heat and mass transfer between water droplets and

the tower

In a HVAC system, the cooling/heating coil handles the supply air to anticipatedconditions The heat exchanger model can be obtained by the energy and massbalance on the water and air side of the coil The physics-based approach for heatexchanger modeling results in a large number of governing equations Solutions to

② lumped-parameter solutions, and ③ analytical solutions

For the numerical solution, air-cooling coil is physically divided into numeroussegments, and the inlet and outlet variables of each segment are calculated in turn[26,27] With this approach, the distributions of air temperature and humidity in the

eval-uated, but it requires large computational cost

The use of a lumped-parameter model can be, however, more preferred than theuse of a numerical solution due to its lower computational costs In alumped-parameter model, the enthalpy difference between air and cooling medium

is treated as the driving force of simultaneous heat and mass transfer Such a

Analytical solutions may be able to evaluate the thermal performance of ferent types of heat exchangers in an accurate manner but less computational cost

Bielski and Malinowski [32] used the analytical method to solve a set of partial

the results obtained from their analytical solution with that from the numerical andsemi-analytical solutions developed in other studies to validate their analytical

exchanger by using the integral method In their model, the temperature distribution

temperature distributions as well as a determined time function The model was

the coupled heat and mass transfer process in parallel/counter indirect evaporativecoolers under real operating conditions The analytical solution model was vali-dated by comparing its results with those from numerical integrations Xia et al

chilled water wet cooling coils and DX wet cooling coils, respectively, under bothunit and non-unit Lewis Factors With the analytical solutions, the distributions of

could be predicted, and the differences in the thermal performances of the cooling

developed a linearized model for air-cooling coil with small disturbances that wassolved by means of Laplace transform The model could be used to study the

different initial conditions

In recent years, the state-space model for air-cooling coil was put forward by

the dynamic characteristics of a multiple-input-multiple-output (MIMO) system.The detailed information about the state-space modeling will be introduced in the

Fans/pumps are the main power equipment in HVAC system, which consume over

rate, pressure head, and input power The model could assist us to design a

per-formance was considered

1.2.1.5 Duct/Pipe Model

duct and pipe are the two important factors concerned in the physics-based model

model for duct and analyzed the resistance loss of duct system in HVAC system

duct system model included physical characterization, air leakage, and heat

con-vective heat transfer in duct and investigated the turbulent concon-vective heat transfer in

The air-conditioning system is usually designed with respect to the thermal acteristic of the air-conditioned room, and the system control scheme is developedbased on the requirement of indoor thermal environment The physical model forair-conditioned room can be obtained according to energy and mass balance of airwithin a zone Heat is transferred to the zone through the supply air, conduction

conservation For a zone, generally, each heat transfer element (wall, window,

of each heat transfer element using the matrix algebra techniques The heat balancemethod has been adopted for studying dynamic responses of air temperature indoors

As early as the 1980s, Metha et al [47] established a dynamic response model for theindoor air temperatures with the assumption that the air in the room was well mixed

between room air and surrounding walls were taken into account in different ways,

thermal response time was used for the dynamic simulation The study concluded

by energy balance on the room air, two walls, and the ceiling The dynamic roommodel was employed for the control analysis on a HVAC system

In thermal network zone model, the building is divided into a network of nodes

this method varies based on the selection of nodes on which energy balance isapplied Zone models exist with different levels of complexity: from simple

‘well-mixed’ models with one air node representing the whole air volume in theroom to complex CFD models solving the equations of conservation of mass,

thermal response and indoor air distribution were computed considering the outdoorair temperature, solar radiation, indoor heat sources, and other thermal boundaryconditions Although the CFD results have been shown to be accurate, the calcu-

into several air zones each of which is assumed to be well-mixed as one air node Wu

model is much closer to real situations while saving much computational timecompared to the CFD model So, it may be the best choice for studying the roomtemperature distributions and dynamic thermal characteristics of indoor air

