Automated Optimisation Method for Wind Farm Noise

Introduction

Environmental noise frequently represents one of the most controversial and restrictive aspects of proposed wind energy developments. Wind farm noise assessments require a careful balance to be struck between utilising the full available generating potential of a site and controlling the risks of excess noise generation. As developable land suitable for wind farms becomes less and less available, more challenging sites are having to be selected.

At the same time, improvements in turbine technology have meant that a variety of noise-controlled modes are available for modern variable speed machines. This allows noise reductions to be applied, in the required conditions, at a marginal but sometimes appreciable energy cost.


Approach

Historically, the operational strategy for a wind farm would be determined manually such that the consented noise limits would not be exceeded at neighbouring noise-sensitive locations, based on predictive models which now increasingly take into account not only wind speeds but also wind direction.

But there can be different strategies that result in compliance, and it is sometimes difficult to determine the optimal one to take. Which is better: fewer turbines operating without any constraint or a greater number with some operational constraint?

The authors have developed a new approach to this challenge, by considering not just noise but also the associated energy generation of each turbine, and determining a solution using iterative optimisation techniques.

Main body of abstract

The proposed method starts by considering each specific wind condition (wind speed and direction). The noise associated with the different turbines modes is then determined using models developed based on HLA’s extensive experience in wind turbine noise propagation. But the associated power production of each strategy can also be determined using the manufacturer supplied power curves.

It is then possible to consider the efficiency of different operational strategies, and seek to determine one which maximise power generation whilst still meeting the required noise limits, using a “branch and bound” discrete optimisation algorithm. This method considers all locations and turbines together and systematically, which is not possible manually in practice, and explores the large number of theoretical possibilities in the most optimal way by constructing a “tree” of different possibilities and excluding less promising approaches. This algorithm was found to be realistically applicable even for large wind farm projects.


Conclusion

Operational strategies developed using this approach have in simple cases shown improvements in the potential energy generation of a wind farm of as much as 2% over a standard approach, which can be significant over the life of a project.

The authors are of the view that this novel approach to wind farm operational optimisation in respect of noise versus energy losses represents a clear improvement on current standard methods, and offers more confidence that a good operational strategy has been determined for a project.

The model is based on standard assumptions, which could be developed further in future research. For example, it assumes uniform wind conditions across a site; through detailed operational knowledge of the relative wind speeds actually experienced between individual turbines, with ongoing involvement over the initial operational life of a wind farm, the optimisation process could be further developed.



Learning objectives
Inefficient operation of a wind farm can result in avoidable loss of revenue. The operation of turbines in a way in which compliance with noise limits is predicted, whilst at the same time considering the consequential implications on energy generation, can help minimise unnecessary losses. The paper will explain a new approach which combines both aspects and objectively determines an optimal approach.