It is well known that the layout of a large wind turbine cluster aects the energyoutput of the wind farm. The individual placement and distances betweenturbines will in uence the wake spreading and the wind velocity decit. Manyanalytical models and simulations have been made trying to calculate this, butstill there is a lack of experimental data to conrm the models. This thesis isdescribing the preparations and the execution of an experiment that has beenconducted using about 250 small rotating turbine models in a wind tunnel. Theturbine models were developed before the experiment and the characteristicswere investigated. The main focus was laid on special eects occurring in largewind turbine clusters, which were named Global Turbine Array Eects.It was shown that the upstream wind was little aected by a large windfarm downstream, even though there existed a small dierence in wind speedbetween the undisturbed free stream and the wind that arrived to the rstturbines in the wind farm. The dierence in wind speed was shown to beunder 1% of the undisturbed free stream. It was also shown that the densityof the wind farm was related to the reduced wind velocity, with a more densefarm the reduction could get up to 2.5% of the undisturbed free stream at theupstream center turbine. Less velocity decit was observed at the upstreamcorner turbines in the wind farm.When using small rotating turbine models some scaling requirements hadto be considered to make the experiment adaptable to reality. It was concludedthat the thrust coecient of the turbine models was the most important parameterwhen analysing the eects. One problem discussed was the low Reynoldsnumber, an eect always present in wind tunnel studies on small wind turbinemodels.A preliminary investigation of a photo measuring technique was also performed,but the technique was not fully developed. The idea was to take oneor a few photos instantaneously and then calculate the individual rotationalspeed of all the turbine models. It was dicult to apply the technique becauseof uctuations in rotational speed during the experiment, therefore thecalculated values could not represent the mean value over a longer time period.