Physics-based models provide good generalization capability but lack of accuracycompared with the data-driven ones In addition, the calibrations of physics-based

of parameters Major methods used for data-driven modeling of HVAC systems

Due to the heavy thermal inertia existing in the HVAC system, many processessuch as dynamics of exit air temperature of air-cooling/heating coils are changing

GðsÞ ¼YðsÞ

1

where, K is static gain; L is apparent dead time of the process;s is time constant; a,

dead time have simple structure and a small number of parameters to be determined

these parameters

The data mining and machine learning algorithms such as ANN and support vectormachine (SVM) are often applied to complicated and nonlinear system like HVACsystems The network is trained by a supervised learning algorithm, and theSVM-based approach projects the nonlinearly separable data into higher dimensionalfeature space through a mapping function in which it can be separated linearly.The ANN method is often employed to estimate the performance of HVAC

per-formance of an automobile air-conditioning system using ANN They developed amultilayer feed-forward networks (MLFFN) for predicting the performanceparameters such as compressor power, heat rejection rate in the condenser,

compressor speed, cooling capacity, and condensing temperature Ertunc and

using ANN techniques They predicted the condenser heat rejection rate, exitrefrigerant temperature of condenser as well as dry- and wet-bulb temperatures ofthe leaving air stream with respect to the following parameters: inlet air temperatureand humidity, airflow rate, refrigerant flow rate, water flow rate, absolute pressure,and temperature of the refrigerant at the inlet of the condenser

The SVM method has been applied to load prediction in building HVAC systems

model based on measured data of cooling load over a period The global optima ofSVM penalty parameter, intensive loss function, and kernel function were found byusing the ant colony optimization (ACO) To improve the load forecasting capacity,the SVM model was often combined with the other data analysis algorithms, e.g.,hybrid SVM combined with autoregressive integrated moving average (ARIMA),hybrid SVM combined with kernel principal component analysis (KPCA), andhybrid SVM combined with particle swarm optimization (SAPSO) algorithm

and provides the bestfit for the data In order to build the models using data miningalgorithms, large amount of training and testing data is needed No physicalinterpretation of the developed model is possible and the performance degradeswhen conditions deviate from training and testing conditions These algorithms are

The fuzzy logical model is developed based on the expert knowledge, and it isimplemented through the if-then-else statements or rules which are written in theform of a table or database Fuzzy logical model is developed usually for the control

temperature and relative humidity in refrigeration system by considering their

control (MMPC) strategy based on Takagi-Sugeno (T-S) fuzzy models for perature control of air-handling unit (AHU) in HVAC systems The overall HVACcontrol system was constructed by a hierarchical two-level structure The higherlevel is a fuzzy partition based on AHU operating range to schedule the fuzzyweights of local models in lower level, while the lower level is composed of a set ofT-S models based on the relation of manipulated inputs and system outputs cor-respond to the higher level There are also other new fuzzy logical models com-

developed with fuzzy logic require experiences and comprehensive knowledge

may not be readily available for many HVAC components and thus presents a

exogenous (ARX), autoregressive moving average exogenous (ARMAX), ARIMA,Box-Jenkins (BJ), and output error (OE) The mathematical expression for the

relationship is given below:

aðq1ÞyðsÞ ¼b1ðq1Þ

h1ðq1ÞuðsÞ þ

b2ðq1Þ

where a; b1; b2; h1; h2 are polynomials; q1 is back shift operator; uðsÞ; yðsÞ, and

The model ARMAX is superior to ARX as it incorporates the time series of error

in the model structure which is essential for capturing the dynamics of the error andbetter control performance ARIMA is a generalization of ARMAX, modeling thestationary and non-stationary data into a single step, and consists of autoregressive,

Box-Jenkins (BJ), autoregressive with external inputs (ARX), autoregressivemoving average with external inputs (ARMAX), and OE models to identify the

