System-Oriented Dynamic Modelling and Power Consumption Simulation for an Industrial Chiller Plant

(1)

Master of Science Thesis

KTH School of Industrial Engineering and Management Energy Technology TRITA-ITM-EX 2020:614

Division of Energy Technology SE-100 44 STOCKHOLM

System-Oriented Dynamic Modelling

and Power Consumption Simulation

for an Industrial Chiller Plant

(2)

-2-

Master of Science Thesis

TRITA-ITM-EX 2020:614

System-Oriented Dynamic Modelling and Power Consumption Simulation for an

Industrial Chiller Plant

Yanan Zhang Approved

Date

Examiner

Hatef Madani Larijani

Supervisor

Hatef Madani Larijani Bryant Zhong

Commissioner Contact person

Abstract

Nowadays, the amount of energy consumed by Heating, Ventilation and Air Conditioning (HVAC) systems poses threats to the improvement of global environment. Hence people have been dedicated to the development of energy-saving retrofit projects via optimized simulation in these chiller plants. However, recent research only focuses on the improvement of one or two devices instead of whole operating system, and researchers seldom pay attention to the comparison between different optimization strategies.

As a result, by using Modelica/Dymola framework, a system-oriented dynamic model is constructed via the on-site investigation of an industrial chiller plant which located in Chengdu, China in this project. This system is composed of the individual control of multiple chillers, water pumps and cooling towers with a relative mean absolute error (RMAE) of 1.4% in the energy consumption simulation and hence can be regarded as a reliable model for further optimizations.

After the implantation of two-months simulated data as system configuration, three optimization strategies including Baseline Strategy (BS), Sequence Optimization (SO) and Holistic Optimization (HO) within the full range cooling load ratio are investigated to explore the potentials of energy savings for this model. Among all three optimization approaches, HO strategy shows the outperforming reduction in energy consumption by 10.8% especially at low range of cooling load ratio. Moreover, based on HO method, two more advanced strategies which named as Holistic Optimization based on Cooling Load Time Frequency Analysis (HOTF) and Holistic Optimization based on Cooling Load Distribution Analysis (HOCL) are explored. The results turn out that both optimization strategies on this model can further decrease the energy consumption by 2.6% and 4.0% accordingly. In addition to this, HOTF also shows its own strength on low cost and operation simplicity.

(3)

-3-

Table of Figures

Figure 2.1 High Efficiency Chiller Plant Standard ... 16

Figure 2.2 Disintegration of a Chiller Plant Model ... 20

Figure 2.3 Chiller Part Load Performance curve - Yu’s Project ... 21

Figure 2.4 Chiller COP at Different CLR for OP1-OP4 - Yu’s Project ... 22

Figure 2.5 Chilled Water Pump Consumption at Different CLR for OP1-OP4 - Yu’s Project ... 22

Figure 2.6 Probability Density Distribution of Cooling Load Ratio - Bo’s Project ... 22

Figure 2.7 Operation Sequence of Chillers based on the Strategy - Bo’s Project ... 22

Figure 2.8 Power Consumption of Chillers under AⅠ, AⅡ and AⅢ Strategies - Bo’s Project ... 23

Figure 2.9 Water Flow Rate of Energy Optimization Methodology - Sundar’s Project ... 24

Figure 2.10 Water Temperature Set-Point of Energy Optimization Methodology - Sundar’s Project ... 24

Figure 2.11 Total Energy Consumption of Chiller Plant under Different Operation Strategy – Sundar’s Project ... 24

Figure 3.1 Methodology of Dynamic Modelling Process ... 25

Figure 3.2 Control Logic Diagram of Holistic Optimization Strategy ... 28

Figure 3.3 Devices Mix based on Strategies (HO1, HO2 and HO3) ... 30

Figure 3.4 Total Power Consumption under HO1, HO2 and HO3 Strategies ... 30

Figure 3.5 Device Mix of HOCL Strategy... 30

Figure 4.1 On-site Picture of Cooling Towers ... 32

Figure 4.2 On-site Picture of Water Pumps ... 32

Figure 4.3 On-site Picture of Chiller CSR6604_1 ... 32

Figure 4.4 On-site Picture of 3 Chillers ... 32

Figure 4.5 Layout of the Chiller Plant System ... 33

Figure 4.6 Air Temperature from 2020.07.01 to 2020.08.31 ... 34

Figure 4.7 Environment Temperature and Cooling Load for Accuracy Check ... 34

Figure 4.8 Cooling Load Data Simulation Model ... 36

Figure 4.9 Internal Heat Source of the 2 Target Buildings ... 36

Figure 4.10 Simulated Cooling Load Data of the 2 Target Buildings ... 36

Figure 5.1 The Diagram of the Chiller Model from HVAC Library ... 39

Figure 5.2 Connection Logic of Chillers ... 40

Figure 5.3 Connection Logic of Chilled and Cooling Water Pumps ... 41

Figure 5.4 The Interface of Water Pumps After Programming ... 42

Figure 5.5 The Diagram of Cooling Tower Model ... 43

Figure 5.6 Connection Logic of Cooling Towers Model ... 44

Figure 5.7 The Interface of Cooling Towers After Programming ... 45

Figure 5.8 Connection Logic of Cooling Load Model ... 45

Figure 5.9 Weather and State components ... 46

Figure 5.10 Sequence controlling Logic ... 46

Figure 5.11 Mass flow rate Controlling Logic ... 47

Figure 5.12 Valve opening controlling Logic ... 48

(7)

-7-

Figure 5.14 The Physical carrier of Temperature Difference Controlling Logic ... 49

Figure 5.15 Interface Icon, cable (left) and bus (right) ... 49

Figure 5.16 Global View of Baseline Model – Version 1.0 ... 50

Figure 5.17 Global View of Holistic Optimization Model – Version 1.0 ... 50

Figure 5.18 Global View of Holistic Optimization First Layer Model – Version 2.0 ... 51

Figure 5.19 Global View of Holistic Optmization Second Layer Model – Version 2.0 ... 51

Figure 5.20 Comparison of Total Power Consumption ... 52

Figure 6.1 Total Power Consumption under BS, SO and HO Strategies ... 53

Figure 6.2 Energy Consumption Distribution under BS, SO and HO Strategies ... 53

Figure 6.3 Station COP under BS, SO and HO Strategies ... 54

Figure 6.4 Total Chillers Compressor Power under BS, SO and HO Strategies ... 54

Figure 6.5 Single Chiller Compressor Power under BS, SO and HO Strategies ... 54

Figure 6.6 Number of Chillers in Working under BS, SO and HO Strategies ... 55

Figure 6.7 Compressor Power Ratio of 3 Chillers under BS, SO and HO strategies ... 55

