One of the most difficult problems that wind turbine designers face today is estimating the dominant loads that will have the greatest influence on the design of the turbine structure. Often, these loads are caused by extreme events in the atmospheric inflow that are rare enough they may only occur once in 20 years. Just as often, they are caused by the worst case of interaction between normal turbulence and the turbine aeroelastic response. Their uncommonness makes extreme design loads nearly impossible to measure and difficult to simulate.

The wind energy community relies heavily on international design standards, such as the International Electrotechnical Commission (IEC) 61400-1, which specifies discrete loading conditions for turbine design. Unfortunately, these loading conditions are based on previous experience and measurements of extreme wind events that occurred in the USA and Europe. Thus, it is not clear if the design conditions derived from these limited measurements will cover the range of sites now being developed, as wind energy continues to grow worldwide.

In response to these limitations, loads extrapolation requirements were introduced into the IEC standard in 2005. Extrapolated loads are based on statistical fits to aeroelastic simulations driven by standard turbulence models, in addition to deterministic load cases. Because they can be tuned to specific site conditions and are driven by actual turbine control and dynamics, extrapolated loads are thought to be superior for determining extreme loads. Many users expressed some initial uncertainty about their implementation, so the IEC decided that further study of methods of load extrapolation was needed to improve consistency and facilitate proper application. The IEC Safety Standard committee commissioned the Loads Extrapolation and Evaluation Exercise (LE3) subcommittee to look at extrapolation techniques in more depth. The contents of this special issue are largely the results of studies originating from this subcommittee. There are also three papers in this issue not originating in the IEC group. They are the articles by Velkamp, Agarwal and Manuel, and by Larsen and Hansen.

All of the papers in this issue have one thing in common: the use of probabilistic methods, which are gaining prominence in many applications. Probabilistic methods have the advantage of being driven by turbulent wind descriptions that may be characteristic of a larger range of wind sites and conditions, but they also have the potential disadvantage of being more complex to implement, requiring more design data, and also possessing the potential for misapplication. Thus, there is a need for further research in this area; this special issue contains an excellent sampling of some of the latest work.

The paper by Moriarty, entitled ‘Database for validation of design load extrapolation techniques’, discusses two large data sets created for the LE3 subcommittee and used for the validation of load extrapolation techniques, and is the foundation for much of the work done by the LE3 subcommittee. The largest data set consists of 5 years of operational data for a 5 MW turbine, while the second data set represents a typical data set used by a designer in the design process. Analysis of both data sets reveals dominant wind speeds that must be thoroughly examined when using extrapolation techniques.

The paper of Natarajan et al., entitled ‘Identification of contemporaneous component loading for extrapolated primary loads in wind turbines’, discusses the important issue of combining extreme loads in orthogonal directions at a common location. An example would be choosing the blade edge root bending moment that occurs coincidentally with an extreme flap bending moment. Often, orthogonal loads are extrapolated separately, and the worst case is assumed to be the combination of both largest extremes, but Natarajan et al. provide a method for calculating a more realistic estimate.

The papers of both Moriarty and Freudenreich and Argyriadis focus on load partial safety factors used in the IEC standard. The motivation behind both of these papers is to find an equivalent safety factor that can be used as an alternative to the full loads extrapolation. One drawback of extrapolation is that it requires more simulation and post-processing than traditional methods. These safety factors may be a quicker method for obtaining estimates of the long-return period loads early in the design process with fewer simulations. Freudenreich and Argyriadis examine the effect of different turbine designs, control systems and turbulence levels on the required safety factor. Moriarty examines the dependence of the required safety factor on number of simulations and load location. Freudenreich and Argyriadis also compare the loads predicted using extrapolation and those with older deterministic methods. The extrapolated loads are often more conservative.

Fogle et al. examine the number of simulations required to obtain accurate estimates of long-term loads. The work includes a convergence criterion to indicate the number of simulations required for a good extrapolation estimate given an objective criterion. Fogle et al. explain that the required number is a function of the loading type and also of the dominant wind speeds, but fairly independent on the time separation between maxima.

The work of Collani et al. describes a new statistical method for loads extrapolation that uses a unique set of distributions to model the empirical data as opposed to the current standard of a single distribution. This method also inherently predicts the uncertainty in the long-term loading estimate.

The paper by Veldkamp, entitled ‘A probabilistic approach to wind turbine fatigue design’, is the only one in the special issue that focuses on fatigue design loads. He has studied the calibration and use of partial safety factors in common civil engineering practice. Wind turbine loads have high levels of uncertainty, which make them the dominant driver of the required safety factor. Veldkamp also introduces a simple cost model that can be used to optimally design turbines for economic rather than for safety considerations, which drive life-critical civil structures.

Agarwal and Manuel use statistical loads extrapolation to explore field data from an operating offshore turbine. The methods of extrapolation are similar to those applied to onshore turbines, but the environmental variables must also include the wave height. The effect of the additional wave variability on tower loads is explored.

Finally, the paper of Larsen, entitled ‘Rational calibration of four IEC extreme load cases’, examines the calibration of extreme environmental conditions such as gusts. In the current IEC standard, these extreme wind events are deterministic and based on measurements at just a few locations. Larsen introduces more statistically sound methods that can be site specific and that include the effects of complex terrain.

We hope this special issue provides a useful framework for the current state of design load determination in the wind energy industry, and we look forward to a continuing technical dialogue in this area of in the future.