| 논문명 |
에너지 실시간 모델 예측 제어를 위한 인공신경망 변수 최적화 / Optimized neural network for Real time Model-based Predictive Control / 2-4 : 빌딩시뮬레이션 I |
| 저자명 |
김영민(Kim, Young-Min) ; 김용세(Kim, Yong-Se) ; 안기언(Ahn, Ki-Uhn) ; 박철수(Park, Cheol Soo) |
| 주제어 |
모델 예측 제어 ; 반응표면법 ; 인공신경망 ; 유전알고리즘 ; Model Predictive Control ; Response Surface Method ; Neural Network ; Genetic Algorithm |
| 요약2 |
Recently, Model-based Predictive Control (MPC) based on a data-driven model are highlighted for building energy savings. The data driven model is based on a correlation between inputs and outputs instead of relying on the first principles. The reason for the data-driven model is its simplicity: requiring far less information, time and assumptions of the model. In this study, Artificial Neural Network (ANN) is employed for the MPC. One of the problems for the MPC using ANN lies in tuning ANN parameters(the number of hidden layers and nodes). However, there is no explicit and definite solution to select the optimal ANN parameters. For the study, the authors used a response surface method to identify the ANN’s behavior according to the parameters with the given data. Then, a Central Composite Faced-design (CCF) was applied to tune the number of nodes. With the predictions using the optimized ANN, GA was used to solve the optimal control problem for MPC. |