Figure 6.8 Chiller COP under BS, SO and HO Strategies ... 55

Figure 6.9 Chiller Part Load Rate under BS, SO and HO strategies ... 55

Figure 6.10 Condensing Temperature under BS, SO and HO strategies ... 56

Figure 6.11 Evaporating Temperature under BS, SO and HO Strategies ... 56

Figure 6.12 Cooling Load Distribution under BS, SO and HO strategies ... 56

Figure 6.13 Total Pumps Power Consumption under BS, SO and HO Strategies ... 57

Figure 6.14 Single Pump Power Consumption under BS, SO and HO strategies ... 57

Figure 6.15 Number of Pumps in Working under BS, SO and HO Strategies ... 58

Figure 6.16 Pumps Power Consumption Ratio under BS, SO and HO strategies ... 58

Figure 6.17 Water Transport Factor of Pumps under BS, SO and HO Strategies ... 58

Figure 6.18 Temperature Difference of CHW and CW under BS, SO and HO Strategies ... 59

Figure 6.19 Water Flow Rate of CHW and CW under BS, SO and HO strategies ... 59

Figure 6.20 Power Consumption of Cooling Towers under BS, SO and HO Strategies ... 60

Figure 6.21 Condensing Temperature under BS, SO and HO strategies ... 60

Figure 6.22 Efficiency of Cooling Tower CT1 under BS, SO and HO Strategies ... 60

Figure 6.23 Efficiency of Cooling Tower CT2 under BS, SO and HO strategies ... 60

Figure 6.24 Total Power Consumption under HOB, HOTF and HOCL Strategies ... 61

Figure 6.25 Number of Operating Chillers under HOB, HOTF and HOCL Strategies ... 61

Figure 6.26 Energy Saving Performance under HOB, HOTF and HOCL Strategies ... 62

Figure 6.27 Energy Saving Ratio under HOB, HOTF and HOCL Strategies ... 62

Figure 6.28 Station COP under SO, HOB, HOTF and HOCL Strategies ... 62

Figure 6.29 Weather Conditions on the Date 2020.07.18 ... 63

Figure 6.30 Cooling Load Distribution on the Date 2020.07.18... 63

Figure 6.31 Energy Consumption during the Typical Day ... 64

Figure 6.32 Average Station COP during the Typical Day ... 64

(8)

-8-

Figure 6.34 Real-Time Station COP during the Typical Day (Bottom) ... 64

Figure 6.35 Real-Time Number of Chillers in Working during the Typical Day (Top) ... 65

Figure 6.36 Real-Time Chillers Power Consumption during the Typical Day (Bottom) ... 65

Figure 6.37 Real-Time Part Load Rate of Single Chiller during the Typical Day (Top) ... 65

Figure 6.38 Real-Time COP of Single Chiller during the Typical Day (Bottom) ... 65

Figure 6.39 COP Curve of Chillers Applied in this project ... 66

Figure 6.40 Real-Time Total Power Consumption of Water Pumps during the Typical Day (Top) ... 66

Figure 6.41 Real-Time Water Flow Rate during the Typical Day (Bottom) ... 66

Figure 6.42 Real-Time Number of Chilled Pumps in Working during the Typical Day ... 67

Figure 6.43 Real-Time Number of Cooling Pump in Working during the Typical Day ... 67

Figure 6.44 Real-Time Water Transport Factor of Chilled Pumps during the Typical Day ... 67

Figure 6.45 Real-Time Water Transport Factor of Cooling Pumps during the Typical Day ... 67

Figure 6.46 Water Transport Factor Improvement Ratio of Pumps during the Typical Day ... 68

Figure 6.47 Energy Saving Ratio of Pumps during the Typical Day ... 68

Figure 6.48 Real-Time Power Consumption of Cooling Towers during the Typical Day (Top) ... 68

Figure 6.49 Real-Time Efficiency of Cooling Towers during the Typical Day (Bottom) ... 68

Figure 7.1 Total Energy Saving Ratio Comparison of the two Projects ... 70

(9)

-9-

Table of Tables

Table 2.2 Cooling Load Range - Yu’s Project ... 22

Table 2.3 Operating Decision of Chiller Plant - Sundar’s Project ... 23

Table 2.4 Performance Measures of Chiller Plant with Two Centrifugal Chillers – Sundar’s Project ... 24

Table 3.1 Operating Decisions under 3 Common Operation Strategies ... 26

Table 3.2 Ranges of Cooling Load Ratio of HOTF strategy ... 29

Table 3.3 Optimal Strategy of under CLTF Analysis ... 30

Table 4.1 Outdoor Environment Information ... 34

Table 4.2 Real-Time Cooling Load for Real Industrial Scenario ... 34

Table 4.3 System Configurations for Accuracy Check ... 35

Table 4.4 System Configurations of 3 Common Operation Strategies (BS, SO and HO) ... 35

Table 4.5 Key Inputs of Cooling Load Data Simulation ... 36

Table 4.6 System Configurations of 3 HO Strategies (HOB, HOTF and HOCL) ... 37

Table 5.1 The MAE, MBE, RMAE and RMBE of Power Consumption Simulation Results ... 52

Table 5.2 The MAE, MBE, RMAE and RMBE of Chiller Performance Simulation Results ... 52

Table 6.1 Four Ranges of CLR ... 53

Table 6.2 Total Electricity Consumption under SO, HOB, HOTF and HOCL Strategies ... 62

Table 6.3 Energy Consumption during the Typical Day ... 63

Table 6.4 Cooling Load Distribution during the Typical Day ... 64

Table 6.5 Competitive Strategies Order in different CLR Ranges ... 65

Table Appendix 1.1 Table of nameplate parameters of Chillers, Water Pumps and Cooling Towers ... 77

Table Appendix 2.1 Controlling Logic for Sequence control Model ... 78

Table Appendix 2.2 Meaning of Every Signal for Sequence control Model ... 78

Table Appendix 2.3 Controlling Logic for Mass flow rate Control Model ... 78

Table Appendix 2.4 Meaning of Every Signal for Mass flow rate Control Model ... 78

Table Appendix 2.5 Controlling Logic for Valve opening control Model ... 79

Table Appendix 2.6 Meaning of Every Signal for Valve opening control Model ... 79

Table Appendix 2.7 Meaning of Every Signal for Temperature Difference Control Model ... 79

Table Appendix 2.8 Programming for cables and bus ... 79

Table Appendix 3.1 The Comparison of Power Consumption Between the Simulation Data and On-Site Date ... 81

(10)

-10-

Table of Symbols

Abbreviation Meanings Abbreviation Meanings

3DE 3D Experience HOB Basic Holistic Optimization

ACs Air Conditionings HVAC Heating, Ventilation and Air Conditioning

AE Absolute Error IEA International Energy Agency

AHU Air Handling Unit IGSD Institute of Governance and Sustainable Development

BE Bias Error IoT Internet of Things

BMS Building Management System MAE Mean Absolute Error

BS Baseline MAU Make-up Air Unit

CH Chiller MBE Mean Bias Error

CHP Chilled Water Pump OS Operation strategy

CHW Chilled Water PAU Primary Air Unit

CLR Cooling Load Ratio PID Proportion Integration Differentiation

COP Coefficient of Performance RAD Return Air Dust

CP Cooling Water Pump RMAE Relative Mean Absolute Error

CT Cooling Tower RMBE Relative Mean Bias Error

CT Cooling Tower SAD Supply Air Dust

CW Cooling Water SO Sequence Optimization

CWP Chilled Water Pumps Tair Air temperature

CWP Cooling Water Pumps Tscd Set-point of condensing water temperature

DAE Differential Algebraic Equations Tscw Set point of chilled water temperature

DBM Dymola Behavior Modelling Twb Outdoor Wet-bulb temperature

DRC Direct Refrigerant Cooling VFD Variable Frequency Drive

DS Dassault Systems VSD Variable Speed Drive

Dymola Dynamic Modelling Laboratory WWV Western Wisdom Valley FCU Fan Control Unit Xair Air humidity (relative humidity)

(11)

-11-

1 Introduction

This chapter is the introduction of this thesis project including the background of the energy situation of the world and China, the motivation and highlight of this simulation project as well as the thesis structure.

