AI for climate adaptation - methods for creating a more reSilienT dRinking wATer prodUction and Supply (STRATUS)
| Reference number | |
| Coordinator | IVL Svenska Miljöinstitutet AB |
| Funding from Vinnova | SEK 3 189 360 |
| Project duration | November 2021 - March 2024 |
| Status | Completed |
| Venture | Climate adaptation in the built environment |
| Call | Climate adaptation in the built environment |
Important results from the project
** Denna text är maskinöversatt ** The project´s objective was to investigate how AI techniques can improve drinking water production in view of climate change. The project succeeded in predicting the turbidity of the raw water source, although the central sensor delivered incorrect values for a period. The dose model could not be implemented due to insufficient data from the sensor. The prediction of the quality impact in the distribution network failed with existing data, quantitative data will be required to a different extent than what is available now.
Expected long term effects
** Denna text är maskinöversatt ** The project succeeded in predicting the turbidity of the raw water well, even though we used publicly available data sources, which shows that it can be a feasible way for actors to approach AI modeling, which often falls short of the quality of the data. När Sandviken Energi installs a meteorological station, improved training data is expected to increase the accuracy of the model. In the long run, this can lead to more efficient operation of waterworks and improved water quality in the distribution network, which benefits both Sandviken and other cities.
Approach and implementation
The layout of the project assumed that the data was of good quality and that sensors were in place. During the course of the project, several changes in direction occurred due to technology or a change in focus from the problem owner. These meant that the project, on a meta level, came to be about how to make the best of the situation when the data is as it usually is - of low quality and not aggregated. The method we used, where we used public data, makes it possible for more people to create a proof of concept, and investigate possibilities before going further and collecting their own data.