| 학회 | 한국화학공학회 |
| 학술대회 | 2021년 가을 (10/27 ~ 10/29, 광주 김대중컨벤션센터) |
| 권호 | 27권 2호, p.1485 |
| 발표분야 | 공정시스템 |
| 제목 | Optimal operation of ash deposits removal system in the pulp process using a machine learning |
| 초록 | This work, we proposed optimal sequence of operating sootblower for ash deposits removal using a machine learning to maximize recovery boiler efficiency. The proposed modified Q-learning algorithm derived the Q-matrix which is a function that predicts the expected dynamic reward (priority for deposits removal) of performing a given action (sootblowing) and a given state (sootblowing location). The reward is sequentially updated through reward update matrix which consider the decrease heat transfer rate according to ash deposits. To calculate decrease heat transfer rate, the computational fluid dynamics (CFD) model is developed for temperature prediction of each sootblowing locations. In additions, based on the mathematical equation of heat transfer rate and the predicted temperature, reward is defined considering the deposit thickness growth rate and thermal conductivity of ash deposits. To demonstrate the effectiveness of this study, the process model was developed to predict power generation. As a results, the power generation increase by 214 kW without any retrofitting of boiler through the optimal sequence. Keywords: Ash deposits, Sootblower, Modified Q-learning algorithm |
| 저자 | 조형태1, 임종훈1, 김유림1, 박현도1, 김태복2, 박한신2, 김정환1 |
| 소속 | 1한국생산기술(연), 2무림피앤피(주) |
| 키워드 | 공정최적화; 공정제어; 인공지능 기반 공정기술 |
| 원문파일 | 초록 보기 |