1.1 Background

1.1.1 Worldwide Energy Efficiency Situation

Within the last 10 years, energy consumption increased by more than 100Mtoe per year with annual increase of 1.5%, this number further increased to 2.9% in 2018. Among all the countries, China, America and India together consumed at least 2/3 of global produced energy and this in turn pushed the electricity generation rising by average 3.7% annually. Although the coronavirus, COVID-19, hit the global economy in the first quarter of 2020 resulting in a 3.8% decrease of global energy demand. From a long term of view, the consumption still shows a predictable raising since population and the scale of industry keep enlarging. The improvement in energy efficiency is negligible due to the continuous rising of energy-intensive industries. According to the account of IEA (International Energy Agency), from 2015, the increasing rate of energy efficiency slowed down by just 1.2% in 2018 [1]_{. This is due to the fact that although technologies}

and processes are becoming more efficient, structural factors like people’s lifestyle and energy-intensive industries are dampening the impact of these technical efficiency gains on energy demand.

The issue of energy efficiency is even worser in the industry and building sectors. Data from IEA show final energy usage in buildings grew from 2080 Mtoe (2010) to 3060 Mtoe (2018) in 8 years[2]_{. The booming of}

energy services in buildings sector – particularly cooling and heating demand, appliances and connected devices – weighs much more than the improvement of energy efficiency. This results in an increase of CO2

emissions from building sector and got to a record in the year 2018 to 3Gt CO2. On the other hand, there

exist millions of factories running with a large amount of low-efficiency equipment that needs to be improved or replaced by new technologies. From the global overview, the share of large economies with high energy intensity improvement goes down, which means the increase of economic output region with less efficient industries.

Therefore, enhancement in energy efficiency, implement of sustainable industries, management in energy consumption and energy-saving projects have become vital issues for every country in the world.

1.1.2 Situation in China

Since 2005, the total primary energy consumption of China has increased from 1829.5 Mtoe to 3247.9 Mtoe in the year 2018. Meanwhile, the annual decreasing rate of energy intensity was around 4% while recently showed a slow-down trend (only 3.1% in the year 2018)[3]_{. Among all sectors, industry and residential sectors}

consume much higher than the other sector in China, according to the data from IEA, they occupied 49.3% and 16.7% of the energy consumption of China in 2017[4]_{. In 2018, the majority energy consumption of}

industry sector in China came from chemistry, construction materials, iron and steels, and nonferrous metals. With the increasingly significant effect of supply-side structural reform in China, improvement of production efficiency based on industrial structure optimization has become the most important driver to reduce the energy intensity. While in the residential sector, energy intensity per unit area is rising year by year, due to people’s lifestyle as well as comfort service in rooms. Thus, energy-saving from the construction envelope, Heating, Ventilation and Air Conditioning (HVAC) system and energy management become much more important with the development of materials science, efficient technologies and digitalization. In China, more than 15% of electricity is consumed for cooling of China’s entire society, which increases nearly 20% every year on the average. China’s HVAC industry has an annual output value of more than 100 billion dollars and creates over 3 million jobs. Nowadays, China manufactures more than 80% of world’s

(12)

-12-

total residential air conditioners (ACs) and more than 60% of world’s total refrigerators. According to the ‘Green and High-Efficiency Cooling Action Plan’ [5,6]_{published by IGSD, China will focus on the energy}

efficiency levels of HVAC products and targets to improve the efficiency of cooling products on the market by 30% by the year 2020.

In China, the electricity load for Air Conditionings (ACs) in large and medium-sized cities represents 60% of the summer peak electricity load. Hence there is a large potential of the energy-saving for air conditioner and affiliated devices, and the energy efficiency of main cooling products can be improved by 30-50%. That’s the reason why more and more energy companies put their eyes on energy-saving retrofit businesses. Since the publication of the 12th Five-year Plan for Energy-saving and Emission Reduction (2013) by the Chinese government, the energy-saving retrofit business developed rapidly among China. At the same time, the simulation for HVAC systems also developed quickly, due to the little primary investment and easiness to operate. The simulation results make great sense for energy-saving retrofit projects.

1.2 Motivation

The processes of energy-saving and cooling projects are complex, related to the on-site investigation, data collection, system designing, equipment selection, etc. To implement this project, high investment, time cost and trial-and-error cost are needed. Under such a situation, it is necessary to set up the technical-financial model and an energy-efficient system dynamic model to evaluate the energy-saving potential and maximum benefit with appropriate investment. With the simulation of HVAC and energy-efficient systems, engineers hence can project the variation of the environment in buildings, including heat acquire and loss, heating and cooling load, indoor climate and human comfort, and then put forward proper energy-saving suggestions. The simulation for HVAC systems also benefits industrial businesses. This can help the company cut the first investment cost, promote the digitalization development, and offer a visual energy solution to customers. What’s more, the knowledge about the mutual connections among physical parameters of the HVAC system also has the guidance meanings to the engineering projects. There is plenty research about energy simulation of chiller plant systems. However, most of them are lack of the systematic analysis of devices behavior as well as the cooperation among devices. Also, the optimized operation strategies proposed by these research usually focus on devices themselves lacking of analyzing cooling load distribution, which may result in the mismatch of supply and demand balance and energy loss.

The purpose of this master’s thesis is to develop a system-oriented dynamic model for an industrial chiller plant by the Modelica framework based software Dymola. The model is to evaluate the energy consumption performance under different operation optimization strategies, determining for guiding the implementation of energy-saving retrofit processes. The goals of this thesis project are as follows:

 Set up a system-oriented dynamic model of a chiller plant based on the real industrial situation and weather conditions;

 Model validation and accuracy checking;

 Analyze the behavior of each devices of the chiller plant under SO and HO strategies under full range of Cooling Load Ratio (CLR), aiming to evaluate the energy saving performances under 2 strategies;

 Based on the HO strategy mentioned above, analyse Cooling Load Time Frequency (CLTF) as well as Cooling Load Distribution (CLD) and further employ 2 advanced HO strateiges: HOTF strategy and HOCL strategy for comparison,seperately;

 Analyze the energy saving performance as well as behavior of devices under the three HO strategies with two-month cooling load data.

(13)

-13-

1.3 Highlights

This thesis work aims for the dynamic modelling based on an industrial chiller plant, evaluate and verify the energy saving potentials of different operation strategies. In recent years, researchers have developed different dynamic models for chiller plant system to investigate the optimal system configurations and controlling logics with enhanced energy performance. However, most of these papers focus on partial equipment and split them out of whole system, resulting the ignorance of the impact and cooperation between devices. In addition to this, the researches of energy saving performance under optimization strategies usually focus on the simulation results in the format of case study, which means, the systematic research about the impact of devices behaviors on energy consumption within full range of cooling load is of lack.

