Cluster analysis integrated in network rehabilitation model
Country: The Netherlands
Period: 2021 & 2022
Demand prediction plays an important role in fully automated drinking water treatment plants. In fully automated drinking water treatment plants, the operator has developed into a process and process automation supervisor, whereas the demand prediction algorithm is the input for the calculation of the production setpoints. For more than two decades, two software packages have been applied at Dutch drinking water utilities. Drinking water company Oasen had reason to engage with Spatial Insight and request the developments to develop a new algorithm and software package for the seven supply areas it serves. Demand production allows ‘flat’ production flows, being optimal in terms of water quality, energy usage and leveraging treatment capacity on days with high demands (thus allowing postponement of investments). At the same time, looking ahead allows optimal planning of well extractions, wich is important for permits and water quality.
Approach and solution
The data science team of Spatial Insight applied the machine learning principle to historic demand data of Oasen of the supply area Lekkerkerk. In a desk study the calculated predictions and setpoint calculations were compared with the existing software. These results were promising: flatter production, more accurate prediction without any manual input (like the definition of bank holidays). The algorithm was packaged into software that could be installed at pumping station Schuwacht. During 3 months, SI-Demand ran in parallel with the existing software on a dedicated virtual machine of Oasen. This gave our client confidence that SI-Demand could be installed on a local industrial PC on the treatment works of pumping station Schuwacht before Christmas. Even with the abnormal situation over Christmas, SI-Demand performed as desired.
Contribution to the organisation’s strategy
SI-Demand allows Oasen to calculate optimal and ‘flat’ production setpoints with robust standard software. Optimal setpoint will help to optimize water quality, save energy, and leverage production capacity and permits optimally. SI-Demand opens the door to next steps in advanced water distribution with anomaly detection, pump energy optimization and leakage localisation.
Bas Bouwman, Sector Manager Klant & Bedrijfsvoering of Oasen has invited Spatial Insight to propose the implementation of SI-Demand in the other supply areas of Oasen.
Ignaz Worm (ignaz.worm [@] spatial-insight.nl)