Reducing computational effort in the calculation of annual energy produced in wind farms

Metaheuristic methods are commonly used in the optimization of wind farms by means of turbine micro-siting. The typical pattern search used by these methods to explore the solution space makes it necessary to repeatedly evaluate the objective function (and hence the annual energy produced by the wind plant under optimization) a large number of times. For each case, before evaluating energy production, it is necessary to calculate the wind speed deficit at the position of each turbine due to the wake effect: a very time-consuming task. This paper presents a set of algorithms and ideas to reduce the computational time spent on the calculation of the wakes, and therefore in assessing the annual energy production of a wind farm. The improvements proposed here can be applied to any wind turbine layout, but in the case of rhomboidal-type arrangements of turbines, their regularity leads to the achievement of much greater reductions in the computation effort. Furthermore, for this case, the computer time results almost independent of the number of turbines. These improvements have been successfully applied to a set of cases, and show that the computation time for the calculation of the yearly energy production of a wind farm with 100 turbines can be reduced between 300 and more than 20,000 times, by using the most efficient of the strategies proposed. Crown Copyright (C) 2014 Published by Elsevier Ltd. All rights reserved.