This thesis work in first develops a comprehensive system-oriented dynamic model for general industrial chiller plant. Then, the simulation runs under 3 common industrial operation strategies (Baseline Strategy, Sequence Optimization Strategy, and Holistic Optimization Strategy) within full range of the cooling load data. Finally, 2 more advanced holistic optimization strategies (HOTF strategy and HOCL strategy) are introduced. Along with basic holistic optimization (HOB) strategy, all three HO strategies are simulated using 2-months load data and compared with SO strategies to evaluate their energy saving potentials. The innovation of this thesis work is written as follows:

 System-Oriented Dynamic Model for the Industrial Chiller Plant: The thesis work develops a dynamic chiller plant model based on on-site investigation data and designed layout, and the simulation model covers the whole system instead of using only one or two devices. So, the thesis work is feasible and has practical guiding significance to a real project.

 Devices Behavior and Energy Savings Research of the Chiller Plant within the Full CLR: Instead of only analyzing the energy saving potential of SO strategy and HO strategy, this thesis work also pays attention to the behaviors of every device when the simulation runs in the full range of cooling load ratio, in order to reveal the interaction among chillers, pumps and cooling towers of an industrial chiller plant.

 Intruduce 2 advanced Holistic Optimization Strategies: Based on cooling load time frequency analysis and cooling load distribution analysis, this thesis work introduces two advanced HO strategies, HOTF strategy and HOCL strategy.

1.4 Structure of Thesis

The research of this thesis project is to model and simulate the chiller plant system for 2 buildings located in the zone of Western Wisdom Valley in Chengdu City, Sichuan Province. A system-oriented dynamic model is built up according to the on-site investigation situations including the layout, devices configurations, etc. The accuracy of this simulation model is checked with real-time weather conditions and cooling load. Then, the model is employed to evaluate the devices behavior and energy saving performance under three common operation strategies. Finally, the research introduces two advanced holistic optimization strategies and evaluate their impact on energy saving performance with 2-months simulated results. The structure of this thesis is as follows:

 Chapter 1 is the introduction of the thesis. In this chapter, the background of energy efficiency situation in both China and worldwide is described to show the necessity of HVAC system retrofit projects. Based on the Modelica Language and Dymola software, this thesis project is implemented and aims for evaluating the energy performance of different operation strategies for industrial chiller plants.

 Chapter 2 is the theory and literature review of this thesis working. First, the concept of High Efficiency Chiller Plant is introduced as the basic knowledge. Then, the two simulation processes

(14)

-14-

of HVAC simulation are introduced, steady state approach and dynamic approach. The next section gives a brief introduction to Modelica and Dymola. Finally, recent researches of chiller plant dynamic modelling are introduced, including the research of cooling load ratio, sequence optimization and holistic optimization strategies;

 Chapter 3 introduces the methodology of this thesis project. Firstly, the modelling process and accuracy check process of this system-oriented dynamic model for this industrial chiller plant are introduced. Secondly, the simulation runs under 3 common operation strategies within full range of the CLR, so that the energy saving performance as well as devices behavior would be evaluated. Finally, the further exploration of Holistic Optimization is implemented, 3 HO strategies with 2-months load data, to evaluate their energy-saving performance;

 Chapter 4 gives the system description of this thesis project. Firstly, the layout and environment conditions are acquired by on-site investigation. Then, the system configurations are introduced for accuracy check, common operation strategies and further exploration on holistic optimizations;

 Chapter 5 introduces the modelling process of this thesis project. The process is divided into 2 parts. The first part is physical devices modelling, including chillers, water pumps and cooling towers; and the second part is controlling logic modelling, including sequence control, mass flow rate control, temperature difference control and valve opening control. Then, the connection bus is programmed and set up. The final step is to combine two parts with the interaction bus to become a whole chiller plant system model with accuracy check;

 Chapter 6 gives the simulation results of this thesis project. The first section compares the simulation results among Baseline Strategy (BS), Sequence Optimization (SO) Strategy, as well as Holistic Optimization (HO) Strategy. The simulation results cover the full range of cooling load, to evaluate the energy performance and devices behavior under different Cooling Load Ratio (CLR). The second section compares the simulation results among Basic Holistic Optimization (HOB) Strategy, Holistic Optimization based on Cooling Load Time Frequency Analysis (HOTF) Strategy, as well as the Holistic Optimization based on Cooling Load Distribution Analysis (HOCL) Strategy;

 Chapter 7 is the discussion of this thesis project. In this Chapter, the simulation results of this project are compared with the results of other researches, in order to discuss the pros and cons of this simulation project;

 Chapter 8 is the last chapter of this thesis report, including the conclusion of this thesis project, as well as the future work for this chiller plant simulation task.

(15)

-15-

2 Literature Review

In this Chapter, the theory of this simulation project is introduced. Section 2.1 introduces the concept and indicators of high efficiency chiller plant. Then, the HVAC simulation process is introduced in Section 2.2. There are two methods for system modelling, steady state approach and dynamic approach. Due to the time variable character of chiller systems operation, the dynamic modelling has obvious advantages with its reliability and numerical stability. In Section 2.3, Modelica based dynamic modelling tool, Dymola, is introduced to be used for this thesis project due to its superior characteristics. Recent research on chiller plant dynamic modelling are also introduced in Section 2.4, including the research of cooling load ratio, sequence optimization and holistic optimization strategies.

2.1 High Efficiency Chiller Plant

2.1.1 Concept

The energy in public constructions is mainly consumed in the form of electricity. Among them, the major energy-consuming equipment includes central air conditioning, water supply and drainage, lighting, elevator and other systems. Specifically, the energy consumption of the central air conditioning system accounts for about 40%~60% of the energy consumption of the whole building. Therefore, a high-efficiency chiller plant for air conditioning system is the key of energy conservation in public buildings.

High-Efficiency Chiller Plant is to improve the energy efficiency of the air conditioning system through in-depth design, and to optimize working status. In this process, an energy-efficient controlling system with related digital solutions developed by energy simulation tools and 3D modelling tools plays a central role in the smart connection to the chiller plant. This efficient connection drives related equipment runs under an optimized mode to decrease the energy consumption for the whole system. Compared with the general chiller plant, the advanced high-efficiency chiller plant enhanced the operation efficiency by 20% to 50% and meanwhile cut the cost by 30%[7]_.

2.1.2 Indicators

The first step to enhance the operating efficiency of a chiller plant is to determine how efficiently the plant is now operating. The evaluation refers to several indicators, while CPE is a direct indicator to value the efficiency of the whole chiller plant.

