Meeting the demand for resources (water, energy, food, etc.) is a big challenge, in which continuous optimisation can take place. Historically the demand for many of these resources was met through centralized, large scale infrastructures systems. The transition towards more sustainable resource use could involve the transition towards de-centralized, small scale, systems which facilitate the re-use of waste streams which are also produced de-centrally. The main challenge to achieve this is to reduce costs to such an extent that these systems can compete with traditional centralised systems.
DOWDupont has a large chemical industrial facility located in Zeeuws-Vlaanderen, in the south of the Netherlands. Currently the main water supply for this facility are the water basins of De Biesbosch located 120 km to the North-East. The company is scouting possible alternatives to become independent of this remote water supply so as to reduce transport costs a to mitigate insecurity of supply in the future. Climate change and socio-economic development could reduce water availability from De Biesbosch, which means that ensuring future operations of the plant requires alternative water sources.
The transport of water poses a unique problem because pipeline infrastructure needs to be installed. Determining the best location of such a pipeline can have large effects on the costs, which in turn determines the feasibility of a project. Knowledge is lacking on how to effectively model the effect of landscape characteristics (waterways for example) on the optimal routing of a pipeline (see Figure). Historically, this problem has been solved by involving experts which would evaluate an area “by hand” to determine the optimal route. This approach is effective when a centralised system involving one supply and one demand location is involved. Moving to a decentralized systems, with n supply locations and m demand locations would greatly benefit from an algorithm which can do a preliminary scan of the optimal pipeline route.
The problem at hand can be compared to the Euclidean Steiner Tree Problem in the sense that the shortest route between terminals needs to be found. However, in this case we are dealing with landscape elements and obstacles with varying spatial costs as well as varying costs for size of pipelines. A model should provide the project multiple best alternatives for a decentralized water network.
- Develop (methodology) an algorithm that solves the above described optimization problem.
- Write a function in a scripting-language to implement the algorithm
- Apply the algorithm to the DOW case
- https://www.youtube.com/watch?v=bQPAqNIFVlI (The black dots would represent the demand and supply locations, the intersections of the blue lines represent the “hubs”. The challenge is to convert this problem to one which includes landscape characteristics and obstacles such as waterways as shown in the figure.)
- Two Heuristics for the Euclidean Steiner Tree Problem, Derek R. Dreyer∗ Michael L. Overton, September 30, 2002, https://people.mpi-sws.org/~dreyer/papers/steiner/steiner.pdf
- This project needs to be done in either Python or R. Visualizations of results could be done in ArcGIS or QGIS, but the development and calculation of the route optimization needs to be in a scripting language. The ideal candidate has already experience with Python or R.
- The ideal candidate is creative and willing to develop a routing algorithm for this case as a new method/concept and is also able to implement this in a script.
Theme(s): Modelling & visualisation