Title page for etd-0710120-153850

URN	etd-0710120-153850	Statistics	This thesis had been viewed 134 times. Download 2 times.
Author	Yu-Sung Chang
Author's Email Address	kobe3536@yahoo.com.tw
Department	Institute Of Mechanical Engineering
Year	2019	Semester	2
Degree	Master	Type of Document	Master's Thesis
Language	zh-TW.Big5 Chinese	Page Count	94
Title	The Diagonal recurrent neural network applied to temperature control of variable refrigerant flow air conditioning systems
Keyword	Energy saving controls Variable frequency air conditioning systems Variable refrigerant flow Diagonal recurrent neural networks Diagonal recurrent neural networks Variable refrigerant flow Variable frequency air conditioning systems Energy saving controls
Abstract	The modern air conditioning systems apply variable refrigerant flow (VRF) techniques to be the VRF air conditioning systems for Energy saving, which also called variable frequency driver air conditioning systems in market. The VRF air conditioning systems can change the refrigerant flow rate by changing the compressor speed, and usually applied to home and small office space. The modern VRF air conditioning systems use variable speed compressors with nonlinear characteristics, and applied to the varied room space and refrigerating load requirements, that leads the conventional PID control becomes insufficient for the requirement of modern VRF air conditioning systems. This study establishes the dynamic model of VRF air conditioning system and utilizes the multi-layer neural networks (MLNN) and diagonal recurrent neural network (DRNN) applied to temperature control for VRF air conditioner system, that can enhance the adaptability and energy saving. The multi-layer neural network applied to the temperature control of VRF air-conditioning system. The simulation results show that use the sigmoid function as activation function can converge stably with satisfied adaptability and effectively control the indoor temperature. The DRNN has simple structure with effective learning strategy and fast convergent speed. The DRNN can be applied to build diagonal recurrent neural network controller (DRNC), which can be widely applied to nonlinear dynamic system identification and control. The simulation results reveal that the DRNC can enhance the adaptability and be available to room temperature control with better energy saving performance than conventional PID control.
Advisor Committee	Ming-Hui Chu - advisor
Files	etd-0710120-153850.pdf indicate in-campus access immediately and off-campus access at 2 years
Date of Defense	2020-07-02	Date of Submission	2020-07-12

Back to Results