베이지안 최적화 기반 모델 참조 적응제어를 적용한 가변속 냉동시스템의 강인제어 = Robust Control of Variable Speed Refrigeration System Using Bayesian Optimization Based Model Reference Adaptive Control

다국어 초록 (Multilingual Abstract)

A compressor constant speed system and a variable speed refrigeration system(VSRS) are widely used in cooling systems such as the oil cooler system(OCS). Traditional constant speed systems, using on/off or hot-gas bypass methods, suffer from low energy efficiency and limited temperature stability. In contrast, VSRS can significantly improve energy efficiency and thermal precision by controlling the electronic expansion valve(EEV) and compressor speed simultaneously. However, VSRS is inherently difficult to model accurately due to frequent load fluctuations, nonlinearities, and large time constants, which often limit the effectiveness of conventional fixed-gain controllers like the PID. The purpose of this study is to design a model reference adaptive controller(MRAC) optimized for VSRS and to propose a new "simplified-MRAC"(S-MRAC) structure. Conventional MIT rule-based MRAC includes a pure integrator in its adaptation rule. In systems like VSRS where residual errors persist due to slow thermal responses, this integrator can lead to an excessive accumulation of errors, resulting in parameter windup and control chattering. Furthermore, while hybrid-MRAC(H-MRAC) structures using auxiliary PI controllers offer some stability, they often mask the core adaptive performance and create a dependency on additional gain tuning. To overcome these structural limitations, this study proposes the S-MRAC architecture, which removes the auxiliary PI controller to verify the standalone robustness of the adaptive law. The design eliminates the pure integrator in the adaptation rule to prevent unintended parameter divergence. Additionally, a low-pass filter(LPF) was integrated to filter out high-frequency noise and mitigate chattering in the manipulated variables, ensuring smooth operation of the inverter and EEV. To ensure a systematic design process, the adaptation gains were determined using Bayesian Optimization(BO), which probabilistically explores the parameter space to satisfy predefined performance constraints without manual trial-and-error. The validity of the proposed S-MRAC was demonstrated through MATLAB/Simulink simulations and experimental verification on an actual OCS testbed. The results showed that the controller maintained stable tracking of the reference model even under significant model uncertainties and external heat load disturbances. By effectively suppressing the interference between the oil outlet temperature and superheat, the S-MRAC achieved high-precision control that outperforms conventional methods. This study is significant in providing a robust and easy-to-implement adaptive control framework for complex multi-variable refrigeration systems, highlighting the potential for maintenance-free operation in industrial environments.

목차 (Table of Contents)

• 제1장 서 론 1
• 1.1 연구 배경 및 목적 1
• 1.2 연구 범위 및 내용 4
• 제2장 가변속 냉동시스템의 제어 7
• 2.1 가변속 냉동시스템의 구성 7
• 2.2 실험 장치의 구성 및 설계 사양 10
• 2.3 가변속 냉동시스템의 동특성 분석 13
• 제3장 모델 참조 적응 제어기 구조 및 설계 20
• 3.1 모델 참조 적응 제어기(MRAC) 개요 및 기본 구조 20
• 3.1.1 Hybrid MRAC(H-MRAC) 기본 구조 및 설계 21
• 3.1.2 H-MRAC의 문제점 22
• 3.2 Simplified MRAC(S-MRAC)의 구조 및 설계 23
• 3.2.1 S-MRAC의 참조 모델 설계 25
• 3.2.2 매개변수 업데이트를 위한 MIT 규칙 26
• 3.2.3 Lyapunov 안정성 평가 30
• 3.3 Bayesian 기법을 이용한 S-MRAC 설계 파라미터 최적화 32
• 3.3.1 Bayesian 최적화 개요 32
• 3.3.2 Bayesian 최적화 프로세스 33
• 제4장 시뮬레이션 결과 및 고찰 35
• 4.1 시뮬레이션 조건 및 방법 35
• 4.2 Matlab 기반의 H-MRAC의 시뮬레이션 결과 및 고찰 37
• 4.3 Matlab 기반의 S-MRAC의 시뮬레이션 결과 및 고찰 38
• 4.4 PI 제어기와의 제어성능 비교 40
• 4.5 S-MRAC의 설계 파라미터 영향 분석 42
• 4.6 S-MRAC의 모델 불확실성 영향 분석 46
• 4.7 Low Pass Filter(LPF)의 영향 분석 49
• 제5장 실험 결과 및 고찰 51
• 5.1 실험 조건 및 방법 51
• 5.2 H-MRAC의 실험 결과 및 고찰 52
• 5.3 S-MRAC의 실험 결과 및 고찰 54
• 5.4 PI 제어기와의 제어성능 비교 55
• 제6장 결 론 58