Posters

Abstract

The energy yield assessment (EYA) of an offshore wind farm in the vicinity of another one(s) has become the norm as more and more wind farm projects are deployed in all the favourable offshore sites around the world. The effects of the neighbouring wind farms on the energy yield of a potential new site are already accounted in most of the EYA studies as part of the commonly labelled "external wake losses". However, there is little information, and most of it focused on the North Sea about the uncertainty and potential biases that these wind farms introduce to the measurement campaigns, hence, the wind resource and gross energy estimations of new offshore sites. Since these effects can be quantified with the same modelling tools used to assess the intra-farm interactions in the EYAs, it is essential to make sure that their predictions are precise and unbiased under different climate conditions, distances between farms, and wind farms characteristics, especially considering the extra-large projects that are being designed nowadays. Therefore, this work presents the validation of different numerical tools used for the wind resource estimation of a 200 km^2 region in the Moray Firth off the coast of Scotland. This region corresponds to the lease area of the future Caledonia wind farm project, which is located next to the 950MW Moray East wind farm project that is owned and operated by Ocean Winds. The measurements are obtained from two Floating Wind Lidars (FWL) deployed at the Caledonia site at 3 km and 10 km downstream from the last turbine of Moray East. Both FWL are aligned with the prevailing wind direction, hence, they experience constant wakes from Moray East. The wake modelling techniques evaluated in this case study are 3 multi-scale simulations and 3 engineering models run in time series mode. Two of the multi-scale simulations are based on the WRF model with the Fitch Wind Farm Parameterization, while the third simulation correspond to the GRASP-LES model with the Actuator Disk implementation. The engineering models included in this study are the TurbOPark, the NO-Jensen and the Eddy Viscosity model. Their outputs are connected to a time series obtained from a WRF mesosale run in order to obtain the wind speed values at the FWL and at each turbine location of Moray East in the time-domain. The validation exercise is performed for multiple weather episodes within a year of available data that is concurrent between the FWL and the 10-min SCADA of the Moray East turbines. This work also presents the comparison in the time domain between the power prediction and the actual production of the operational wind farm. Large discrepancies are found between the measurements and models, and among all the methodologies. When compared to the SCADA data of Moray East, the relative biases of power range between -5% to 14%, though an improvement from 0.66 to 0.74 is found in the coefficient of determination between methods based on physical downscaling compared to the engineering models.