A new heuristic mutation operator for wind farm layouts is proposed.
The operator uses machine learning build a model of velocity deficits across a wind farm.
New layouts are generated by moving turbines to positions of low predicted velocity deficit.
The operator is evaluated in conjunction with an evolutionary strategy on five wind farm scenarios.
Results show an improvement compared to other standard approaches for wind farm layout optimisation.
Correct placement of turbines in a wind farm is a critical issue in wind farm design optimisation. While traditional “trial and error”-based approaches suffice for small layouts, automated approaches are required for larger wind farms with turbines numbering in the hundreds. In this paper we propose an evolutionary strategy with a novel mutation operator for identifying wind farm layouts that minimise expected velocity deficit due to wake effects. The mutation operator is based on constructing a predictive model of velocity deficits across a layout so that mutations are inherently biased towards better layouts. This makes the operator informed rather than randomised. We perform a comprehensive evaluation of our approach on five challenging simulated scenarios using a simulation approach acceptable to industry . We then compare our algorithm against two baseline approaches including the Turbine Displacement Algorithm . Our results indicate that our informed mutation approach works effectively, with our approach identifying layouts with the lowest aggregate velocity deficits on all five test scenarios.
- Wind farm;
- Layout optimisation;
- Velocity deficit;
- Wake effect;
- Evolutionary strategy;
- Informed mutation operator;
- Turbine displacement algorithm
Copyright © 2015 Elsevier Ltd. All rights reserved.