Simulation of urban wastewater systems using artificial neural networks: embedding urban areas in integrated catchment modelling

Abstract

The urban wastewater system is an important part of integrated water management at the catchment level, yet, more often than not, inclusion of the system and its interaction with the surrounding catchment is either oversimplified or totally ignored in catchment modelling. Reasons of complexity and computational burden are mostly at the heart of this modelling gap. This paper proposes to use artificial neural networks (ANN) as a surrogate for the simulation of the urban wastewater system, allowing for a more realistic representation of the urban component to be incorporated into catchment models within a broad scale modelling framework. As a proof of concept, an integrated urban wastewater model is developed and its response in terms of both quantity and quality in combined sewer overflow (CSO) discharges and treatment plant effluent are captured and used to train a feedforward back-propagation ANN. The comparative results of the integrated urban water model and the ANN show good agreement for both water quantity and quality parameters. The resulting trained network is then embedded into a MIKE BASIN catchment model. It is suggested that ANN models greatly improve the level at which broad scale catchment models can accurately take into account urban-rural interactions.

Multiobjective optimisation on a budget: Exploring surrogate modelling for robust multi-reservoir rules generation under hydrological uncertainty

“…Typical surrogate models include polynomials (Crestaux et al, 2009;Sudret, 2008), Radial Basis Functions (RBFs) (Mugunthan et al, 2005;Regis and Shoemaker, 2007b;Shoemaker et al, 2007), Artificial Neural Networks (ANNs), Support Vector Machines (SVMs - Dibike et al, 2001;Zhang et al, 2009a) and Kriging (Sacks et al, 1989;Santner et al, 2003). ANN, in particular, have been used extensively in water resources research (Behzadian et al, 2009;Broad et al, 2005Broad et al, , 2006Fu et al, 2012;Fu et al, 2010;May et al, 2008). Khu et al (2007) examined various applications of evolutionary computation based surrogate models to augment or replace the conventional use of numerical simulation and optimisation within the context of hydro-informatics.…”

A Hybrid Surrogate Modelling Strategy for Simplification of Detailed Urban Drainage Simulators

“…Application of data-driven approaches in various aspects of urban water management domain has been growing rapidly during the past decade (Eggimann et al 2017). Due to the advantages addressed above, data-driven surrogate modelling approaches are not exempt in this regard (Fu et al 2010;Gradano and Le Roux 2012;Nadiri et al 2018). However, in most of data-driven approaches the input-output mapping is performed in a black (or grey) box manner, neglecting most of the mechanisms inside the simulator and solely focusing on the input-output data.…”

Receiving water body quality assessment: an integrated mathematical approach applied to an Italian case study

“…• The complexity of a given model introduces uncertainties in the modelling process that, sometimes, are not clearly identifiable and assessable (Mannina & Viviani 2010). • Such approaches are usually computationally demanding, often requiring the definition of specific ad hoc models able to improve such aspects (Fu et al 2010). • System complexity take as a consequence difficulties in the definition of reliable parameters values; such a consideration took to the investigation of probabilistic models giving the advantages of not focusing on single deterministic values of the parameters and of providing probabilities associated to specific model responses (Benedetti et al 2010).…”