|

Modelling wastewater treatment plants: Kaggle competition

What is the best digital approach for modelling wastewater treatment plants? A data science competition tackles this challenge.

 

Modelling wastewater treatment plants

In wastewater treatment plants, there are many biological, chemical, and physical processes going on. Creating a digital model of these processes allows for making predictions about future behaviours and states depending on changing circumstances. For example, the amount of oxygen needed during the water treatment depends, among other things, on the outside temperature.

Such digital models of wastewater treatment plant come in three main types: mechanistic, data-driven, and hybrid. Mechanistic models emphasize understanding the causality between inputs and outputs, often referred to as physical models. Data-driven models, also known as AI models, leverage patterns within extensive datasets to make predictions. The hybrid model aims to synergise the strengths of both mechanistic and data-driven approaches, creating an optimised and comprehensive model for a wastewater treatment plant.

The Kaggle competition

In Kaggle competitions, data scientists, machine learning experts, and coding enthusiasts come together to tackle problems in data science. From predicting stock prices to understanding complex datasets, Kaggle competitions provide a dynamic space for participants to showcase their skills, learn from others, and contribute to cutting-edge solutions.

The goal of the Kaggle competition related to DARROW is to learn which of the three model types is best suited for modelling wastewater treatment plants. This competition is an initiative of the international water association’s (IWA) working group on hybrid modelling in water and wastewater treatment. Our partners from Ghent University are heavily involved in this working group. Hence, the data for the Kaggle challenge comes directly from the DARROW demo site in Tilburg. The best result will be awarded 1000€. The competition will close on March 1, 2024.