ABSTRACT This graphical abstract illustrates the proposed framework for daily-scale prediction of effluent total phosphorus (T-P) and sensitivity-based optimization of coagulant dosing in a full-scale DAF system, integrating data preprocessing, machine learning, and operational decision support. Effective phosphorus control in dissolved air flotation (DAF) systems is essential for regulatory compliance in full-scale wastewater treatment plants, yet operational decisions are often constrained to daily time scales due to limited online sensing. This study proposes an interpretable, feature-engineered machine-learning framework for daily prediction of effluent total phosphorus (T-P) and sensitivity-based optimization of coagulant dosing in a full-scale municipal DAF system (410,000 m3/day). Long-term operational, water-quality, and meteorological data (1,096 days) were preprocessed using three-sigma outlier screening and multivariate imputation by chained equations (MICE). Mechanistically informed features capturing influent loading, operational conditions, short-term variability of effluent T-P (1–3 day difference-based features), and seasonal effects were incorporated. Among the evaluated models, Random Forest achieved the best performance (Test R2 = 0.818; RMSE = 0.032 mg/L), corresponding to a prediction error within 20% of the discharge limit (0.2 mg/L). SHAP analysis identified influent T-P, coagulant dosage, and short-term variation as dominant drivers across seasons. A sensitivity-based autoregressive simulation indicated that optimized dosing could reduce coagulant consumption by 32–51%, yielding an estimated annual cost saving of 1.53 billion KRW while improving effluent stability. The proposed framework demonstrates the practical value of daily-scale, interpretable machine learning for data-driven DAF operation.

ABSTRACT This graphical abstract illustrates the proposed framework for daily-scale prediction of effluent total phosphorus (T-P) and sensitivity-based optimization of coagulant dosing in a full-scale DAF system, integrating data preprocessing, machine learning, and operational decision support. Effective phosphorus control in dissolved air flotation (DAF) systems is essential for regulatory compliance in full-scale wastewater treatment plants, yet operational decisions are often constrained to daily time scales due to limited online sensing. This study proposes an interpretable, feature-engineered machine-learning framework for daily prediction of effluent total phosphorus (T-P) and sensitivity-based optimization of coagulant dosing in a full-scale municipal DAF system (410,000 m3/day). Long-term operational, water-quality, and meteorological data (1,096 days) were preprocessed using three-sigma outlier screening and multivariate imputation by chained equations (MICE). Mechanistically informed features capturing influent loading, operational conditions, short-term variability of effluent T-P (1–3 day difference-based features), and seasonal effects were incorporated. Among the evaluated models, Random Forest achieved the best performance (Test R2 = 0.818; RMSE = 0.032 mg/L), corresponding to a prediction error within 20% of the discharge limit (0.2 mg/L). SHAP analysis identified influent T-P, coagulant dosage, and short-term variation as dominant drivers across seasons. A sensitivity-based autoregressive simulation indicated that optimized dosing could reduce coagulant consumption by 32–51%, yielding an estimated annual cost saving of 1.53 billion KRW while improving effluent stability. The proposed framework demonstrates the practical value of daily-scale, interpretable machine learning for data-driven DAF operation.

Chlorophyll-a (Chl-a) is a key indicator of eutrophication and algal blooms in freshwater ecosystems. This study aimed to improve Chl-a prediction through a novel integrative approach that combines the optical and structural characteristics of dissolved organic matter (DOM) with machine learning and explainable artificial intelligence techniques. Weekly data were collected over one year (Apr 2023–Apr 2024) from four sites in the Yanghwa Stream, South Korea. Machine learning models (Random Forest, Gradient Boosting Machine, and Extreme Gradient Boosting) were developed using 34 input variables encompassing 15 traditional water quality parameters, 8 ultraviolet–visible (UV–Vis) optical indices, and 11 liquid chromatography–organic carbon detection (LC–OCD) components. Extreme Gradient Boosting showed the best performance (R 2 = 0.7318, RMSE = 4.3827). SHapley Additive exPlanations analysis identified total nitrogen, total organic carbon, spectral slope ratio (S R ; ratio of S 275–295 to S 350–400 ), and Molecularity (nominal average molecular weight of the humic substances fraction derived from LC-OCD) as major contributors, highlighting the importance of DOM composition in Chl-a variability. Counterfactual analysis further revealed model-guided adjustment scenarios for reducing high Chl-a concentrations to below the threshold (12.2 mg/m 3 ; 75th percentile), emphasizing the importance of variables such as total nitrogen, pHosphate, Molecularity, and S R as key indicators within the learned model. In conclusion, this integrated approach combining DOM characteristics with explainable artificial intelligence techniques enhances prediction accuracy and interpretability and provides specific and practical strategies for the effective control of Chl-a in freshwater systems.

