Wastewater treatment modelling

1. Wastewater modelling
Modelling wastewater treatment plants provides a means of:
- representing the internal behavior of a process
- designing a new process
- testing a process, without expenditure on mechanical equipment
- mimicking an existing facility to evaluate key parameters and diagnose/rectify the process
- providing a means of controlling an existing or proposed facility, and
- predicting the future performance of an existing facility.
Model testing of an existing facility is normally relatively inexpensive, faster and simpler than practical testing. It can be conducted without the risk of damage to the facility. A very significant body of research is based on modelling of wastewater processes, ranging from mimicking existing processes (including so-called ‘digital twins’) to aiding the development of new processes.
Of greatest practical significance is the use of modelling for computer-aided design (CAD). A number of commercial software packages, or platforms, exist which are used to provide designs of wastewater treatment plants. These are predominantly based on mechanistic models which use a broad set of mathematical expressions to represent the biochemical, chemical and physical processes which define the operation of wastewater treatment schemes.
2. Mechanistic modelling: commercial CAD software
Commercial software packages include (in alphabetical order):
- BioWin (EnviroSim): Established and widely used 'black box' software with a user-friendly interface used for general wastewater treatment plant simulations and process design.
- Generative Design Generator (Transcend): providing complete engineering packages from a limited set of design inputs for CAS, SBR, MBBR, AGS (Nereda), Wetlands, and MBR technologies.
- GPS-X (Hydromantis): Modelling/simulation tool interfacing with MATLAB for controller design, and including a Monte Carlo analysis tool for generating probabilistic distributions of model outputs.
- SIMBA# (Ifak): Simulation system for water and wastewater systems, including activated sludge, aeration systems, and sewer networks.
- STOAT (WRc): Freeware tool for simulating WwTP and sludge processing performance, used to optimise the response of the works to changes in the influent loads, works capability, or process operating conditions.
- SUMO (Dynamita): Open-source platform offering flexibility for customisation through modification of the model algorithms by the user.
- WEST, Wastewater Treatment Plant Engine for Simulation and Training (DHI): Platform which includes an editor for creating or modifying mathematical models and a configuration tool for building complex models.
There are also some CAD platforms, such as ArcGIS and QGIS, which are Geographic Information System (GIS) software packages providing spatial analysis and mapping of water and wastewater systems to assist with the plant layout. GIS is also included in other platforms, including Transcend’s Generative Design Generator.
Whilst the most comprehensive software packages incur a fee – of anything up to ~$4,500 (as of January 2025) depending on the features required – some are currently (as of January 2025) available free of charge. This applies to the teaching and research version of ASIM (Activated Sludge Simulation, developed by EAWAG), the simplified version of GPS-X (called ‘GPS-X Lite’), and STOAT.
All the commercial CAD software platforms predicting, or otherwise representing, the performance of biological treatment processes are based on a series of mathematical models originally developed in the 1980s and 1990s. These models, abbreviated as 'ASM' ('Activated Sludge Models'), have been modified over the years to account for different biochemical and/or physicochemical components of the wastewater treatment process.

3. Origins of mechanistic modelling: ASMs
Activated Sludge Models (ASMs) were originally introduced by the International Water Agency (IWA) in 1987, the first basic model being ASM1 (Henze et al, 1987). ASM1 can calculate COD and N removal (both nitrification and denitrification), oxygen consumption and sludge production. It represents a key development in wastewater process modelling and lays down the foundation for subsequent models.
ASM2 (Henze at al, 1995) extends ASM1 to include biological and chemical phosphorus removal though the action of phosphate accumulating organisms (PAOs), growing only under aerobic conditions, with the associated anaerobic, anoxic, and aerobic reactions. ASM2d (Henze at al, 1999) extends ASM2 to include the denitrification activity of the PAOs.
ASM3 refines the original ASM1 model to more accurately describe the kinetics of substrate uptake and oxygen consumption. ASM3 is more sophisticated than ASM1 but relies on data or appropriate assumptions regarding the proportion of readily and slowly biodegradable COD (RBCOD and SBCOD respectively).
Mechanistic models based on ASM – both commercial and non-commercial – have been pivotal in the development of wastewater treatment processes. A more recent development is the application of artificial intelligence to the area.
4. Artificial intelligence (AI): machine learning
Machine learning (ML) is the use and development of computer systems able to learn how to perform tasks without following explicit instructions, through using algorithms and statistical models to analyse and draw inferences from patterns in data. ML approaches in wastewater treatment are currently mostly used for decision support, using inductive inference to generalise the correlations between input and output data which is then used to make informed decisions in new circumstances. The correlations are thus informed by data, rather than a set of predetermined mechanisms with associated defined mathematical relationships.
Among the various ML models, the artificial neural network (ANN) algorithm predominates in wastewater treatment. This model comprises interconnected nodes that work much like neurons in the human brain. Using algorithms, they can recognise hidden patterns and correlations in raw data, cluster and classify it, and use this information to gradually improve the process.
Research into the application of ML to wastewater treatment modelling increased dramatically from around 2015 onwards, with a significant effort in the prediction of membrane fouling in MBRs (Yang & Li, 2025).