Development of an Empirical Obstacle Wake Model for Small Wind Turbine Micrositing

Abstract

The way in which wind flows around and downstream of buildings in the atmospheric boundary layer is of great interest in wind engineering. Building wakes are characterized by a large reduction in downstream flow velocity and increased levels of downstream turbulence, both of which are detrimental to the power output of a small wind turbine. This paper describes the development of a new empirical model that predicts flow properties in the wake of a building for the purpose of small wind turbine micrositing. The new model has been developed in the form of a feedforward, backpropogation artificial neural network ($$). All training data was obtained from wind tunnel measurements taken in the wakes of model obstacles submersed in a simulated atmospheric boundary layer. To validate the wind tunnel simulation, a field experiment was carried out in which the flow was measured in the wake of an obstacle situated on an open field. The mean difference between the field data and wind tunnel data (from a geometrically similar test) was 3.7% for wind speed and 13.0% for the turbulence intensity. Overall, wind tunnel data was determined to be acceptable for use to develop a new model. The new model has mean errors of 0.6% and 4.4%, when predicting the velocity and root mean squared velocity, respectively (the mean error of the model is defined as the difference between the $$ predictions and the wind tunnel data). Currently, the model is somewhat limited in scope; wake properties can be predicted for solid ‘block’ obstacles, with varying height, width and depth, but only with flat roofs. Future work includes improving the model’s generality (the ability to predict wake properties when presented with inputs not used during training), training the model to predict the effect of roof shape, and investigating how to combine the wake effects of multiple obstacles.