摘要
排水管网系统是重要的城市基础设施之一,对其管网水位的精准预测可以有效减少或消除道路积水、管道溢流等灾害。当前雨水管网水位与降雨量存在正相关、滞后、峰延等复杂关系,同时与前m时刻的管网水位也存在密切关系,导致现有方法在精度上难以满足实际应用要求。为此,提出基于长短期记忆网络(LSTM)的雨水管网水位预测模型。该模型充分考虑到时间序列的影响因素,提高了预测的准确性。以湖州市某一区域为试验区,选定管网的10个节点,使用模拟数据和实测数据对模型进行了验证。结果表明,所提方法的预测准确率超过89%,达到了实际应用要求。
The drainage pipe network system is one of the important urban infrastructures.Accurate prediction of the water level in the pipe network can effectively reduce or eliminate disasters such as road flooding and pipe overflow.Currently,the water level in the rainwater pipe network has complex relationships with rainfall,including positive correlation,lag effect,and peak delay.It also has a close relationship with the previous m moment′s water level,making existing methods difficult to meet the requirements of practical applications in terms of accuracy.To address this issue,a water level prediction model for the rainwater pipe network based on Long Short-Term Memory(LSTM)networks is proposed.This model takes into account the influential factors of time series and improves the prediction accuracy.A specific area in Huzhou City is selected as the experimental zone,and 10 nodes of the pipe network are chosen for model validation using simulated and measured data.The results show that the proposed method achieves a prediction accuracy exceeding 89%,meeting the requirements of practical applications.
作者
王佳佳
盛政
胡文军
WANG Jiajia;SHENG Zheng;HU Wenjun(School of Information Engineering,Huzhou University,Huzhou 313000,Zhejiang,China;Zhejiang Province Key Laboratory of Smart Management&Application of Modern Agricultural Resources,Huzhou 313000,Zhejiang,China)
出处
《智能计算机与应用》
2025年第5期156-162,共7页
Intelligent Computer and Applications
关键词
降雨量
雨水管网
水位预测
长短期记忆网络(LSTM)
rainfall
rainwater pipe network
water level prediction
Long Short-Term Memory(LSTM)networks
