[1]何媛滨,方正,康丹,等.基于BP神经网络的合流制截污管溢流实时控制方法[J].中国给水排水,2024,40(3):120-123.
　HEYuan-bin,FANGZheng,KANGDan,et al.Real-time Control Method for Combined Sewer Overflow Based on BP Neural Network[J].China Water & Wastewater,2024,40(3):120-123.
点击复制

基于BP神经网络的合流制截污管溢流实时控制方法

中国给水排水[ISSN:1000-4062/CN:12-1073/TU] 卷: 第40卷 期数: 2024年第3期 页码: 120-123 栏目: 出版日期: 2024-02-01
Title:
Real-time Control Method for Combined Sewer Overflow Based on BP Neural Network
作者:
何媛滨1， 方正1， 康丹2， 刘涵1， 陈铁2
（1.武汉大学 土木建筑工程学院，湖北 武汉 430062；2.武汉市政工程设计研究院有限公司，湖北 武汉 430023）
Author(s):
HE　Yuan-bin1， FANG　Zheng1， KANG　Dan2， LIU　Han1， CHEN　Tie2
（1. School of Civil Engineering， Wuhan University， Wuhan 430062， China； 2. Wuhan Municipal Engineering Design & Research Institute Co. Ltd.， Wuhan 430023， China）
关键词:
合流制系统;  神经网络;  实时控制方法;  截污管溢流
Keywords:
combined sewer system;  neural network;  real-time control method;  interceptor sewer overflow
摘要:
提出了一种基于BP神经网络的合流制截污管溢流控制方法，通过使用当前时刻前一段时间每5 min的降雨、控制点水位、蒸发量和雨前干期等数据对控制点未来一段时间的水位进行预测，当控制点预测水位长时间超过设定值时，则实施开闸溢流。通过对近2年共3 000 组观测数据进行训练测试，发现该模型的平均绝对误差（MAE）为0.111 5、均方根误差（RMSE）为0.156 5、平均百分比误差（MPE）为0.045 3%、纳什-苏特克利夫效率系数（NSE）为0.938 6，表明该模型在整体预测上满足应用要求。现该模型已部署至武汉市智慧水务系统，并在2023年7月8日的降雨事件中，预测到水位将会上升到22.7 m以上，实施开闸溢流，避免了城区内洪灾害。
Abstract:
This paper proposed a method for controlling combined sewer overflow based on BP neural network. The water level at the control point for a future time span was predicted by using the data of rainfall, water level at the control point, evaporation and antecedent dry period during every 5 minutes before the current time, and the gate overflow was implemented when the predicted water level at the control point exceeded the set value for an extended period. A total of 3 000 sets of observation data in recent 2 years were trained and tested. The mean absolute error (MAE) of the model was 0.111 5, the root mean square error (RMSE) was 0.156 5, the mean percentage error (MPE) was 0.045 3%, and the Nash-Sutcliffe efficiency coefficient (NSE) was 0.938 6, indicating that the model met the overall prediction requirements in application. The model was deployed in Wuhan Smart Water System. In the rainfall event on July 8, 2023, the water level was predicted to rise to more than 22.7 m, and the gate overflow was implemented, thereby avoiding waterlogging in the city.

相似文献/References:

[1]姜楠,封莉,林子茵,等.基于动态均值感知器的拍门污水泄漏精准检测系统[J].中国给水排水,2024,40(17):118.
　JIANGNan,FENGLi,LINZi-yin,et al.Precise Detection System for Sewage Leakage in Pneumatic Flap Doors Based on Dynamic Mean Perceptron Neural Network[J].China Water & Wastewater,2024,40(3):118.

更新日期/Last Update: 2024-02-01