• Energy Research
• 2016-2025close

A novel method was developed to detect the optimal onshore wind farm layout driven by the characteristics of all commercially-available wind turbines. A huge number of turbine combinations (577) was processed, resulting in 22,721 generated layouts. Various assumptions and constraints were considered, mostly derived from the literature, including site features, wind conditions, and layout design. For the latter, an irregularly staggered turbine array configuration was assumed. Wake effects were simulated through the Jensen's model, while a typical turbine thrust coefficient curve as a function of wind speed was originally developed. A detailed cost model was used, with levelized cost of energy selected as primary and capacity factor as secondary objective function. The self-organizing maps were used to address a thorough analysis, proving to be a powerful means to straightforwardly achieve a comprehensive pattern of wind farm layout optimization. In general, the two optimization functions basically match, while for higher wind potential sites, increasing capacity factor did not necessarily result in decreasing levelized cost of energy. The latter may be minimised by reducing the total number of turbines or the overall wind farm capacity, as well as maximising rotor diameters or minimising rated wind speeds; increasing rated power or hub height is only beneficial for mid-potential sites. The mere maximisation of wind farm energy production is a misleading target, as corresponding to mid-to-high values of levelized cost of energy. In contrast to previous studies, the use of turbines with different rated power, rotor diameter or hub height should be avoided.

A wind turbine (WT) site optimization procedure was developed and applied on two different onshore and one offshore sites, supplied with tall met masts and belonging to exemplary wind climates. In addition to detecting the most suitable WT for each site, Kohonen's self-organizing maps (SOMs) were employed to improve investigation of those parameters mostly influencing site optimization. Three years (2013-2015) of 1-h vertical observations from local met masts and a database of 377 onshore and 23 offshore commercial WTs were used. As a result, maximizing capacity factor (CF) was confirmed as a good objective function, though not the best, which was minimizing levelized cost of energy (LCoE). In general, these two conditions do not necessarily match: variously setting WT parameters may either result in an LCoE reduction or CF increase, but both conditions do not occur concurrently. A key finding was that minimum LCoE cannot be achieved by indefinitely increasing the WT hub height, but rather through detection of an optimum value obtained as a unique solution of the optimization procedure. Furthermore, the capability of SOM to recognise the cluster structure of all parameters influencing WT site optimization shed further light on their mutual relationship, thus proving to be an ideal tool to address the non-convex nature of this issue.

The reliability of ERA5 reanalyses for directly predicting wind resources and energy production has been assessed against observations from six tall towers installed over very heterogeneous sites around the world. Scores were acceptable at the FINO3 (Germany) offshore platform for both wind speed (bias within 1%, r = 0.95−0.96) and capacity factor (CF, at worst biased by 6.70%) and at the flat and sea-level site of Cabauw (Netherlands) for both wind speed (bias within 7%, r = 0.93−0.94) and CF (bias within 6.82%). Conversely, due to the ERA5 limited resolution (~31 km), large under-predictions were found at the Boulder (US) and Ghoroghchi (Iran) mountain sites, and large over-predictions were found at the Wallaby Creek (Australia) forested site. Therefore, using ERA5 in place of higher-resolution regional reanalysis products or numerical weather prediction models should be avoided when addressing sites with high variation of topography and, in particular, land use. ERA5 scores at the Humansdorp (South Africa) coastal location were generally acceptable, at least for wind speed (bias of 14%, r = 0.84) if not for CF (biased by 20.84%). However, due to the inherent sea–land discontinuity resulting in large differences in both surface roughness and solar irradiation (and thus stability conditions), a particular caution should be paid when applying ERA5 over coastal locations.

A review spanning across a 40-year period (1978-2018) and including a total of 332 applications has been addressed on theoretical and empirical wind resource extrapolation models applied in wind energy, which can be grouped into three main families: (i) the logarithmic models; (ii) the Deaves and Harris (DH) model; (iii) the power law (PL). Applied over 96 very heterogeneous locations worldwide, models have been tested against observations at upper extrapolation height and assessed by location characteristics, extrapolation range skills, and application economical advantages. The logarithmic models can nowadays be considered unsuitable for extrapolating wind resource to hub height of current multi-MW WTs, mainly because exhibiting a limited extrapolation range capability (about 10-50m median bin). Finer scores in extrapolating wind resource (mean absolute bias of 3.3%) and in predicting energy output (10.1%) were achieved by the DH model, also showing remarkable extrapolation range skills (10-80m median bin). However, although among the most economical and forward-looking solutions, its need for accurate z0 assessment and u* observations resulted so far in great limitations to its large-scale application for wind energy purposes (less than 1%). Eventually, the PL confirmed the most reliable - and largely most commonly used (73.5%) - approach for wind energy applications. Out of the plethora of PL models developed in the literature, the PL(?)-?lower and the PL(?)-?I were the finest in predicting both extrapolated wind resource (mean absolute error of 4% and 4.4%, respectively) and energy output (8.9% and 5.5%), also exhibiting extrapolation range skills meeting modern WTs requirements. By contrast, the PL using ?=1/7 returned among the worst scores, yet resulting - since the simplest - the solution most frequently applied (19.6%). This study also demonstrated that extrapolation tools requiring the most expensive instrumentation equipment do not necessarily return the finest scores.