In many potable reuse projects, we largely rely on traditional control logics along with the conservative set points to take a unit process or entire treatment train offline to prevent the public consumption of unsafe water. While it represents a safe solution, it does not necessarily reflect the best management strategy due to potential losses of revenue as well as operational and maintenance challenges due to intermittent operations resulting from advanced treatment trains going frequently offline. Artificial intelligence (AI) and machine learning (ML) have the ability to constantly and quickly adapt and process large amounts of data in real-time with great accuracy, it may be an ideal platform for supporting our decision for managing potable reuse projects and assets. Unfortunately, AI/ML has not been widely used in the water industry despite its huge potential. The objective of this paper is to demonstrate if AI/ML can be successfully applied in potable reuse projects. To do that, AI/ML concepts have been applied to operational data obtained from two different potable reuse projects (one pilot, one full-scale). Feedforward artificial neural networks (ANNs) using Pytorch were developed to predict:1. RO permeate TOC in a pilot scale IPR train 2. Energy use of the RO facility and third stage RO fouling in a full-scale IPR facility Despite substantial changes in feed water quality and other operational parameters, the ANN model developed well predicted RO permeate TOC with R2 value of 0.945. Similarly, despite changes in RO feed water quality, operational parameters, and variations in membrane ages in the RO trains, the ANN models well predicted energy use (R2 of 0.971 to 0.989) and third stage fouling as reflected in the specific flux (R2 of 0.991) for plant optimization. This study clearly showed that AI/ML is a powerful tool to support our decision making in potable reuse projects while reducing errors and boosting regulatory and public confidence.
This presentation is available to AMTA Members only.
Speaker
- Ufuk Erdal
Company
- Arcadis
Event
- AMTA/AWWA Membrane Technology Conference, Las Vegas
Session
- AMTA/AWWA Membrane Technology Conference
Date
- 02/22/22
Media
Keywords
- Artificial Intelligence, Direct and Indirect Potable Reuse, Data Analytics
Reference
- 9715-DP2765