폐수처리공정에서의 인공지능기반 알고리즘 성능 향상 연구 = A Study on Improving the Performance of Artificial Intelligence Based Algorithms in Wastewater Treatment Processes

다국어 초록 (Multilingual Abstract)

Considering wastewater properties from the rapid development of industrial growth, it is necessary to have an optimal treatment system for this wastewater to mitigate the environmental impact of discharges and comply with environmental regulatory standards. The quality of industrial wastewater effluent varies depending on factors such as wastewater source and treatment process (physical, chemical, biological). Governing equations and data-based prediction models are used to determine treatment facilities' process characteristics and behavior. However, a lot of analysis is required to comply with environmental regulations. Therefore, a typical wastewater treatment facility's operation relies on experienced people and their knowledge. The 4th Industrial Revolution, the next generation industrial revolution, consisting of the convergence of information and communication technology (ICT), has already been introduced to Korea and worldwide in various fields such as artificial intelligence (AI), the Internet of Things (IoT), and big data. However, in the domestic environmental technology field, the research and development of artificial intelligence technology is insufficient compared to other fields. Although various methods of applying artificial intelligence technology to wastewater treatment facilities are being proposed abroad, research and development in Korea have not yet been conducted on various water treatment systems. Also, there is a lack of standardized data-based model development cases and guidelines that can be optimally applied. Therefore, to apply artificial intelligence technology to wastewater treatment facilities, it is necessary to utilize the results derived from various studies and their comparison and analysis. Recently, the prediction models that can forecast the effluent water quality from wastewater treatment have been actively developed from AI-based machine learning and deep learning models and can be widely applied to wastewater treatment processes.
In this study, we compared and analyzed the performance of machine learning algorithms (SVM, RF, ANN) and deep learning algorithms (LSTM) using the daily flow rate and BOD concentration of effluent from the wastewater treatment process. In addition, we compared and analyzed the performance of a similar wastewater treatment process. The applicability of transfer learning was analyzed in two wastewater treatment processes.
This study assessed the performance of single and modified algorithms based on machine learning and deep learning for wastewater treatment process. More specifically, this study adopted support vector machine (SVM), random forest (RF), and artificial neural network (ANN) for machine learning as well as long short-term memory (LSTM) for deep learning. The performance of these (single) algorithms were compared with that of modified ones processed through hyperparameter tuning, ensemble learning (only for machine learning), and multi-layer stacking (i.e., two layers of LSTM units). The daily effluent of wastewater treatment process observed between 2017 and 2022 in the Cheong-Ju National Industrial Complex was used as input to all tested algorithms, which was evaluated with respect to mean squared error. For the model performance evaluation, discharge and biochemical oxygen demand are selected as dependent variables out of nine measured parameters. Results showed that the performance of any machine learning algorithms was superior to their competitor LSTM. This is mainly attributed to a small amount of input data provided to the LSTM algorithm and unstable effluent wastewater characteristics. Meanwhile, hyperparameter tuning improved the performance of all tested algorithms. However, ensemble learning for machine learning and two-layer stacking for LSTM generally resulted in performance degradation as compared to that of single algorithms, regardless of dependent variables. Therefore, this calls for a careful design and evaluation of modified algorithms, specifically for model architecture and performance improvement processes.
This study assessed the feasibility of transfer learning from one wastewater treatment process to another using two popular deep learning algorithms. Specifically, convolutional neural network (CNN) and long short-term memory (LSTM), which consisted of four and three hidden layers, respectively, were used as benchmark algorithms for transfer learning. Input data for both deep learning and transfer learning were provided from two wastewater treatment plants with identical treatment trains in series (located in Jinju and Cheongju City) over the five-year period from 2018 to 2022. Performance evaluation was also done not only against two deep learning algorithms but also against those adopting two transfer learning strategies, one for freezing all hidden layers developed from the pre-trained model and the other for training the last hidden layer only among multiple ones, with respect to Mean Squared Error (MSE). We found that the performance of both CNN and LSTM was relatively comparative regardless of dependent variables, discharge and biochemical oxygen demand (BOD), whereas the prediction accuracy of both algorithms was slightly higher for discharge than for BOD due to its low variability. When transfer learning which froze all hidden layers of the existing model was applied to two benchmark algorithms, the predictive performance of both algorithms was found to slightly improved only for discharge. Also, there was no measurable variation in the prediction accuracy of benchmark algorithms using the other transfer learning approach. Potential applications of transfer learning include the rapid reuse of the existing models (developed from source domains) for target domains which are hard to develop new prediction models due to the lack of data in deep learning.
The results derived from machine learning, deep learning, and transfer learning in wastewater treatment can be used to develop data-based model development standardization guidelines. This might be due to that the primary data could be used in future intelligent wastewater treatment algorithms and scenarios.

