Identification of wind turbine clutter in weather radar measurements

London Array and Greater Gabbard are the first and second largest wind farms in the world. These off shore wind farms are located in the North Sea, to the East of London. They have a combined capacity of 1.13 GW and occupy an area of more than 100 km2 each. With continued progress in renewable energy production the off shore wind farms studied here are set to be expanded further, and with consented plans for wind farms of more than four times the size there could be even greater problems for the UK weather radar network.

The turbines stand 130m above sea level, but even if the main beam of the radar passes above this height the side lobes can be intercepted by the rotating blades causing significant reflectivity returns in higher elevation scans. Due to the large coverage of the wind farms, they appear as substantial regions of high reflectivity clutter. Of which, one of the farms appears only during anomalous propagation conditions, which makes the use of a standard clutter map redundant to remove the affected pixels. A clutter tail or shadow that resembles precipitation can be observed behind the wind farms on some occasions, so these pixels must also be correctly identified as clutter.

The network of UK Met office radars are in the process of being upgraded to dual-polarimetric capabilities. In this study data from a recently upgraded polarimetric radar located to the North West of London, UK, is utilised. The different polarimetric measurements are compared to assess their ability in aiding the differentiation between ground clutter, wind farm clutter, sea clutter, and precipitation. Membership functions are proposed from a large database of measurements and Fuzzy Logic is used to individually classify the pixels.

Current algorithms to classify ground clutter or anomalous propagation heavily rely on the texture of phidp for the classification. However, the texture of phidp for wind farms is slightly different from ground clutter and has some overlap with the texture of phidp obtained in precipitation. So normal clutter and precipitation membership functions and identification methods cannot be used, and wind farm clutter should be specifically targeted. Preliminary results show a good ability to initially separate clutter from meteorological returns, and then the algorithm shows some skill in identifying the clutter source as sea, ground or wind farm.