Monitoring water quality with sensors placed at strategic points in the distribution network provides an online picture of water quality during its distribution and in the vicinity of the tap. Currently, a picture of water quality is only available a few days after its delivery and/or at only a few locations. With a clear vision of what is desired from the network, and with a well-founded strategy to determine the number of sensors needed and their optimal placement, one can achieve an optimally performing sensor network. Both within and outside the water company, this ensures realistic expectations with regard to the added value and the limitations of water quality sensors.
This projects aims to provide a suitable, optimal placement method for wide application by the Dutch water companies. Concretely, the objective is the development of a method and research tool for the optimal placement of a small number of fallible sensors in a robust configuration in a distribution network, so that the representativeness of measurement values, the monitoring of operational management and incident detection are optimised.
Report on objectives, numerical evaluation and optimisation
The problems involved were approached from a modelling perspective. To begin with, a study was done of how one can describe the performance of a sensor network, and what objectives one should bear in mind when developing a sensor network. Then, on the basis of a combination of hydraulic models, contamination scenarios and an optimisation method (genetic algorithm), a program was written for the evaluation and optimisation of sensor network designs. This program was first tested in the Leeuwarden-Bergum subarea of the Vitens Innovation Playground, and subsequently implemented there as well. In the process, optimal configurations for various objectives were determined, as were inter-relations and trends. The results were incorporated in a report, including recommendations for the installation of a water quality sensor network.
Better performance through numerical optimisation
The results of the model calculations are unequivocal: numerical optimisation leads to a clearly better performing configuration of a specified number of water quality sensors, than is the case when the same number of sensors is evenly distributed, or when the placement is based on feeling and experience. In this research, the average chance of detection, for example, is almost doubled, and the average time before first detection is cut by several hours. The optimised networks generally even function better than the existing uniformly-distributed configurations in meeting objectives for which they weren’t optimised. This very clearly demonstrates the usefulness of such a numerical approach.
The methods used and developed for the optimal placement choice of water quality sensors make it possible to weigh different objectives against each other, and to define realistic expectations about the benefits of such a network. This contributes to both internal decision-making and to external communications regarding decisions and actions in this area. Costs can be weighed against the degree of protection given; and the outcome of these calculations provide the framework conditions for a response the company can strive for.