JinXian,ZhangLei,MaJie,et al.Study on Synergistic Dosing of Dual Chemicals in Water Treatment Plants Based on MIMO Model[J].China Water & Wastewater,2026,42(15):61-66.
Study on Synergistic Dosing of Dual Chemicals in Water Treatment Plants Based on MIMO Model
China Water & Wastewater[ISSN:1000-4062/CN:12-1073/TU]
volume:
第42卷
Number:
第15期
Page:
61-66
Column:
Date of publication:
2026-08-01
- Keywords:
- water treatment plant; machine learning; coagulant dosage; ferric chloride (FeCl3); polyaluminum chloride (PAC)
- Abstract:
- A multi-input multi-output (MIMO)-based dosage prediction model was established to tackle the challenge of synchronous and accurate dosage determination in dual-chemical coordinated dosing for water treatment plants. A large-scale water treatment plant in northern China was taken as the research object, and 65 880 sets of operational and water quality data were collected over the entire year of 2024. Three algorithms—XGBoost, SVR, and RF-Elman—were adopted, and a multi-head attention mechanism along with a penalty term was introduced to strengthen the model’s responsiveness to key indicators such as pH and turbidity. Meanwhile, mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2) were employed as evaluation metrics. The results showed that XGBoost achieved the best predictive performance for both ferric chloride and polyaluminum chloride dosages, with test-set R2 values reaching 0.992 4 and 0.982 8, respectively, and RMSE and MAE both lower than those of the other models. Feature importance analysis indicated that influent temperature, sodium hypochlorite and ozone pre-oxidation parameters, and the constructed pH regulation equation were key influencing factors. Robustness validation demonstrated that XGBoost offered a lower risk of water quality exceedance and higher engineering safety under extreme water quality conditions, achieving the best overall performance.
Last Update:
2026-08-01