Their study manifested that these numerical models could all be potentially used forimproving the performance of the thermal environment control system Otherexamples of statistical model applications in HVAC areas include the following:predicting the room temperature variations for both short-term and long-term

processes in an HVAC system depend on their previous values, a time seriesregression model (i.e., ARX, ARMAX, and ARIMA) captures these correlations byincluding the process variables from the previous sampling times, which will result

in a favorably accurate model of the process dynamics

Hybrid models use physics-based models as the model structure, and their modelparameters are estimated from the measured data These models provide physicalmeaning and are superior to data-driven models in terms of generalization capa-bility Hybrid models choose appropriate parameters to capture regular patterns ofthe system which cannot be well modeled with physical equations The determi-nation of model parameters requires knowledge of both the physical phenomenaand the data from the process

water-cooling coils in a static state The model was built based on the heat transfermechanism and the energy balance principle The key model parameters that rep-

to identify parameters in the partially known models of AHU elements includingair-cooling coil, electric heater, and humidifier In another study by Yao et al [69],

the dynamic model were determined by test data with ANN method

Hybrid models provide better accuracy than physics-based models and better

to develop In order to develop hybrid models, both the knowledge of underlying

HVAC sub-systems, the underlying physical phenomenon could be very

complicated to model, and for the other systems, the input–output data may not be

hybrid models also need to be updated timely when the operating conditions deviatefrom the training data in order to ensure higher accuracy

State-space model can be obtained directly from the input and output data surements [70] However, it is often derived from a series of physics-based equations

the study of heat transfer characteristics of building wall Afterward, the state-spacemethod has been employed by limited researchers to develop dynamic models for

air-conditioning systems [75,76] In this book, the state-space method will be used

to develop dynamic models of all components in HVAC system, and thesestate-space models can be used to study dynamic characteristics of HVAC equip-

following:

(1) Unlike the ANN, ARMAX, and ARX models, it can be of physics-basedmodel which allows us to understand the essences of dynamic relationshipbetween the input perturbations and the output response variables of thesystem

(3) It has uniform representation and hence enhances the integration of systemmodeling

(4) It provides not only information of system outputs, but also that of the systemstates which may be used as feedback signals to improve control performances.(5) It is an important basis for the development of MPC (model-based predictivecontroller) that may have good application prospect in the HVACfields [77,78].(6) Lastly, state-space model is expressed in the form of matrix that is convenientfor computer calculation

various areas Garg et al [80] developed a deterministic quantitative model based on

graph theoretical methodology to compare various technical and economical features

of wind, hydro, and thermal power plants and also used to evaluate and rank the powerplants in ascending or descending order in accordance with the value of their suit-

with the help of graph theory, then by variable adjacency matrix and then by a

automatic synthesis of mathematical models by using graph theory for optimization ofthermal energy systems The topology of the graph was stored in the computermemory, and the computer model of the respective system could be constructed

graph-based model of manufacturing system The matrix models and the variablepermanent function models have been developed to carry out decomposition, char-acterization, and the total analysis Wang et al [84] employed the graph-theory-basedmethod to describe theflexible flow circuits of different liquefaction processes of LNGspiral wound heat exchanger The graph-theory description helps to solve thedistributed-parameter model in a proper way

limited although the topology of a HVAC system may make the theory particularly

analysis of air-conditioning system and integrate different components of theair-conditioning system and takes into account the effect of interaction betweenthese components for the better understanding of the structure of theair-conditioning system, the graph-theory approach is employed in this book for

modeling in HVAC system mainly include the following:

air-conditioning system and identifying interactions among these differentcomponents;

(2) Makes it possible to automatically develop the model of the whole system byusing models of components from a library and information about the topology

of the system (i.e., the components comprising the system and theirinterconnections);

(3) Development of matrix models and the variable permanent function models tocarry out decomposition, characterization, and overall analysis

