Identification of conceptual models for sewer flow

Abstract

Climate change will cause more extreme events to happen in the future, along with which is the higher frequency of rain. Besides, the growth of population and urbanization also results in more burden to the urban drainage systems. In order to understand the response of a drainage system, physically-based distributed models (PDM) such as Mike Urban+ are often used. However, these models require intensive computational power, which makes them not feasible for scenario analyses or real-time control. Conceptual models are preferred since they can produce fast simulation results. Nevertheless, conceptual models’ parameters are concepts that cannot be directly derived from reality, but through calibration or establishment of empirical links with the physical systems.

This study proposes a method to derive parameters of the conceptual models based on the drainage systems’ own physical characteristics using the storage-outflow (SQ) relationship. Two conceptual models are built in Python script. The first one optimizes and computes the SQ relationship for transport stretches of various lengths and slopes. The second model uses the obtained SQ relationship to simulate two virtual cities’ discharges, each city is divided into 4 different levels of lumping.

The conceptual model for transport stretch performs accurately with low errors. However, discharge hysteresis in long and flat pipes cannot be captured precisely. In comparison to using Mike Urban+, the conceptual models for simulating city discharges are up to 70 times faster with very high accuracy. The NSEs of different lumping levels are around 0.94 while the peak errors both in magnitude and time are small. It is recommended that the cities should be divided into compartments with the areas of around 7−15 ha for high quality results and low computational time. The quality of conceptual models does not depend on the shape of the city. Results show that the conceptual models in this study are capable of simulating outflow independently of any other data than the physical characteristics of the pipes.

Further work still needs to be conducted to evaluate the stability of the conceptual models and increase their accuracy. However, this study’s conceptual models show great potential for fast and accurate discharge simulation in an urban drainage system without the need of using any other PDM.