GIS-Neural Network Hybrid Approach for Capacity Analysis of Metro Vancouver Sewer System

This paper will discuss how a trained neural network-based flow prediction model can be embedded within a Geographic Information System (GIS) framework to efficiently estimate peak dry-weather flows (PDWF) at several hundred locations in Metro Vancouver sewer system. The seamless computation process only takes a very short time typically in minutes. The paper will also show tabulated peak dry-weather flow using conventional Unit-Flow methodology, and compare the accuracy of both methods against the field measured dry-weather flow in various independent case studies. Although the traditional approach of using a unit sewage flow rate per capita is simpler, Metro Vancouver has found that the GIS-Neural Network based program was more consistent, accurate and extremely time efficient in representing the peak dry-weather sanitary flows in service catchments of various areal and population sizes. This hybrid method allows Metro Vancouver to save cost in actual flow monitoring, capacity upgrading, routine service billing and operational flow bypassing situations.