Accurate prediction of air-conditioning load is the critical basis for energy-savingoperations of central air-conditioning system including adjusting the starting time ofcooling to meet start-up loads, minimizing or limiting the electric on-peak demand,

developed many forecasting models for air-conditioning load prediction, such asARIMA model, exponentially weighted moving average model (EWMA), multiple

individual forecasting model has its own limitations For example, the ARIMA and

short-term forecasting; the ANN model needs large amount of data to train, and it isvalid only within the scope of the training data; the MLR model may result in bigforecasting errors because nonlinear relationships always exist in actual situations;the GM uses the operations of accumulated generation to build differential equa-tions for prediction, and it may be more suitable for long-term forecasting Toovercome shortcomings and make full use of advantages of individual forecastingmodel, the combined forecasting method is introduced in this book The theory ofthe combined forecasting method is based on a certain linear combination of var-

forecasting model is greatly improved, and the forecasted result will present a

the optimal combined forecasting method whose deviation reaches the minimumand is less than that of each individual forecasting method

Optimization Model

Control functions of building automation systems (BASs) can be divided into twocategories, local control functions and supervisory control (or energy management)functions Local control functions are the basic control and automation that allowthe building services systems to operate properly, which are established based onthe dynamic models of controlled objects (e.g., state-space model) The supervisorycontrol functions determine optimum operation points of equipment in HVACsystem aiming at the minimum energy input or operating cost to provide thesatisfactory indoor comfort and healthy environment Aiming at energy saving ofcentral air-conditioning systems, many researchers have done a lot of work on

that of the air-distribution sub-systems were particularly paid attention to by theother studies [97–99] As well known, local optimal operations will not always lead

to global optimization results A central air-conditioning system normally consists

of a primary chilled water plant and numerous air-handling and air-distributionsub-systems When the whole HVAC system is taken into account, there ought toexist a trade-off of energy consumption among different equipment in this system It

is easily understood that increasing the fan speed of the cooling tower increases fan

the chilled water set point temperature reduces chiller power but increases pump

power, and so does increasing the supply air set point to chiller power and fan

systems as a whole, and global optimization model is suggested for the supervisorycontrol of whole system

Searching optimization method is equally important in realizing global mization control strategies Many optimization algorithms have been developed in

concepts of genetic and evolutionary algorithm have been introduced as well to

features and roles in solving certain optimization problems, but they have their ownlimitations in real applications The direct search methods may be less computa-

gradient-based and nonlinear programming methods are only applicable for solvingthe optimization problems that can be directly presented by a continuous function;the ANN relies too much on the input data for training and is unable to reliablyextrapolate beyond the calibration range Although genetic can quickly evaluate alarge solution area and work quite well for problems with high degree of com-plexity, it has the biggest disadvantage due to its nature of evolution that may easilyevolve away from a bad solution and lead to a large number of iterations for

large complicated nonlinear systems that consist of many strongly coupledsub-systems There will encounter big challenge in solving the global optimizationproblem in which a large number of equipment are involved and numerous decisionvariables are to be determined The high-dimensional problems will easily result in

what kind of optimization algorithm (mentioned above) is adopted

In this book, a global optimization model for central air-conditioning system isdeveloped The global optimization problem is formulated based on the energymodels of components in HVAC system, including chiller, water-to-air surface heat

‘high-dimensional disaster’ problem of global optimization model for some

Due to the large quantity and complexity of devices in central HVAC systems, this

enhanced BAS that electronically integrates the mechanical devices through ing, computing, data processing, and actuating However, due to the consideration

sens-on initial cost, building systems are generally under-sensed with near-zero sensor

redundancy Physical variables of our interest in HVAC systems and buildings may

be measured with only one sensor or even not measured For example, the outdoorintake ratio in AHUs is seldom acquired Meanwhile, many sensors in buildingHVAC systems are improperly installed, wrongly placed, damaged, or gradually

transmitters could be inaccurate or totally wrong Because of zero redundancy in

measurements Using erroneous data or wrong information could lead to a icant energy penalty or even direct failure of control and operation algorithms.Sensor errors generally comprise precision degradation, reading bias, drifts,noise, or sensor failure Conventional approaches for correcting the errors andimproving the accuracy of measurements from various sensors and meters in real

