Welcome to Task 23 – Background for Research

Task is currently inactive

Onshore turbines have an early history of poor reliability. In the early 80s many turbines were installed in various complex terrain sites in California. Frequent failures were common. At one point in the mid 80s a survey of gearbox health revealed nearly 90% of gear boxes had failed or were in some stage of failure. Support structures failed, blades failed, yaw drives and main frames failed. These same machines operated with a better level of reliability in other, less severe sites in Europe. Extensive testing revealed that the predicted loads for these machines underestimated the actual loads. Aerodynamic load models did not compare well with the dynamic loads from test data. Coupled structural response predictions did not compare well. Armed with convincing evidence and detailed measurements researchers began to improve the models. New aerodynamic features were added to capture some of the complexity. Model tuning using empirical data was introduced as a practical approach to improving predictions where engineering models failed to capture complex physical processes. Ultimately the models and load prediction procedures improved enough to dramatically improve turbine reliability. But it wasn’t without comprehensive model comparisons with both code to code (to isolate differences due to different executions of similar theory) and code to test data (a less precise comparison but more closely reflective of reality).

By comparing predictions from different codes with identical inputs, many of the output variations were traced back to surprising differences in the interpretation of nearly identical theories. These exercises have lead to dramatic improvements in load prediction by stimulating discussions about how to implement physical models within the context of wind turbine applications. The success of this process depends on carefully controlling all the inputs to the codes. With test data it is impossible to know all the inputs precisely and therefore direct comparisons are impossible. Usually statistical comparisons are more revealing but even these require comprehensive data sets for code validation. So analysts seek such comprehensive data sets from controlled turbine experiments where far more parameters are measured than can be justified for verification of a commercial prototype. . These data sets often are the result of research collaborations among many interested participants, which benefit the entire industry by improving reliability of all the products with improved analytical tools.

Recently, a new working group was formed under subtask 2 of IEA Annex 23, entitled “Offshore Model Comparisons”. This offshore modeling working group is envisioned to be just this kind of collaboration, for the benefit of the entire industry. Fueled by the success of the onshore wind energy industry, the offshore application of wind energy is generating great interest. Most experts expect similar success for the offshore industry. However, the offshore application is even more complex, fraught with new load prediction challenges such as waves, currents, different support structures and combining different stochastic loading sources in ways that are tractable in the design process. These issues can only be addressed with accurate models.

Currently conservative offshore design practices, adopted from marine industries, are enab ling offshore development to proceed but if offshore wind energy is to be economical, reserve margins must be quantified and uncertainties in the design process must be reduced so that appropriate margins can be applied. Uncertainties associated with load prediction are usually the largest source and hence the largest risk. Model comparison is the first step in quantifying and reducing load prediction uncertainties.