Cyanobacteria outbreaks are costly to manage and can pose significant public health risks, but one process engineer has been utilising machine learning to assist in the prediction and detection of nuisance algae in drinking water reservoirs and treatment plants.
Presenting at the Australian Water Association NSW Conference, Jacobs Process Engineer Brendan Gladman said the advances in sensor technology and use of machine learning allows for the generation of more insights from data.
“In practice, the water industry’s response to cyanobacteria has been largely reactive. The traditional way of determining the level of risk involves sending samples away to a laboratory for testing,” he said.
“There is an operational need for utilities to ensure the supply of safe and aesthetically acceptable drinking water.
“We wanted to develop a model that forecasts the concentration of cyanobacteria using water quality data in order to improve decision making and preventative risk management, including use of alternate sources or to implement additional treatment.”
Gladman said the model must be trained and its outputs are specific to a single reservoir, but there is potential for the approach to be applied in other water bodies too.
The success of modelling is dependent on access to good quality data and Gladman said, with the increase in data from online sensors, it’s hoped predictive models will be able to be developed to accurately and reliably predict a range of water quality parameters, including cyanobacterial blooms.
“We hope to realise the full potential of using machine learning for forecasting water quality. The message is that there is a means of extracting insight from data which has been ignored or not used to its full potential,” he said.
Gladman said that machine learning could definitely become a fundamental tool to improve the management of drinking water quality.
Register for the Australian Water Association NSW Conference to hear more from Brendan Gladman on machine learning for outbreak prediction.