Modelling Hail and Convective Storms With WRF for Wind Energy Applications
Leading Edge Erosion (LEE, material loss) on wind turbine blades reduces electrical power production by up to 5% and causes substantial repair costs. LEE is linked to material stresses induced by impacts from falling hydrometeors. Here we present a detailed evaluation of the WRF model simulation of hail relative data from the RADAR observations for a region of the US with high wind turbine installed capacities.
Impact fatigue caused by collision with rain droplets and hailstones is a severe problem for wind turbine blades. Although rain droplets fall at only modest velocities (of the order of 10 ms-1), the tip of WT blades rotate quickly (70-140 ms-1), thus the impact velocity (and kinetic energy transfer) is large. Each impact on the blade leading edge adds to the accumulated material damage such that eventually, the leading edge material may crack, causing a loss in the coating material and degrading the blade aerodynamics (and hence lift and electricity protection). The resulting erosion of the wind turbine blade leading edge is a key and growing source of wind turbine maintenance and repair. There is evidence that at some sites >99% of the kinetic energy transferred by hydrometeor impacts on wind turbine blades could be from hail. Thus, to characterize the potential for wind turbine blade leading edge erosion at sites across the US it is essential to properly characterize the frequency, intensity, and characteristics of hail.
Until recently hail occurrence was poorly (and subjectively) reported, but the deployment of dual-polarization RADAR at National Weather Service (NWS) sites across the USA has revolutionized our detection abilities. Implementation of new microphysics parameterizations for the Weather Research and Forecasting (WRF) model have also greatly enhanced model capabilities. Here we present a detailed evaluation of the WRF simulation of hail relative data from the RADAR observations for a region of the US with high wind turbine installed capacities.
The WRF simulations exhibit some fidelity in terms of the occurrence of hail. For example, in the comparison of WRF hail output to RADAR observations in a 100-km circle centered at a single RADAR station in west Texas, WRF models the hourly presence of hail with a proportion correct of 0.73, and an odds ratio of 3.14. For the more intense events (i.e. hourly WRF hail accumulation and hourly number of RADAR hail reports above their respective median values), the proportion correct increases to 0.85, and the odds ratio to 4.85. However, the WRF simulations exhibit a positive bias in terms of both (1) the frequency of hail (WRF models the presence of hail in 30% more hours than are estimated from RADAR data) and (2) the spatial extent of hail. During hours of widespread hail, the spatial area over which WRF simulated hail is 3.5 times the area over which hail is indicated by the RADAR output. A larger validation exercise is underway which will include all nine of the RADAR stations contained within the WRF simulation domain. This larger data set will enable a better understanding of the sub-regional variability of the hail climate, and the accuracy of WRF at estimating the frequency and distribution of severity of hail events in this area of the Great Plains.