(2) statistics-driven data fusion [114–116] The essence of a physical sensor

environment to bring the working sensor back to its normal condition A sensorcalibration is the fundamental method of correcting suspicious sensors Generally,all sensors in a dynamic system should be checked regularly against standard

example, for temperature measurements, sensors should be calibrated every

12 months; for pressure gauges, it is desired for every six months Beside the sensorcalibration, statistic data fusion methods may also be applied to obtain the repre-sentative value of physical variables With a data-driven method, different data orinformation sources (for instance, direct measurements from physical sensors andindirect measurements from models) are integrated in a data fusing process to obtainthe accurate, complete, or dependable information The main procedure of

A sensor calibration is more preferable over a data fusion method since theformer works frontend on a sensor itself for maintaining the quality of directmeasurements Meanwhile, a calibration is the most effective method in reducingsystematic errors and eliminating failure of sensors Despite the necessity, a sensorcalibration is barely carried out regularly on various sensors in building HVAC

challenges to conduct a regular calibration on sensors are as follows:

(1) Time and monetary cost A complete calibration process of an individualsensor includes multiple steps, from removing a working sensor from a system,conducting a calibration, to reinstalling it back; any of the steps could betime-consuming and expensive

(2) Disruption to a normal operation Removing and reinstalling a sensor willmore or less disrupt the normal operation of HVAC systems Missing mea-surements from the removed sensor also need to be covered temporarily toresume the operation during the process

(3) Access to various sensors Due to the space and installation constraints, it

meter in a pipeline, a temperature sensor hiding behind the ceiling) from itsworking environment

(4) Large quantity of sensors Building HVAC systems have a large sensor work to acquire different types of information (e.g., temperature, humidity,

In addition to these challenges, there is one more limitation directly associatedwith a conventional calibration A physical sensor after calibration may not have afavorable working environment, as that in the calibration, to work properly and

that the commonly preinstalled supply air temperature sensor in compact rooftop airconditioners cannot accurately measure the real temperature of supply air Due tothe compact size, poor air distribution, and intensive thermal radiation of gasheating chamber, errors associated with the sensor could be up to 19.2 °C and

be erratic

In addition to acquiring improved accuracy and resiliency against errors, an idealcalibration process should be conducted as in situ, hence avoiding the differences inthe medium and changes of working environment and the associated effects on themeasurements Some studies have recently been conducted in the area of automated

[122] and self-calibration [121], are used synonymously as virtual in situ calibration

the sensor calibration function can be depicted with a linear model; therefore, acalibration problem was transformed to obtain the unknown gains and offsets.Slightly over-sampling was assumed for general applications in order to solve thelinear system of equations A virtual standard concept was proposed by Dulev et al

treated as a general parameter estimation problem in a study conducted by

response The average measurement errors were reduced from 74.6 to 10.1 % afterthe implementation of the method An iterative registration and fusion approach

calibration approach was to minimize the squared distance error through the

The redundancy was utilized to calibrate one sensor against the others The relativecalibration relationships, as temporal correlations between pairs of co-located

maximize the consistency of the pairwise functions among sensors Wirelessthermistors were tested to evaluate the proposed method A self-calibration method,formulated as an inference problem on a graphical model, was investigated for

was then applied to obtain the solution for the problem

tech-nology for the measurement of supply air temperature in packaged air-conditioningunits The measurement error, which is up to 19.2 °C and erratic, can be improved

robustness over a wider range of operating and fault conditions The book willarrange a chapter to present an innovative virtual in situ calibration algorithm

potentially automated to handle the aforementioned challenges and limitation of aconventional calibration

