摘要
针对上海市中心城区A水厂构建了基于时供水量预测模型的智能水量调度系统并对实践经验进行总结。该系统采用XGBoost算法构建时供水量预测模型,预测未来1~4 h内的瞬时供水流量,并结合出水机泵搭配模型与进水量计算模型,实现对出厂水流量、出水机泵及进水量的动态调整。研究结果表明,时供水量预测模型的预测准确性良好(平均相对误差为7.4%),智能水量调度方案执行率达到80%以上,相较于传统的人工经验调度模式,有效降低了人为因素导致的操作不一致性、缩短了应急响应时间,同时减少了出水泵房电耗成本。此外,精准的水量预测不仅减少了不必要的机泵启停次数,还优化了清水泵房的经济运行模式。该研究可为自来水厂智能化转型提供可行的技术路径,并验证了智能水量调度在提升供水服务质量、节能降耗方面的效益。
This study developed an intelligent water scheduling system based on hourly water supply forecasting model for waterworks A in central Shanghai and summarized practical implementation experiences.The system uses the XGBoost algorithm to build an hourly water supply forecasting model that predicts instantaneous water flow rates 1-4 hours ahead.Combined with effluent pump configuration and influent flow calculation models,the system dynamically adjusts plant effluent flow,pump operations,and influent volume.Results show that the hourly water supply forecasting model achieves good prediction accuracy with a mean relative error of 7.4%.The intelligent scheduling system execution rate reaches over 80%.Compared to traditional manual scheduling,the system effectively reduces operational inconsistencies caused by human factors,shortens emergency response time,and decreases pump station electricity costs.Additionally,accurate water demand forecasting reduces unnecessary pump start-stop cycles and optimizes the economic operation of the clear water pump station.This research provides a practical technical approach for intelligent transformation of waterworks and demonstrates the benefits of intelligent water supply scheduling in improving water service quality and reducing energy consumption.
作者
杨澜
李柱
杨瑜玲
王子瑜
徐力克
崔露苑
宋朝阳
YANG Lan;LI Zhu;YANG Yu‑ling;WANG Zi‑yu;XU Li‑ke;CUI Lu‑yuan;SONG Chao‑yang(Water Production Branch,Shanghai Chengtou Water Group Co.Ltd.,Shanghai 200086,China)
出处
《中国给水排水》
北大核心
2025年第18期88-95,共8页
China Water & Wastewater
关键词
时供水量预测模型
梯度提升决策树
智慧水厂
智能水量调度
hourly water supply forecasting model
Gradient Boosting Decision Tree(GBDT)
intelligent waterworks
intelligent water scheduling