목차 (Table of Contents)

• Ⅰ. 서론 1
• 1. 연구의 배경 및 필요성 1
• 2. 연구의 목적 4
• 3. 논문의 구성과 내용5
• Ⅱ. 이론적 배경 7
• 1. 전국 공공폐수처리시설 운영현황7
• 1) 공공폐수처리시설 현황7
• 2) 공공폐수처리시설 종류 및 운영현황9
• 3) 공공폐수처리시설 설치 효과 9
• 4) 폐수 배출특성 및 배출허용기준 10
• 5) 공공폐수처리시설의 방류수 수질기준 10
• 2. 연구대상 폐수처리시설 시설 및 운영현황 12
• 1) C 산업단지 폐수처리시설 12
• 2) J 산업단지 폐수처리시설 14
• 3. 수질예측 연구 변천 과정 18
• 1) 활성슬러지공정모델(Activated Sludge Model) 18
• 2) 인공지능 활용 24
• 4. 인공지능의 발전 29
• 5. 기계학습 연구 38
• 1) 기계학습(Machine learning, 머신러닝) 알고리즘 분류 38
• 2) 선형회귀(Linear regression, LR) 39
• 3) 서포트 백터 머신(Support vector machine, SVM) 42
• 4) 의사결정 트리(Decision tree) 44
• 5) 랜덤 포레스트(Random forest, RF) 45
• 6) 앙상블(Ensemble)기법 46
• 7) 파이썬(Python) 프로그램 50
• 8) 케라스(Keras) 51
• 6. 기계학습 알고리즘 성능 분석 52
• 1) R2 (Coefficient of determination) 52
• 2) MSE (Mean squared error) 53
• 3) RMSE (Root mean square error) 53
• 4) MAPE (Mean absolute percentage error) 54
• 7. 심층학습(Deep Learning_딥러닝) 연구 55
• 1) 심층학습 개념 55
• 2) 심층학습 알고리즘 56
• 3) 최적화 알고리즘(Optimization algorithm) 73
• 8. 전이학습(Transfer Learning) 76
• 1) 전이학습의 개요 76
• 2) 전이학습 방법 79
• Ⅲ. 폐수처리 공정에서의 기계학습 및 심층학습 알고리즘 성능비교 84
• 1. 서론 84
• 2. 실험방법 85
• 1) 폐수처리공정 85
• 2) 단일 알고리즘 87
• 3) 변형 알고리즘 88
• 3. 결과 및 고찰 90
• 1) 단일 알고리즘 성능 비교 90
• 2) 매개변수 조정 후 성능 비교 91
• 3) 변형 알고리즘 성능 비교 92
• 4. 소결론 93
• Ⅳ. 동일한 폐수처리 공정에서의 전이학습 적용 유무에 따른 심층학습 알고리즘 성능평가 95
• 1. 서론 95
• 2. 실험방법 97
• 1) 폐수처리공정 선정 및 입력 데이터 구축 97
• 2) 벤치마크 알고리즘 100
• 3) 전이학습 103
• 4) 모델 개발 및 평가 환경 103
• 3. 결과 및 고찰 104
• 1) 심층학습 알고리즘 성능 104
• 2) 전이학습 알고리즘 성능 105
• 3) 전이학습 적용 방법 변화에 따른 성능 106
• 4. 소결론 108
• Ⅴ. 종합결론 110
• 참고문헌 112