Functions of building control are typically fulfilled in a two-level structure: a visory level and a local level The supervisory-level controller comprises a group oflogics and typically resides in a central station It takes actions based on preset con-ditions and/or rules, or the commands from the operators, such as heating/coolingswitch-over, components sequencing, event scheduling, etc However, this type ofrule-based responsive control, with local on/off and proportional–integral–differential(PID) controllers, is expensive in the long run since they operate at a non-optimal

super-efficiency [125] In addition, neither of the two control objectives, thermal comfortand energy savings, can be explicitly expressed in the conventional control laws [126]

An advanced building control system needs to possess some kind of predictivecapability with building dynamics and deterministic and stochastic disturbancesconsidered and can evaluate different objectives rather than decoupled simple refer-ences One of such enabling technologies is termed model-based predictive control(MPC) [127] It includes a building system model, a cost function, a set of constraints,and an optimization solver and runs in a receding horizon manner From a generalpoint of view, an MPC is a strategy, rather than a control law, which uses a combi-

handling constraints, cross-coupling issues, and multiple objectives in a nonlinearmultiple-input-multiple-output (MIMO) system, MPCs have been greatly utilized inthe manufacturing process industry since its introduction in the 1970s and

Research on MPC in building systems has just recently started to thrive[130–133] Upon the literature review, wefind the majority of these studies are in thearea of general MPCs, which usually couple a simulator with an optimization solver

for the applications With this approach, the mathematical structure of the simulator

is invisible to the outside solver Stochastic global search strategies, such as geneticalgorithm, simulated annealing, particle swarm optimization, are applied to identifythe optimums This approach is generally associated with very high computationalcost and not suitable for online implementation For example, Zhang and Hanby

were possible, a search for 13 hourly commands took over 24 h to complete

was recommended followed by the logistic regression on the data to extract somesimple rules for real implementation Another approach of formulating an MPC is touse an explicit linear model of a physical system and make the gradients visible tothe optimization solver This allows the application of powerful tools, such as the

Study of utilizing the good features of linear MPC on multivariate HVAC systems is

MPCs to a variable air volume system to achieve an acceptable indoor air quality.The ventilation condition in the six zones with MPC was found to be improved

comfort control in a single-zone one-actuator air-conditioning building Two casestudies in terms of an MPC and different metabolic rate and clothing index were

pro-gramming method in the canonical form to the predictive control of intermittentheated buildings The mathematic summation of energy consumption and thermal

the radiator was used as the control input

ther-mal comfort and energy savings explicitly in the formulation of MPCs In addition,the method and application of a linear classical MPC for MIMO nonlinear buildingssystems have not been fully explored The book will arrange a chapter to investigatethe methodologies and potentials of utilizing linear classical MPCs for nonlinearsystems, considering both thermal comfort and energy cost

are summarized as below:

This chapter introduces background of the topic related to this book, gives a

physics-based, the data-driven, and the hybrid Afterward, the proposed methods to

Chapter 2 illustrates in detail the state-space modeling process for HVACcomponents including water-to-air surface heat exchanger, chiller, cooling tower,air-conditioned room, and duct (pipe) as well as fan (pump) Then, the HVACsystem modeling based on the component state-space models is given Meanwhile,experimental validations for all the state-space models are presented.

Chapters Approaches

Virtual measurement modeling

(1) Modeling for virtual sensing; (2) Modeling for virtual calibration Physics-based modeling.

Chap.6

Control design based

on state-space model

Theory and case study on:

(1).State feed-back control;

(2).Model-based prediction control

State feed-back control;

Model-based prediction control

Combined-forecasting method;

Analytic hierarchy process (AHP);

Combined residual error correction;

ANN,GM, ARIMA, MLR.

Chap.8

Energy analysis model for HVAC system

(1).Energy models for HVAC;

(2).Evaluation program for energy-saving

Case study on:

(1) Total air volume control;

(2) Constant static pressure control; (3) Trim & Respond reset control.

Trim & Respond static pressure

set-point reset logic control;

Constant static pressure control;

Total air volume control;

State-space method;

First-order Taylor linearization.

Transient response with

state-space models.

Fig 1.1 Basic structure of this book