CPE =_𝑊 𝑄𝑐𝑜𝑜𝑙𝑖𝑛𝑔

𝑐ℎ𝑖𝑙𝑙𝑒𝑟+𝑊𝑐ℎ𝑖𝑙𝑙𝑒𝑑 𝑝𝑢𝑚𝑝+𝑊𝑐𝑜𝑜𝑙𝑖𝑛𝑔 𝑝𝑢𝑚𝑝+𝑊𝑐𝑜𝑜𝑙𝑖𝑛𝑔 𝑡𝑜𝑤𝑒𝑟 (2-1)

In the equation 2-1, CPE means chiller plant efficiency,

Qcooling is the output cooling capacity of the chiller plant, with the unit KW;

Wchiller is the power consumption of the chillers, with the unit KW;

Wchilled pump is the power consumption of chilled pumps, with the unit KW;

Wcooling pump is the power consumption of cooling pumps, with the unit KW;

Wcooling tower is the power consumption of cooling towers, with the unit KW.

The optimization target of the central air conditioning system is to minimize the energy consumption to large extent with the capacity to cover the cooling load at the same time (keep Qcooling without change). Thus,

the solution is to decrease the total energy consumption of every device:

min𝑃𝑘 = 𝑊𝑐ℎ𝑖𝑙𝑙𝑒𝑟+ 𝑊𝑐ℎ𝑖𝑙𝑙𝑒𝑑 𝑝𝑢𝑚𝑝+ 𝑊𝑐𝑜𝑜𝑙𝑖𝑛𝑔 𝑝𝑢𝑚𝑝+ 𝑊𝑐𝑜𝑜𝑙𝑖𝑛𝑔 𝑡𝑜𝑤𝑒𝑟 (2-2) In this equation, Pk is the total power consumption of all the equipment involved, with the unit of KW.

(16)

-16-

CPE which is the COP (coefficient of performance) of whole chiller plant, means how much of cooling capacity can be offered by the chiller plant when 1 KW electricity has been consumed. The cooling performance can also be valued in KW/ton, measuring the total cooling capacity in KW divided by the tons of refrigerant, this can also be calculated from the value of COP divided by 3.517.

As the figure 2.1 shows [8]_{, when the value of refrigerant tons is higher than 0.85 KW/ton (4.13KW), the}

efficiency of the chiller is at a good level, and the excellent level means this value higher than 0.7 KW/ton (5.02KW). If the value is less than 1 KW/ton (3.517KW), the chiller plant needs to be improved.

Figure 2.1 High Efficiency Chiller Plant Standard

2.1.3 Energy-saving Retrofit for A Chiller Plant

As mentioned before, central air conditioning is one of the indispensable and important operating systems for modern buildings, including hotels, shopping malls, industrial factories, etc. For a large amount of industrial or commercial zones, the equipped HVAC systems are less energy efficient with poor maintenance resulting in huge energy waste. What’s more, the universal application of variable frequency technology in China, one of the important energy-saving technologies for the HVAC system, still lags compared with that applied in western countries. Thus, it is of great meaning for China to develop the energy-saving retrofit project, especially the application of variable frequency technology, to enhance the energy efficiency of the central air conditioning systems[9]_.

As the most energy-consumption and the most important component, the energy-saving retrofit for a chiller plant has huge energy-saving potentials. By considering the devices chosen process, the chillers, chilled water pumps, cooling water pumps as well as cooling towers, they generally are designed to cover the largest amount of capacity under the worst cooling load working conditions. However, the controlling systems are not always considered to be as smart as thought, as a result, the devices usually work with a full load. In the real situation, these devices should be working under the non-full load conditions in most of the time, e.g., during the evening or cloudy days, which would lead to a large sum of energy waste.

To realize the efficient and continuous operation of the whole air-conditioning room, it is necessary to implement the energy-saving retrofit project, which is a systematical work of fine design, intelligent operation and maintenance. It includes energy-saving HVAC equipment selection, chiller plant system optimization and energy-saving maintenance.

There are several retrofit solutions for a chiller plant[10]_:

 Upgrade chillers: Select chillers with magnetic bearings and variable speed compressors, which reduce internal friction; allowance for modulation; and the ability to operate at variable speeds instead of simply “on” or “off.” This will increase efficiency at times when the cooling system is at partial load.

 Install High Temperature-Drop Terminal Units: Select terminal units with an entering water temperature as high as possible.

(17)

-17-

 Optimize Heat Rejection: Cooling towers should be selected for the smallest overall size and power consumption for greatest efficiency; Install variable speed fans in cooling towers to save energy during mild weather.

 Properly Size Water Pumps: Oversized pumps cannot be modulated to enough low speeds with poor efficiencies and have increased maintenance issues. Install correctly sized pumps paired with variable frequency drives to maximize energy-savings.

 Install Smart Controllers: Chilled water plants typically perform best when centrally controlled through a building management system (BMS).

Through the system dynamics modelling and simulation, the operation data of the system is extracted for research and analysis, to find the optimal operation status for the minimum of energy consumption. Meanwhile, the equipment configuration and control strategy of the chiller plant need to be matched well with its design parameter and loading characteristics. The air conditioning system is a dynamic changing system due to the correlation among the devices, which means it is no possible to enhance the efficiency of the chiller plant only by the manual operation. By only relying on the smart control system and adjusting all the devices according to the real-time operating status can make the chiller work efficiently. This is the value of energy-saving retrofit.

2.2 HVAC Simulation Process

Modelling and simulation are a cost-effective way to evaluate the design and operation of cooling systems. Modelling refers to the process that the real physical system is represented with mathematical models. Different physical systems (thermal, electrical, and electromagnetic, etc.) with different time-scaled dynamics are involved.[11]_{This usually leads to high-indexed differential-algebraic equations. Simulation is then}

conducted to numerically solve the mathematical equations, which involves computer representation of models, different numerical solvers, solution procedures etc.

Simulating the annual energy consumption of a cooling system has become a key strategy in designing high-efficiency HVAC system that can meet the needs of the industry. The automated exchange data between the energy management platform and energy simulation program is also essential to the future of the HVAC designing process.

Compared to the field experiment, simulation has several advantages [12]_:

 It could control factors that cannot be controlled in a field experiment (e.g., outside weather conditions) to isolate the effects of occupancy;

 Sole consequences of a control strategy on energy consumption can be determined in simulation;

 Calculating energy consumed by an HVAC system might be difficult in a field experiment due to infeasibility of control or lack of metering;

 Simulation is less expensive and less time consuming;

To achieve maximum energy efficiency, and HVAC control strategy that accounts for occupancy must be designed on a case-by-case basis and in the context of specific physical and functional characteristics of a building, where simulation can play a pivotal role.

There are lots of simulation tools have been developed to compute and simulate the HVAC systems. For example, Energy Plus[13,14]_{, eQuest}[15]_{, IDA ICE, etc. Generally, these simulation tools are developed based}

on imperative programming languages such as FORTRAN, C/C++ etc. Most simulation tools currently used in the HVAC community, such as BLAST, DOE-2, and EnergyPlus, have detailed dynamic models for the heating and cooling loads, but do not include dynamic models for practice.

A classification of models and simulation tools can also be done with regard to the internal treatment of transient input signals[16]_:

(18)

-18-

 steady state approach: The algebraic equation system fae (Eq. (2-3)) with unknown variables xa is

solved for each time step, assuming fixed boundary conditions i (inputs) between two time points tn and tn+1. Hence follows a stepwise constant output o.

