유역관리를 위한 심층신경망 수질 농도 예측 모델 개발에 관한 연구

Abstract

This study developed the ‘deep neural network(DNN)’ model combined with several non-linear functions part of neural network algorithm to predict BOD in inflow sewage the day that can overcome the limitations of water treatment power plant(WTPP) applied in the past mechanistic and statistical methods and the complex inflow characteristics behavior. To estimate DNN model, this study developed multi linear regression(MLR) model using stepwise method and principal component analysis(PCA) each. And using derived variable as a days ago and two days ago water charactersitic. It has related because of continuity of sewage. In result, DNN model used derived varible has the highest predictive power. Also, DNN and MLR model’s prediction are increased when used derived variable. Through this sutdy, derived variable as past water characterisitc are verified to forecast BOD and make a basis to use DNN model in WTPP.