Abstract
In-stream tidal devices are ready to be deployed, yet the largest operational commercial array is limited to 6MW. Upcoming government support should see the size of such arrays increase by orders of magnitude, and thus the optimal placement of turbines within tidal arrays is an emerging challenge for successful commercial integration. Hydrodynamic models are required to predict the power produced by an array and the impact on the surrounding environment. The influence of common model inputs to layout optimisation are investigated herein. This is achieved using a shallow water equation based tidal array modelling framework, Thetis, coupled with a low cost analytical wake model (FLORIS) that allows for rapid assessment of the impact of small changes in hydrodynamic results on array micro-siting. The sensitivity of array optimisation at an intermediate development point (43 turbines) is interrogated through both artificial flow field manipulation and the variation of model input pertinent to the optimisation. A small margin exists in which an optimised array layout will perform efficiently for a deviation in flow prediction accuracy. However, incorrect flow predictions by a range sensitive to model inputs had a reduction of almost 8% on array efficiency relative to a control case. The sensitivity of flow field variance by model input changes, on extractable energy and array layout are substantial. Comparing arrays sited on flow fields using different bathymetry resolution leads to a discrepancy on average of over 2% to average array power. Arrays sited for different mesh resolution and friction representation also see average changes up to and exceeding 0.75%. For array developers and the future of this nascent industry, re-calibration of the model at every stage of data collection coupled with early acquisition of bathymetry data is critical not just for accurate power quantification, but also array efficiency. At a regional scale, quantification of the potential tidal resource must consider the consequences of uncertainty in model input data and site constraints that cannot yet be accounted for.