𝑓𝑎𝑒(𝑥𝑎, 𝑖, 𝑜, 𝑡) = 0 𝑤𝑖𝑡ℎ 𝑖(𝑡) = 𝑐𝑜𝑛𝑠𝑡. 𝑓𝑜𝑟 𝑡𝑛≤ 𝑡 ≤ 𝑡𝑛+1 (2-3)

 dynamic or transient approach: All unknown variables xd, xa and boundary conditions i may be

changing any time. The system model can be transformed in a differential-algebraic-equation (DAE) system fDAE (Eq. (2-4)), whereas the transient behavior of dominating variables xd are

described by differential equations. Comparatively fast model parts xa can be modelled by means

of algebraic equations (quasi-static approach), which are balanced continuously with the dynamic model part.

𝑓𝐷𝐴𝐸(𝑥𝑑̇ , 𝑥𝑑, 𝑥𝑎, 𝑖, 𝑜, 𝑡) = 0 (2-4) Even though static approach is able to simulate the HAVC systems and predict the operation cost, while the data stability as well as flexibility would be affected when the boundary conditions are variable. On the other hand, since the dynamic approach implicitly results in a higher effort, for instance costs for model design and simulation, the question for the beneficial aspects of this method may arise.

One answer to this question is basically resulting from the mathematical difference between both ways of modelling. Since a transient behavior of a capacity of any kind can only be described by a mathematical expression like Eq. (2-4) an accurate physical solution for a problem with changing boundary conditions can just be provided by the dynamic simulation. The advantages of dynamic simulation for HVAC systems are shown as following:

 the key variables to be investigated are strongly influenced by the systems capacities,

 control strategies or even suitable control settings shall be investigated,

 increased reliability of the results because of conservative balance equations combined with the display of short transient peaks,

 enhanced numerical stability for models containing unsteady equations and signal feedback at the same time,

 utilization of one common tool for a broad range of applications in the development, planning and optimization process of complex energy systems.

2.3 Modelling and Simulation Tools

2.3.1 Common Dynamic Modelling Tools

In terms of simulation of the HVAC equipment dynamics, a mixture of ordinary differential equations (ODE), partial differential equations (PDE), and differential algebraic equations (DAE) may be required (Wetter 2009a). Most simulation tools currently used in the HVAC community, such as BLAST, DOE-2, and EnergyPlus, have detailed dynamic models for the heating and cooling loads, but do not include dynamic models for the equipment to be interconnected for each other. Apart from this, EnergyPlus also adopts idealized controls to reduce computation time[17]_{. Although TRNSYS has dynamic control models,}

its constant time step poses numerical challenges[18]_{. Further, conventional tools often intertwine model}

equations and numerical solvers in their source codes; this makes it difficult to extend these programs to support control-oriented cases.

Some transient modeling tools have been developed based on Matlab/Simulink® and Modelica. The Thermosys Toolbox[19]_{, originally developed by the University of Illinois now jointing The Texas A&M}

University, is a Matlab/Simulink-based transient modeling tool for air-conditioning and refrigeration systems. Based on block diagram simulation platforms such as Simulink®[20] _{that involve block-diagram type}

(19)

-19-

control the dynamic behaviors of HVAC systems. Such systems are built on input-output components (i.e., observing the causality principle), however, these systems do not reflect relations between physical variables in algebraic equality (e.g., due to mass, momentum and energy balances).

After the introduction of the Modelica® language in 1997[21]_{, equation-based and object-oriented modeling}

has attracted enormous attention from both academia and industry. In the past decade, Modelica has proved great capability for simulating multi-physical systems, through various engineering applications, especially for large, complex, and hybrid systems. The Modelica modeling paradigm, due to its equation based nature, accommodates the causality among system variables, and thus can well embrace the DAE systems that otherwise would be difficult to handle. In addition, simulation platforms such as Dymola (2013) provide a rich library of numerical solvers for PDE, ODE, and DAE systems, as well as algorithms of index reduction that facilitate handling simulation of multi-scale systems. The equation-based paradigm also allows a nearly “decoupled” or “parallel” framework for model development: the numerical solvers are developed and updated by applied mathematicians, while high-quality physical models are constructed by HVAC engineers. Therefore, HVAC engineers can focus on improving the dynamic models, while being relieved of the challenges from selecting and implementing up-to-date solvers.

2.3.2 Modelica and Dymola

The Modelica language has been developed by the Modelica Association, a non-profit organization based in Linköping, Sweden, which gathers members from Europe and the US. It is an ambitious modelling language that has shown potential to bring order to the fragmented world of DAE-based simulation. It draws on the collective experience of a large number of first-generation languages[22]_.

Moreover, Modelica is a freely available, object-oriented language for modelling of large and complex systems. The programming language is mainly used for computer simulation of dynamic systems where behavior evolves as a function of time. It has been designed to deal with multi-domain modelling; this means that several aspects of physics can be treated in the same model.

Since the first Modelica derived tool, Dymola, appeared in 1999, several large industries such as Toyota, Ford, United Technologies, ABB, etc. have adopted it. Efforts to develop building and HVAC system simulation models resulted in various Modelica libraries, such as ATPlus and Building Informatics Environment.

Dymola (Dynamic Modelling Laboratory), a commercial environment, has been developed and kept updating since 1992 by Swedish company Dynasim AB, acquired by the French Dassault Systems in 2006[23]_.

It can be compatible with other programs like CATIA, Simulink, or Excel.

As a result, in this project, the platform 3DExperience R2021x HotFix 1.5 is used, with the application of Dymola inside the platform. Three Modelica Commercial Libraries are needed, Modelica Standard Library – Version 3.2.3, Human Comfort Library – Version 2.8.0 and HVAC Library – Version 2.8.0.

2.3.3 Characteristics of Modelica Language There are four notable characteristics as following[24]_:

 Objective-oriented Modelling

The objects of Modelica language adopt the concept of ‘component’ which corresponds to a single element/subsystem/part in the actual project. Therefore, a complex system can be divided apart into small components. Each component has its physical attributes. After a single component is determined, the relationships among these components can be further determined to model the final functions of the entire system. For these single objects, the establishment of their models is relatively simple. Models simulate specific physical processes in mathematical forms (equations). They not only describe the concrete existing systems in real life but also are applied to describe abstract things. A simple decomposition of the refrigeration system components is illustrated in Figure 2.2. The model can be divided into internal part and

(20)

-20-

external part. The internal part means the interactions of every component inside the model, as well as the functions, and there is no need to take much care. Instead, users need to focus more on the external part. The external part of the model is the interface part, which is used to act on other objects.

 Declarative Physical Modelling

The components in the thermodynamic system influence and restrict each other. For example, the chiller and the cooling tower, the chiller affects the work of the condenser, and the cooling tower also affects the work of the compressor. This mutual influence usually needs to be expressed by implicit equations. Many advanced simulation programs such as Simulink use block diagrams, a sequential modeling technique, which specifies the corresponding input and output for each model. However, there is no clear input/output relationship between components in a thermodynamic system, which is not suitable for block diagram expression. Moreover, the block diagram model needs to specify certain boundary conditions [25]_.

