Clean water is one of the cornerstones of our living, and therefore we need to always thrive for the best and highest quality water. Data-driven water management is now meeting artificial intelligence to fundamentally change the way of ensuring clean water for all. Together with Ramboll, we set out to understand how machine learning environments can bring significant improvements to water utilities in Finland.
EDIT: Read Tekniikka&Talous article about our work at Ramboll.
Starting point: smart water management
Water utilities these days are equipped with plenty of IoT sensors and other data-driven technologies that constantly collect data on different phases in the water supply and demand. These sensitive instruments are, when used correctly, able to draw previously-unattainable information on the treatment of water and empower water utilities to predict their water operations.
Setting the smart monitoring system up and maintaining it can cost up to hundreds of thousands of euros. However, often this data is not analysed until some trouble is already up. Typically, when a problem arrives, data collected by these sensors and water sample laboratory results are sent to an external analyst, who then tries to research the correlation between different parameters and understand what went wrong.
This process of analysing raw data is quite difficult for humans to analyse and it requires a lot of time. What if the analysis process could be done constantly, drawing predictions on the performance of the water facility on a daily basis, even before any disturbance in the system?
Artificial intelligence to reshape the managing of water
With Ramboll, we were able to prove that artificial intelligence, and machine learning in particular, can be leveraged to create more effective water treatment processes, make sure problems can be recognised ahead of time and help direct efforts in those areas early enough. Machine learning, which is often used for predicting changes, can also create new insights that can be used as information for future investments and planning the usage of the water utilities.
For these purposes, Silo.AI developed machine learning algorithms that uses time-series forecasting for Ramboll. The AI solution predicts the quality of the water leaving from the water utilities. The quality of the water is analysed in the context of environmental permissions and terms.
With the machine learning model we were able to show its capabilities to monitor the water treatment process constantly, shifting the focus from trouble-shooting to predictive risk assessment and dynamic optimisation of the facilities. With machine learning, it is possible to use the already invested smart systems and their measurements more efficiently, and identify risks in advance. Our AI system also contributed to finding out new factors that affect the performance of the water utilities.
The data analysed consisted of measurement data from the laboratory and reporting data produced by the water utility’s IoT sensor data. Also other sources of public data such as weather data and network information could be used to form more accurate predictions.
AI era for the water sector requires starting today
The results of the pilot were encouraging. We noticed that the machine learning environment can be used to create more efficient and accurate support for Ramboll’s clients. The algorithms can be leveraged to build cost-effective AI-driven systems on top of the existing IoT infrastructure, that improve the day-to-day operations at water utilities, preventing any surprises and optimising the current usage of the facilities.
After the pilot both Silo.AI and Ramboll look forward to creating Human-in-the-loop AI systems for water treatment, which would take the human-machine collaboration to a new level. With Human-in-the-loop AI, the machine can be given the grunt work and basic analysis, where people get to work on higher-value performing tasks and validate the analysis produced by the AI.
However, the biologic treatment of waste water is a complex environment. As one result of the project, it seems clear that the first step in any water industry AI project is to establish rules on what data needs to be collected and on what level. Quality of data is after all one of the most important factors in creating a working AI system.
As a result of the pilot we now have better knowledge on how to help Ramboll offer AI-driven solutions to their clients. The future projects will focus on optimising the energy efficiency and functioning of the water utilities, but also improve the human-machine collaboration. Together with Ramboll, we can leverage our expertise in both artificial intelligence and water management to manage water resources with an unprecedented efficiency.
