검색 상세

Performance-based Thermal Comfort Controls using Machine Learning : Performance-based Thermal Comfort Controls using Machine Learning

머신 러닝을 활용한 성능 기반 열 쾌적 제어

윤영란 (Young Ran Yoon, 단국대학교 건축공학과 건축환경및설비)

원문보기

  • 주제(키워드) 도움말 Machine learning , Thermal comfort , performance-based
  • 발행기관 단국대학교
  • 지도교수 도움말 문현준
  • 발행년도 2018
  • 학위수여년월 2018. 8
  • 학위명 박사
  • 학과 및 전공 도움말 대학원 건축공학과건축환경및설비
  • 세부전공 건축환경및설비
  • 원문페이지 126p
  • 본문언어 영어
  • 저작권 단국대학교 학위논문은 저작권에 의해 보호받습니다
초록 moremore
With the recent increase in energy consumption in buildings, energy-saving strategies in buildings have become a priority for the energy policies of many countries. Therefore, many recent research studies have emphasized on advanced control methods to attain comfortable thermal conditions while mini...
With the recent increase in energy consumption in buildings, energy-saving strategies in buildings have become a priority for the energy policies of many countries. Therefore, many recent research studies have emphasized on advanced control methods to attain comfortable thermal conditions while minimizing the energy consumption in buildings. One of the most promising approaches is model predictive control, which enables explicit accounting to minimize energy consumption while maintaining conditions that provide thermal comfort within a certain range. However, this approach has limitations such as the necessity of a mathematical model of the building, restricted control strategies, and uncertainty. To address these problems, a new performance-based thermal comfort control method that considers context is required. This method should enhance the efficiency of the thermal comfort control process by using a new modeling technique. In recent years and for many buildings, various types of indoor/outdoor environmental data such as temperature, relative humidity, and air velocity have been collected by building energy management systems. These data reflect the context (i.e., physical characteristics, behavior patterns of the occupants, and environmental information) of each room. An inverse model (or data-driven model) that utilizes these data would be capable of predicting the environmental characteristics of the room and occupant status, which could be used for real-time thermal comfort control. Therefore, we propose a data-driven approach for real-time occupant status and thermal comfort prediction in this report. The results obtained using this method were evaluated using a newly proposed data-driven model that reflects changing situational information in real time using a machine learning technique, i.e., a deep Q-learning algorithm. The results show that the proposed indoor environmental data-driven model for occupant status detection using the classification algorithm is suitable. In addition, the Gaussian-process-based thermal comfort prediction model used in this study was confirmed to have high accuracy, including uncertainty information. Finally, by applying deep Q-learning, we verified the effectiveness of control based on the developed thermal comfort performance indicator. The results confirm that the proposed approach is more efficient than set-temperature control in terms of providing thermal comfort and reducing energy consumption.
목차 moremore
I. INTRODUCTION 1
1.1 Background 1
1.2 Objectives and Hypotheses 5
...
I. INTRODUCTION 1
1.1 Background 1
1.2 Objectives and Hypotheses 5
1.3 Organization of Thesis and Research Scope 7
II. NEW APPROACH FOR PERFORMANCE-BASED CONTROL 11
2.1 Limitations of Current Occupant Status Detection Methods 11
2.2 Existing Thermal Performance Indicators 13
2.3 Proposed Methods for Performance-based Control 21
III. DEVELOPMENT OF OCCUPANT STATUS DETECTION MODELS 25
3.1 Overview 25
3.2 Experimental Procedure and Data Acquisition 26
3.3 Classification Algorithm for Occupant Status 32
3.4 Results and Evaluation 39
3.5 Summary 46
IV. GAUSSIAN PROCESS REGRESSION 48
4.1 Overview 48
4.2 Gaussian Process Model 49
4.3 Gaussian Process Model for Buildings 55
4.4 Dealing with Uncertainties in Inputs for Thermal Comfort 58
V. GAUSSIAN PROCESS MODEL FOR THERMAL COMFORT 60
5.1 Overview 60
5.2 Chamber Experiment 61
5.3 Evaluation of Applicability in Real Buildings 76
5.4 Application of Probabilistic Thermal Comfort for Performance-based Control 81
VI. DESIGN AND EVALUATION OF PERFORMANCE-BASED COMFORT CONTROL 86
6.1 Introduction 86
6.2 Reinforcement Learning for Performance-based Comfort Control 88
6.3 Performance-based Thermal Comfort Control (PTCC) Evaluation Results 99
6.4 Summary 109
VII. CONCLUSIONS 110