Figure 2.2 Disintegration of a Chiller Plant Model

Declarative physical modelling means that there is no specific causal relationship between objects. The behavior of the model can be directly written with the general physical description equations, and the interconnection and function of the model components are as same as the actual physical systems. This method greatly improves the flexibility of components. With this mode, a single component is just like a standard part produced on an assembly line so that it can be accessed at any time on demand. At the same time, it also means that the modelling process of a complicated system can be easy to establish and understand, hence the time for debugging is also greatly reduced.

 Multi-domain Modelling

Modern products often involve multiple physical fields, such as inverter air conditioning systems. Most parts come from the thermodynamics field, the compressor drive motor belongs to the electrical field, and the mechanical port belongs to the mechanical field. The frequency control strategy belongs to the control field. Modelica unify the expressions of parameters in multiple physical fields, building multi-field models. Modelica uses mathematical equations to describe the behavior of components to reflect the internal properties of the components themselves. The behavior of the component is represented by the external port, that is, the connector. The connector itself does not define equations or formulas, but only defines the object that the component acts on the outside [26]_.

The connection between two connectors needs to be of the same type, which means the defined parameters in the connectors need to be consistent. Components from the same field of discipline can be easily connected, and components from different fields are also possible to be connected by corresponding connectors, as long as the existing common parameters. For example, the compressor model of an air conditioner is usually from thermodynamic field but has a mechanical port. And a motor also has a mechanical output port because it belongs to the electrical field. So, the two components can be connected by a flange in the mechanical field to realize the connection between different domains. It is completely consistent with the actual system.

 Continuous-discrete hybrid modeling

In actual physical systems, there are often both continuous time-varying and discrete time-varying characteristics. For the modelling of continuous systems, differential-algebraic equations can be used to describe in Modelica language. For discrete systems, it mainly describes events that occur at certain points in time. Modelica uses differential equations, state tables, Petri nets and other forms to describe discrete

(21)

-21-

events. Models built with Modelica contain both continuous and discrete characteristics, which is hybrid modelling.

In summary, it can be seen that Modelica has a wider range of applications than traditional modelling languages. And from the user's point of view, Modelica's object-oriented declarative modelling is more intuitive and easy to understand.

2.4 Researches of Chiller Plant Systems

In recent years, some researches have studied dynamic modelling of chiller plants for HVAC systems. Dynamic modeling of a chiller plant is a subject of particular importance for control system design and fault detection and diagnosis, while the transient behaviors of the associated processes are, in principle, very complicated and features strong interactions among multiple physical domains. A chiller plant system includes chillers, cooling towers, chilled water pumps, cooling water pumps and other related devices. In order to study the effect of one or two devices performance in whole chiller plant, scientists usually employ dynamic model for individual device.

Chan and Yu [27]_{developed an air-cooled chiller system model, using the operating data to analyze how the}

chiller components interact with each other and discussed how the analysis helped modelling floating condensing temperature control to improve system performance. Deng et al. [28]_{presented a dynamic model}

of air-to-water dual-mode heat pumps screw compressor having four-step capacities. The dynamic responses of adding additional compressor capacity in step-wise manner were studied. Koury[29]_{et al.}

proposed a model for a refrigeration system with distributed parameter model for heat exchangers. Numerical simulations were carried out to verify the possibility of controlling the refrigeration system and the superheating of the refrigerant in the evaporator outlet by varying the compressor speed and the throttling valve position. Most of these research focus on one or two devices of the chiller plant and investigate the effect of partial components performance to the whole system. The dynamic modelling for the whole chiller plant including all devices and the performance analysis for the whole system is seldom developed.

2.4.1 Cooling Load Ratio Research

Chiller Part Load Ratio (PLR), is the ratio between chiller cooling capacity at specific conditions (Qcl, KW)

and rated chiller cooling capacity (Qcr, KW). According to the characters of chillers, the COP of chillers

vary with different PLR values. The PLR could be calculated according to the following formula:

PLR = 𝑄𝑐𝑙

𝑄𝑐𝑟 (2 − 5) For a chiller plant system, different numbers of chillers with diverse rated capacities can be combined and mixed to bear the cooling load of a building. Yu and Chan [30]_{found that, the optimized number and size of}

chillers for satisfying the cooling demand of a building can reduce the energy consumption. The COP performance curve of chillers used in their project is depicted as follows:

(22)

-22-

It is shown in the figure that the chiller COP is increasing with the rising of PLR. So, the station COP could be enhanced by improving the PLR value of chillers during the operation hours through the different device mixes (figure 2.4). What’s more, the increased number of switching steps also improves the energy efficiency of applied constant speed water pumps (figure 2.5). The assessment showed that an electricity saving of 10.1% can be achieved by installing a chiller plant with six chillers of three different sizes instead of four equally sized chillers, which is shown in the table 2.1.

Figure 2.4 Chiller COP at Different CLR for OP1-OP4 -

Yu’s Project Figure 2.5 Chilled Water Pump Consumption at Different CLR for OP1-OP4 - Yu’s Project However, chillers’ COP curve may not be monotonic increasing, which means the highest COP value may not at the PLR=1. In such situation, the method mentioned above may not be the most optimal way to enhance the energy efficiency of a chiller plant. Even though there is 10.9% energy saving for water pumps in their project, the saving ratio is not high enough compared with the method that replacing the constant speed pumps to variable speed pumps. What’s more, purchasing lots of devices and frequently switching devices may result in a complex and expensive chiller plant system.

2.4.2 Sequence Optimization Research

Based on Cooling Load Time Frequency (CLTF) and Probability Density analysis, Bo and Jin [31]_introduced

a method to optimize the operation strategy of chiller plants. Firstly, the probability density curve of their project was calculated and depicted as figure 2.6. Also, the cooling load distribution was divided into 7 ranges according to different chiller mixes (table 2.2).

Table 2.2 Cooling Load Range - Yu’s Project

Then, the switching sequence of chillers was optimized according to the probability density, which is depicted in the figure 2.7.

Figure 2.6 Probability Density Distribution of Cooling Load

Ratio - Bo’s Project Figure 2.7 Operation Sequence of Chillers based on the Strategy - Bo’s Project The figure 2.8 shows the results of their project. It is clearly that, Strategy AⅠ, AⅡ and AⅢ can decrease the power consumption of chillers by 1.4%, 3.4% and 4.3%, respectively. This means the cooling load time frequency and probability density analysis can help to produce advanced operation strategy for chillers.

(23)

-23-

Figure 2.8 Power Consumption of Chillers under AⅠ, AⅡ and AⅢ Strategies - Bo’s Project

However, this research only aimed for the majorization of chillers, and didn’t related to the power consumption of other devices of a chiller plant system.

2.4.3 Holistic Optimization Research

In the last ten years, researchers put their eyes on the advanced controlling logics and operation strategies to a chiller plant. Instead of controlling individual devices, the control mode usually covers all devices and multi-parameters, which is named as holistic control.

