An Evaluation of Graph Algorithms for the Wind Farm Cable Layout Problem under Electrical Aspects

Abstract

The task of the Wind Farm Cable Layout Problem is to design a cable system between turbines and substations such that all turbine output can be transmitted to the substations. This problem can be modelled with different levels of complexity. While a higher level of complexity yields solutions that can be implemented in a real-world setting more readily, problem instances also become more difficult to solve or even remain intractable. More simplistic models are easier to solve but their usability could be inhibited. One such more simplistic model for installation cost minimization contains a network flow and a suitable minimum-cost flow algorithm provides good cable layouts on instances with up to 500 turbines within tens of seconds. The question remains whether those cable layouts are suitable for electrical implementation as well. We propose a workflow to evaluate the cable layouts generated from such algorithms under electrical aspects. This workflow converts the output of cable layout optimization algorithms to power flow models. The power flow models are simulated using the simulation framework eASiMOV. The evaluation of the power flow simulations under electrical metrics shows that output from the minimum-cost flow algorithm and from an approach solving a Mixed-Integer Linear Program perform very well under electrical aspects on a vast majority of input instances. For the remaining minority we are able to identify structures in the solutions that result in a worse performance. These observations can be used by the algorithm engineers as possible directions for future improvements.