A METHOD OF PERFORMING REINFORCEMENT LEARNING FOR THE SIMULTANEOUS PRECISE CONTROL OF VENTILATION DEVICES AND AIR CONDITIONERS IN CLASSROOMS AND A COMBINED VENTILATION-COOLING AND HEATING SYSTEM USING THE SAME

The present invention relates to a method of performing reinforcement learning for simultaneous precise control of ventilation devices, air conditioners, and heaters in classrooms and a composite system for ventilation, cooling, and heating using the same. The present invention can basically operate...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: PARK JEONG HO, YOON MYUNG SUP, YOON WON SIK, SEO SANG MIN
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present invention relates to a method of performing reinforcement learning for simultaneous precise control of ventilation devices, air conditioners, and heaters in classrooms and a composite system for ventilation, cooling, and heating using the same. The present invention can basically operate a cooling/heating system to enter a comfort zone in a psychrometric chart in a thermal environment to allow an occupant to feel comfortable, and operate a ventilation device to satisfy an indoor contaminant maintaining standard and a recommendation standard of the government at the same time. The present invention applies a deep reinforcement learning algorithm using a tree compensation system in the composite operation of ventilation, cooling, and heating to simultaneously consider energy cost aspects and identify an operation pattern (remote control status of a user) preferred by the user to allow an optimal composite operation of ventilation, cooling, and heating. Specifically, the present invention applies a systematic tree compensation technique to the deep reinforcement learning algorithm to overcome temporal difficulty which may require several months for artificial neural network training in virtual simulation techniques in previous studies and suggests the concept for a real-environment contamination source simulation demonstration chamber as in the present invention to allow artificial neural network training and demonstration for a composite system of ventilation, cooling, and heating in the simulation environment. 본 발명은 교실에서의 환기장치 및 냉난방기 동시 정밀제어를 위한 강화학습 수행방법, 이를 이용한 복합 환기 냉난방 시스템에 관한 것으로, 기본적으로 재실자가 쾌적함을 느낄수 있도록 열환경 선도조건(Psychrometric chart)에서의 컴포트 존(Comfort Zone)에 들어올 수 있도록 냉난방 시스템을 작동할 수 있으며, 동시에 정부의 실내오염물질 유지기준 및 권고기준을 만족할 수 있도록 환기장치를 운전할 수 있다. 이러한 환기 및 냉난방 복합운전시 트리보상체계를 활용한 심층강화학습 알고리즘을 적용하여 에너지 비용적인 측면을 동시에 고려하고, 사용자가 보다 더 선호하는 운전패턴(사용자 리모컨 제어여부)을 파악하여 최적의 복합 환기 냉난방 운전이 가능하게 한다. 특히, 강화학습 알고리즘에 체계적인 트리 보상기법을 적용함으로써, 선행 연구들에서 가상 시뮬레이션 기법상 인공신경망 학습에 몇 달이 소요될 수도 있는 시간적 어려움을 극복하고, 본 발명에서와 같이 실환경 오염원 모사 실증챔버에 대한 개념을 제시하여 상기 모사 환경에서의 복합 환기 냉난방 시스템에 대한 인공지능 학습 및 실증이 가능하게 한다.