The Metropolitan Water Reclamation District of Greater Chicago (MWRDGC) operates one of the largest stormwater management systems in the world, including tunnels and massive reservoirs.
THE CHALLENGE
Unfortunately, these reservoirs can be a source of odor, which is not just a minor nuisance but negatively impacts public health and wellbeing. With an extremely complex reservoir system, climate dynamics, and human behavior dimensions, physical modelling was not possible. Our goal was to develop a machine learning algorithm for advanced prediction of odors coming from the reservoir. The solution development process consisted of data processing and quality assurance/quality control, algorithm development, and algorithm testing and evaluation.
THE SOLUTION
Supervised machine learning techniques were used to both predict odor events and identify human-interpretable relationships important to the client’s odor monitoring and reservoir management operations. By creating many algorithms and choosing the best ones, with accuracy measured using receiver operating characteristic curves, we created a highly accurate advanced warning algorithm.
RESULTS
Ensaras used machine learning to develop an accurate algorithm that aided in odor event prediction, and further provided MWRDGC with other key human-interpretable insights to optimise reservoir odor management and benefit public health and wellbeing. This work received the LIFT Intelligent Water Challenge Prize (2nd place) at WEFTEC.
At a Glance
Challenges:
- Odor problems
- Unhappy customers
- A physical model of odor dispersion was not
available
Benefits:
- Accurate odor prediction based on machine learning as part of a cost-effective odor management solution