Xiao and Yaoyu[32]_{developed a dynamic model of a chilled water system for the study of self-optimizing}

operation using extremum seeking control (ESC). Their Modelica based dynamic simulation model demonstrated the effectiveness of proposed control strategy, and the potential for energy saving was also evaluated. Muzaffar Ali[33] _{analyzed the energy performance of a chilled water system with different}

configurations and controlling logics based on dynamic simulation. Combined with Modelica/Dymola modelling program and GenOpt generic optimization program, the optimal system configuration is given according to different cooling loads ranging from 1000KW to 7000KW. This method gives optimal configurations for minimizing the energy consumption with different cooling loads. However, the operation status of every devices is changing time to time, and the optimal configuration would be also changed with times. As a result, the energy consumption may not be minimum for the whole period. Variety of research studies have focused on developing energy-efficient control schemes for multiple chiller plants. However, few studies have conducted optimization of the chiller plant with non-identical number of chillers based on dynamic energy modeling approach [34]_,_{Also, in most situations, chiller plants are not working in an optimal}

strategy, because the chiller plants are either operated at designed condition irrespective of the cooling load or optimized locally due to the lack of overall chiller plant behavior. Sundar et.al, [35]_{developed an overall}

energy model of a chiller plant in Singapore, and investigated the impact of two kinds of optimization strategies: Chiller Optimization (CO) strategy and Energy Optimization (EO) strategy on the energy consumption compared with Conventional Method strategy. It is a chiller plant with 2 centrifugal chillers in working. The operation decisions under the 3 strategy are listed in the table 2.3.

Table 2.3 Operating Decision of Chiller Plant - Sundar’s Project

Compared with Chiller Optimization Strategy, the Energy Optimization Strategy also optimized condensing and evaporating water temperature to decrease the power consumption, and VSD were also applied to chilled water pumps and cooling tower fans to adjust the mass flow rate. Figure 2.9 and 2.10 show the water flow rate and set temperature of different devices under the three strategies of their project. Here, the water pumps are equipped with VSD to save the energy.

(24)

-24- Figure 2.9 Water Flow Rate of Energy Optimization

Methodology - Sundar’s Project Figure 2.10 Water Temperature Set-Point of Energy Optimization Methodology - Sundar’s Project

Figure 2.11 and table 2.4 show the total energy consumption of the chiller plant under 3 strategies in their project. Due to the optimization, both CO and EO strategies have more than 30% of energy savings, and most of saving potentials lies on the chillers. For EO strategy, the consumption of water pumps is also heavily reduced.

Table 2.4 Performance Measures of Chiller Plant with Two Centrifugal Chillers – Sundar’s Project

Figure 2.11 Total Energy Consumption of Chiller Plant under Different Operation Strategy – Sundar’s Project

However, from the flow rate curve in the above figure, the cooling load distribution is stable for their project. Compared with another project employing wide range of cooling load distribution, the saving potentials may be different. Thus, it is necessary to analyze the behavior of every device under full range of cooling load distribution. What’s more, the optimization options could be dozens. For some industrial projects, it has no possibility to change the set point of water temperature due to the production process requirement. For other optimization options, such as water temperature difference, can be used to save the energy consumption.

(25)

-25-

3 Methodology

In this Chapter, the methodology of this thesis project is introduced. The first step is to solve the questions raised in Chapter 1 by developing a system-oriented dynamic model for this industrial chiller plant, including modelling of devices and controlling logic. Then the accuracy of this model is checked by running the simulation based on the real data of system configurations and environment conditions. Secondly, the simulation runs under 3 common operation strategies (Baseline, Sequence Optimization and Holistic Optimization) among the full range of Cooling Load Ratio (CLR), so that the energy saving performance as well as devices behavior would be evaluated. Finally, the further exploration of Holistic Optimization is implemented. Two advanced HO strategies, HOTF strategy and HOCL strategy are introduced to be compared with Basic Holistic Optimization (HOB) strategy. After this, simulation runs under 3 HO strategies during two summer months cooling load data, in order to evaluate the energy saving performance of the 3 strategies compared with Sequence Optimization Strategy.

3.1 Dynamic Modelling Process

3.1.1 Chiller Plant Modelling

Figure 3.1 shows the dynamic modelling process of the target chiller plant system. The whole system model is divided into 2 parts, physical models and controlling logic models.

For physical models, component models including chiller, water pump and cooling tower are called from HAVC Commercial Library in Dymola. Then, these component models are modified according the needs of this thesis project. According to the layout and onsite situation of this chiller plant, subsystems are modelled. The final step is to connect these subsystems according to the operation logic to construct a chiller plant system. At the meantime, necessary sensors including temperature sensors, mass flow sensors are also installed to monitor the medium status of the system.

Figure 3.1 Methodology of Dynamic Modelling Process

Controlling logic models are established according to the energy-saving retrofit planning as well as common industrial operation optimization methods. To realize the controlling optimization of the chiller plant model, there are four basic units of controlling logic models: sequence control model, temperature difference control model, mass flow rate control model, as well as valve opening control model. By mixing these four units, a group controlling model is set up for the holistic optimization of this chiller plant. The function of this group controlling model is then checked to verify the functional correctness.

(26)

-26-

After modelling the physical part and controlling logic part, an interface bus is programmed as connection between these two parts. Finally, the system configuration, environment and load data are set as the input. So far, the chiller plant dynamic model has been established successfully.

3.1.2 Accuracy Check

In order to make comparison between the simulation results from modelling and the measurement data, the error indicators are used for this process. These error indicators include absolute error (AE), bias error (BE), mean absolute error (MAE), mean bias error (MBE), relative mean absolute error (RMAE) and relative mean bias error (RMBE). The related formula to calculate these error indicators are as below [36]_:

3.2 Industrial Optimization Strategies

As mentioned before, the operation optimization for a chiller plant have different approaches. For most industrial projects, chiller plants are operated manually, which is Baseline Strategy (BS). Sequence optimization of devices is commonly used for industrial project to save the energy consumption of a chiller plant. More advanced, some projects also optimize water flow rate, chilled water and cooling water temperature to promote the collaborative operation among different devices. This optimization strategy is called Holistic Optimization (HO) strategy.

These 3 strategies are commonly used in industrial projects. It is obvious energy consumption of HO strategy is much less than that of HO strategy and BS strategy. However, the saving ratio and the behavior of devices under 3 strategies are worth investigating when the chiller plant is operating among the full range of cooling load.

Table 3.1 Operating Decisions under 3 Common Operation Strategies

Operating decisions Baseline _Strategy _OptimizationSequence _OptimizationHolistic Devices ON/OFF

(Chillers, pumps and cooling towers) ON Varied Varied

Chilled water flow rate Fixed Fixed Varied

Cooling water flow rate Fixed Fixed Varied

Temperature difference of Chilled water Varied Varied Fixed

Temperature difference of Cooling water Varied Varied Fixed

Air flow in the cooling tower Fixed Varied Varied

Table 3.1 shows the operational decisions adopted by holistic optimization strategy, along with baseline strategy and sequence optimization strategy. Here, the decisions of baseline refer to the designed condition of the chiller plant. The evaluation method accounts all devices of this system influencing the power consumption and the quality of supplied cooling capacity. The optimized decisions are on/off status for