[1]吴长峰,谢志诚,钟燕.基于注意力双向GRU模型的城市供水加氯量预测[J].中国给水排水,2026,42(17):61-66.
WuChangfeng,XieZhicheng,ZhongYan.Prediction of Chlorine Dosage in Urban Water Supply Based on an Attention-based Bidirectional GRU Model[J].China Water & Wastewater,2026,42(17):61-66.
点击复制
WuChangfeng,XieZhicheng,ZhongYan.Prediction of Chlorine Dosage in Urban Water Supply Based on an Attention-based Bidirectional GRU Model[J].China Water & Wastewater,2026,42(17):61-66.
基于注意力双向GRU模型的城市供水加氯量预测
中国给水排水[ISSN:1000-4062/CN:12-1073/TU]
卷:
第42卷
期数:
2026年第17期
页码:
61-66
栏目:
出版日期:
2026-09-01
- Title:
- Prediction of Chlorine Dosage in Urban Water Supply Based on an Attention-based Bidirectional GRU Model
- Keywords:
- urban water supply; intelligent chlorine dosing; attention mechanism; bidirectional gated recurrent unit (BiGRU); chlorine dosage prediction
- 摘要:
- 针对水厂水处理加氯消毒过程中存在的非线性、多干扰，以及加氯决策缺乏数据支持等问题，提出一种基于融合注意力机制与双向门控循环单元（BiGRU）的智能加氯预测模型。该模型以余氯、供水量、水温、进水量等多种关键因素作为输入，通过BiGRU捕捉加氯过程中的双向时序依赖关系，充分挖掘历史和未来信息，以提高预测的全面性。同时，结合自注意力机制，使模型聚焦于加氯过程中的关键时刻特征，区分不同时间节点上的重要信息，增强对关键时刻变化的敏感度。通过对兰州市供水系统实际运营数据进行实验验证发现，提出的模型在50 d长时间跨度下的预测表现优异（R?=0.928 1），显著提升了复杂供水场景中的加氯量预测精度。
- Abstract:
- To address the issues of nonlinearity, multiple interferences, and the lack of data support for chlorination decision-making in water treatment disinfection processes, this paper proposed an intelligent chlorine dosage prediction model based on an attention mechanism integrated with a bidirectional gated recurrent unit (BiGRU). The model took multiple key factors such as residual chlorine, water supply volume, water temperature, and inflow volume as inputs. By leveraging the BiGRU, the model captured bidirectional temporal dependencies in the chlorination process, fully exploiting both historical and future information to enhance prediction comprehensiveness. Meanwhile, the self-attention mechanism enabled the model to focus on critical temporal features during chlorination, distinguishing important information across different time nodes and improving sensitivity to key variations. Experimental validation using operational data from the Lanzhou water supply system demonstrated that the proposed model achieved excellent prediction performance over a long-term span of 50 days (R2=0.928 1), significantly improving the accuracy of chlorine dosage predictions in complex water supply scenarios.
相似文献/References:
[1]李智力,曾新.武汉市汉口地区供水监测系统的构建及维护运行[J].中国给水排水,2018,34(20):31.
LI Zhi li,ZENG Xin.Construction and Maintenance of Water Supply Monitoring System in Hankou District of Wuhan[J].China Water & Wastewater,2018,34(17):31.
[2]魏锦程,李琳.城市供水管网漏损治理系统化方案编制方法[J].中国给水排水,2022,38(12):1.
WEIJin-cheng,LILin.Systematic Scheme Compilation Method for Water Loss Control of Urban Water Supply Network[J].China Water & Wastewater,2022,38(17):1.
更新日期/Last Update:
2026-09-01