Using machine learning to predict wind turbine power output

Abstract

Wind turbine power output is known to be a strong function of wind speed, but is also affected by turbulence and shear. In this work, new aerostructural simulations of a generic 1.5 MW turbine are used to rank atmospheric influences on power output. Most significant is the hub height wind speed, followed by hub height turbulence intensity and then wind speed shear across the rotor disk. These simulation data are used to train regression trees that predict the turbine response for any combination of wind speed, turbulence intensity, and wind shear that might be expected at a turbine site. For a randomly selected atmospheric condition, the accuracy of the regression tree power predictions is three times higher than that from the traditional power curve methodology. The regression tree method can also be applied to turbine test data and used to predict turbine performance at a new site. No new data are required in comparison to the data that are usually collected for a wind resource assessment. Implementing the method requires turbine manufacturers to create a turbine regression tree model from test site data. Such an approach could significantly reduce bias in power predictions that arise because of the different turbulence and shear at the new site, compared to the test site.

Current gust forecasting techniques, developments and challenges

“…In recent years, studies in meteorology motivated by wind energy extraction have moved from locating areas of high resource to considerations of production efficiency, consistency and control of supply to energy grids, and turbine fatigue, all of which depend on the relative gustiness and spatial profile of the wind to which they are exposed (Clifton et al, 2013;Clifton and Wagner, 2014). The location of turbines at different heights increasingly far from the surface further drives research toward characterising the vertical profile of gust activity, rather than merely a screen level or 10 m prediction.…”

“…gravity wave breaking, deep convection) that are difficult to parametrise using reductionistic approaches. Application with gusts, however, often focusses on the detection (identification) of gusts presently occurring, for mitigation in flight control systems (Tedrake et al, 2009;Afridi et al, 2010;Antonakis et al, 2016), exposure of wind turbines to damage or power fluctuations (Spudic et al, 2009), or prediction of wind power variation as the result of turbulence and wind shear (Clifton et al, 2013;Clifton and Wagner, 2014). Studies in a meteorological context prove harder to find, but there are examples, and an indication that there is considerable promise in such approaches.…”

An error reduction algorithm to improve lidar turbulence estimates for wind energy

“…Although machine learning can be a useful tool for turbine power prediction (e.g., Clifton et al, 2013), it does not appear to be an ideal technique for correcting lidar TI error. Thus, the next steps in the development of L-TERRA will involve further refining the physics-based corrections in L-TERRA-S to improve TI estimates in a more robust manner.…”

“…Equation (3) can be simplified by letting U (z r )z −α r equal a constant β, as in Clifton et al (2013). The power law then becomes the following:…”

“…By averaging tens or hundreds of decision trees, the variance of the overall model is reduced significantly (Friedman et al, 2001). Random forests were evaluated because they are relatively easy to understand and have previously been used for wind energy applications (e.g., Clifton et al, 2013;Bulaevskaya et al, 2015). MARS is essentially a stepwise regression model, where different coefficients and basis functions are used to predict the output depending on each region in the dataset (Friedman, 1991).…”