광산배수 처리를 위한 기계학습 기반 소석회 투입량 예측 연구

초록

In this study, we predicted the lime dosage for treatment facilities (ST and HT) using machine learning (ML) to effectively design and operate mine drainage treatment facilities. After removing the bad data from the original data, the flow rate related to the lime dosage, and metal ion (Fe, Mn, Al) concentrations and the pH which are highly related to OH‒ supplied by lime, were selected as inputs. The convolutional neural network was used, and the data was augmented using fancy principal component analysis to compensate for the limited and imbalanced dataset while maintaining the characteristics of the original data. The test dataset prediction by the ML model demonstrated a lower mean absolute error and higher coefficient of determination (R2) than that of the theoretical calculation equation. The ML application is expected to enhance operational effectiveness of treatment facilities and offer fundamental design data for new facilities.

키워드