상수관로의 노후화는 수질 안전성 저하와 수자원 손실, 유지보수 비용 증가 등의 문제를 야기하며, 이에 따라 지중 매설관의 상태를 신속하고 정확하게 진단할 수 있는 기술의 중요성이 커지고 있다. 특히 내시경 영상(CCTV)을 활용한 관로 점검은 가장 보편적인 방식으로 자리 잡았으나, 판독자의 숙련도에 따라 해석 편차가 발생하고, 대량 데이 터의 신속한 처리에는 한계가 있다. 이러한 배경에서 본 연구는 관종⋅관경⋅용도 등 상수관 메타데이터를 모델에 통합하고, 관로 내 결함의 존재 여부와 유형, 크기를 동시에 예측할 수 있는 다중과제 학습(Multi-task Learning) 기반 인공지능 모델을 제안한다. 제안한 모델은 두 개의 예측 헤드를 통해 결함 판별과 정량적 분류를 병행하도록 설계되었으며, SHAP 기반 분석을 통해 모델의 판단 근거가 상수관로의 실제 결함 특성과 일치함을 확인하였다. 이러한 접근은 수작업 판독의 부담을 경감하고, 관로 상태 기록의 표준화 및 정량화를 통해 예방 중심의 유지관리 전략 수립을 효과적으로 지원할 수 있다.

Water quality incidents in drinking water distribution networks require rapid operational decision-making to limit contaminant propagation and service disruption. This study proposes a hydraulically informed operational shutdown framework that integrates incident occurrence time and utility response time to support real-time decision-making during water quality emergencies. Network-based forensic analysis is applied to estimate contaminant travel times under variable demand conditions and to identify valve isolation actions that are feasible under realistic field constraints. To translate analytical results into actionable operations, the concept of Possible Operated Segments (POS) is introduced, defining the smallest isolation units that can be practically actuated during emergency response. In addition, a Cause–Impact–Duration (CID) analysis is employed to quantify location-specific impacts and the time-dependent escalation of service disruption. By jointly considering hydraulic propagation dynamics and operational feasibility, the proposed framework enables utilities to evaluate when and where intervention is most effective. Application to representative turbidity scenarios demonstrates that the framework provides clear guidance for prioritizing valve operations, containing incident impacts, and improving response timeliness, highlighting its potential as a practical decision-support tool for enhancing emergency response capability and operational resilience in drinking water distribution systems.

인천 적수 사고를 비롯한 수돗물 수질 사고가 연이어 발생하면서, 안전하고 깨끗한 물 공급의 중요성과 함께 상수도관망의 체계적인 유지관리에 대한 필요성이 증대된 실정이다. 이에 본 연구에서는 현장 세척 기록과 수리해석 및 관로 제원 등 데이터를 통합하여 상태가 취약한 상수관로의 최적 플러싱 주기를 결정하는 모델을 제안하고자 한다. 이를 위해 본 연구에서는 문헌 검토를 통해 상수관로의 취약성을 평가하기 위한 6가지 핵심 지표로 평균 유속, 관 연령, 비내식성 밸브 산재 구역, 수질민원 발생 구역, 잔류염소 및 탁도 기준 미달 구간, 법정 세척 이력 부재 구역이다. 이러한 지표를 기반으로 상수관의 상태를 4가지 등급으로 분류하였다. 먼저, S시(2013∼2014 년)의 플러싱 데이터를 사용하여 관로 길이, 관경, 연령을 모델의 독립변수로 고려하여 최적 세척 주기를 예측하는 다중 회귀 모델을 개발했다. 이 모델은 통계적 유의성(R = 0.617, p < 0.05)을 갖으며, 다중공선성은 없는 것으로 나타났다. 개발한 모델의 적용성을 검토하기 위해 B시의 대규모 블록 구역에 모델을 적용한 결과, 취약한 관로가 관망의 말단 및 저유속 지역에 집중되어 있는 것으로 나타났다. 또한, 관경이 작고(100∼150mm) 노후된 배관은 일반적으로 월별 또는 분기별 주기로 더 짧은 세척 간격이 필요한 것으로 도출되었다. 이처럼 세척 간격이 짧은 구간은 세척을 정기적으로 수행하기 보다는 경제성 평가를 통해 개량 의사결정을 수립하는 것이 적절하다고 판단된다. 종합적으로 본 연구에서 제시한 취약관로 상태평가 기준 및 세척 주기 산정모델은 기존의 일률적 세척 주기 운영 방식의 한계를 보완하고, 관로별 상태에 따라 차등화된 유지관리 전략 수립에 기여할 수 있는 정량적 의사결정 도구로서의 활용 가능성을 확인하였다. 향후 다양한 지역에서의 세척 데이터 축적과 현장 실증을 통해 모델의 외삽성을 검증하고, 제한된 예산, 인력, 장비 등 현실적 제약조건을 고려한 최적화된 세척 스케줄링 기법으로의 확장이 필요할 것으로 판단된다. 이를 통해 정기적 세척뿐 아니라 예산 효율성, 유지관리 비용, 세척 효과 분석을 통합적으로 고려하는 종합적 의사결정 지원 모델로 발전시킬 수 있을 것으로 기대된다.

