Municipal wastewater treatment operators and managers cannot control all contamination introduced to a plant, which leads to significant issues with operations. Now, one group of researchers has developed a new data analytics approach to support preventative measures.
Researchers from the University of South Australia have developed a real-time process early warning system that uses online measurements and smart data analytics to support wastewater treatment plant operations.
Lead researcher Dr Christopher Chow said the project was about demonstrating the potential usefulness of an anomaly detection system using online scanning instruments to give indications of changing raw water quality and to allow for real-time treatment response.
“For wastewater treatment plant operations, a significant challenge is posed when unwanted substances make their way into a biological treatment process. It can slow down or stop biological activity,” Chow said.
“This can lead to poor quality effluent for a period of time because operators may not be aware when an unwanted material has entered the system until after a problem has been realised.
“As researchers, we took the opportunity to further develop smart data analytics tools that can provide additional capability, to demonstrate that online data can be used to provide additional information to operators.”
Chow said the project’s objective was to introduce more sophisticated monitoring at the beginning of the treatment system so that operators know when potentially damaging material is present in the raw water and prepare accordingly.
“It works as an extra alarm or early warning so that operators have the insights they need to respond sooner,” Chow said.
“If they know what’s coming into their system, they can take corrective measures, rather than waiting until the plant is not functioning and addressing the issue reactively.”
Chow said advances in instruments have enabled online monitoring systems for faster and more reliable detection of contaminants, as well as treatment optimisation, for wastewater treatment plant operators.
“Most modern treatment plants have already got some kind of monitoring system, such as SCADA. There are usually already some basic alarm functions set up, some parameters that indicate certain anomalies,” he said.
“Our work is about creating an advanced version of this monitoring approach using the latest instrumentation and advanced computing.
“There has to date been a limitation for processing continuous data. Our aim was to develop additional functions so that the data can be used more effectively by adding an extra level of visibility.”
Using data analytics, the smart data processing system incorporates more complicated data streams from UV-Vis spectrophotometer sensors.
“For example, when measuring one parameter, like pH, there is only one point of data per minute. Our approach used a more complicated analytical instrument, which looks at the spectrum of that data point,” Chow said.
“It can tell us a lot more information about the water quality. Every measurement point has a spectrum, so rather than getting one point of data per measurement, this system enables 256 points per measurement.”
Chow said the increased volume of data needs to be processed in a way that provides useful insights to plant operators.
“These data scans need to be sent back into a central computer for processing. This is why we needed to develop new processing systems to allow the operator to interpret the results,” Chow said.
“The insights also need to be simple enough for the operator to get the correct warning. We don’t want operators to have to read complex interpretations of this more complicated data. We want to provide insights in a simple way to make their job easier.”
The benefits of the smart data processing system developed by the research team include clear summaries of data for quick use, data clustering for more effective anomaly detection, classifying data for predictions, and visually exploring the data for general understanding.
“We have sophisticated software to do the processing but we found that operators want specific insights presented in a way that’s easy to interpret. We used an internet browser to display the information, utilising lots of visual representations of anomalies,” Chow said.
“Fixed alarms can be set, but sometimes they are affected by other operational parameters, which can produce a false alarm. But wIth data analytics, we can combine more types of operational data to come up with a much more reliable alarm system.
“This is about introducing additional capability into the system to allow for more accurate warnings.”
Chow said the smart data processing system can also collect one year’s worth of data, enabling pattern recognition that can be used to better understand of the characteristics of the system and potentially create predicative capability to help manage the plant.
“This system can also analyse years of data to allow it to identify potential recurring problems. It can identify patterns within the raw water stream, enabling detailed views of the types of materials that are introduced at certain times,” he said.