Premium
Fuzzy Neural Network for Flow Estimation in Sewer Systems During Wet Weather
Author(s) -
Shen Jun,
Shen Wei,
Chang Jian,
Gong Ning
Publication year - 2006
Publication title -
water environment research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.356
H-Index - 73
eISSN - 1554-7531
pISSN - 1061-4303
DOI - 10.2175/106143005x89553
Subject(s) - artificial neural network , stormwater , environmental science , surface runoff , storm , combined sewer , retention basin , fuzzy logic , hydrology (agriculture) , meteorology , computer science , engineering , machine learning , artificial intelligence , geotechnical engineering , geography , ecology , biology
Estimation of the water flow from rainfall intensity during storm events is important in hydrology, sewer system control, and environmental protection. The runoff‐producing behavior of a sewer system changes from one storm event to another because rainfall loss depends not only on rainfall intensities, but also on the state of the soil and vegetation, the general condition of the climate, and so on. As such, it would be difficult to obtain a precise flowrate estimation without sufficient a priori knowledge of these factors. To establish a model for flow estimation, one can also use statistical methods, such as the neural network STORMNET, software developed at Lyonnaise des Eaux, France, analyzing the relation between rainfall intensity and flowrate data of the known storm events registered in the past for a given sewer system. In this study, the authors propose a fuzzy neural network to estimate the flowrate from rainfall intensity. The fuzzy neural network combines four STORMNETs and fuzzy deduction to better estimate the flowrates. This study's system for flow estimation can be calibrated automatically by using known storm events; no data regarding the physical characteristics of the drainage basins are required. Compared with the neural network STORMNET, this method reduces the mean square error of the flow estimates by approximately 20%. Experimental results are reported herein.