A machine learning-based classification model was applied to identify the main influencing factors affecting O3 advisory (triggered when hourly average O3 concentrations exceed 0.12 ppm), using existing 25 urban air quality monitoring networks data from Seoul and meteorological data from Seoul automatic weather station (Jongno-gu). From May to September 2023, data were collected and analyzed. The dataset comprised 19 variables, including urban air quality metrics (such as O3, PM2.5, PM10, NOX) and meteorological parameters (such as wind speed, temperature, relative humidity, rain probability, and cloud cover), recorded on an hourly basis. Using this data, two classification models were developed: the first model (analysis model, ANM) employed decision tree and random forest algorithms to identify the main influencing factors affecting high O3 concentration events. The second model (prediction model, PRM) was designed to predict the likelihood of O3 advisory for the following day. Through the application of ANM, the main influencing factors affecting high O3 concentration were identified, with PM2.5, PM10, and temperature emerging as significant variables affecting O3 advisory. And both decision tree and random forest models have demonstrated strong classification performance. These results indicate that the models effectively classified the data into category 0 (no O3 advisory) and category 1 (O3 advisory). Additionally, a second classification model (PRM) was developed to predict the likelihood of O3 advisory in Seoul for the following day. This model utilized seven independent variables: temperature, relative humidity, rain probability, cloud cover, and forecasted air quality levels (PM2.5, PM10, O3). Overall, these findings suggest that PRM is a viable tool for predicting next-day O3 advisory. In this study, the application of the proposed classification model methodology based on real-time air quality and meteorological data for a given region is expected to quantitatively explain the performance of PRM and be usefully utilized in reducing O3 exposure for sensitive and vulnerable populations.

Numerous studies have established a correlation between sociodemographic characteristics and water usage, identifying population as a primary independent variable in mid- to long-term demand forecasting. Recent dramatic sociodemographic changes, including urban concentration-rural depopulation, low birth rates-aging population, and the rise in single-person households, are expected to impact water demand and supply patterns. This underscores the necessity for operational and managerial changes in existing water supply systems. While sociodemographic characteristics are regularly surveyed, the conducted surveys use aggregate units that do not align with the actual system. Consequently, many water demand forecasts have been conducted at the administrative district level without adequately considering the water supply system. This study presents an upward water demand forecasting model that accurately reflects real water facilities and consumers. The model comprises three key steps. Firstly, Statistics Korea’s SGIS (Statistical Geological Information System) data was reorganized at the DMA level. Secondly, DMAs were classified using the SOM (Self-Organizing Map) algorithm to consider differences in water facilities and consumer characteristics. Lastly, water demand forecasting employed the PCR (Principal Component Regression) method to address multicollinearity and overfitting issues. The performance evaluation of this model was conducted for DMAs classified as rural areas due to the insufficient number of DMAs. The estimation results indicate that the correlation coefficients exceeded 0.9, and the MAPE remained within approximately 10% for the test dataset. This method is expected to be useful for reorganization plans, such as the expansion and contraction of existing facilities.

In this study, we analyzed how the revenue water ratio(RWR) is affected by changes in conditions of the water supply area, such as the ratio of aging pipes, maintenance conditions, and revenue water. As a result of analyzing the impact of pipe aging and maintenance conditions on the RWR, it was confirmed that the RWR could be decreased if the pipe replacement project to improve the aging pipe ratio was not carried out and proper maintenance costs were not secured. It was also confirmed that an increase in the revenue water could be operated to facilitate the achievement of the project’s target RWR. In contrast, a decrease in the revenue water due to a population reduction could affect the failure of the target RWR. In addition to analyzing the causes of variation in the RWR, the calculation of estimated project costs was considered by using leakage reduction instead of RWR from recent RWR improvement project cost data. From this analysis, it was reviewed whether the project costs planned to achieve the target RWR of the RWR improvement project in A city were appropriate. In conclusion, the RWR could be affected by variations in the ratio of aging pipes, maintenance conditions, and revenue water, and it was reasonable to consider not only the construction input but also the input related to RWR improvement, such as leakage reduction, when calculating the project cost.

The water consumption data of individual consumers must be analyzed and forecast to establish an effective water demand management plan. A k-mean cluster model that can monitor water use characteristics based on hourly water consumption data measured using automated meter reading devices and demographic factors is developed in this study. In addition, the quantification model that can estimate the daily water consumption is developed. K-mean cluster analysis based on the four clusters shows that the average silhouette coefficient is 0.63, also the silhouette coefficients of each cluster exceed 0.60, thereby verifying the high reliability of the cluster analysis. Furthermore, the clusters are clearly classified based on water usage and water usage patterns. The correlation coefficients of four quantification models for estimating water consumption exceed 0.74, confirming that the models can accurately simulate the investigated demographic data. The statistical significance of the models is considered reasonable, hence, they are applicable to the actual field. Because the use of automated smart water meters has become increasingly popular in recent year, water consumption has been metered remotely in many areas. The proposed methodology and the results obtained in this study are expected to facilitate improvements in the usability of smart water meters in the future.