Anomaly detection and treatment for meteorological and wind turbine power measurements

A comprehensive Quality Assurance (QA) process was implemented on the meteorological measurements and power output of the neighboring wind turbine at the Wind Energy Institute of Canada's wind farm on Prince Edward Island. The data, obtained from May 2013 to September 2021, include wind power, speed, direction, temperature, pressure, and relative humidity, averaged over 10 min. The first QA step was the detection of erroneous values due to well-known causes, such as the icing of the cup anemometers, the wake of the neighboring turbine, and the meteorological mast. Subsequent, and mainly novel, procedures were developed to capture systematic errors, nonphysical values, and repeated constant values, to assess the internal consistency, and to deal with abnormally high variation and erroneous small values in the data. Most of the data taken before the calibration of the anemometers in 2017 and some a few months after the calibration were invalid; a fact that shows the importance of yearly calibration. The installation of two instruments on the opposite sides of the mast provided "counterpart " measurements, which greatly facilitated the QA. We also offer some recommendations about the layout of the mast and closest wind turbine and the value of using sonic anemometers which are not affected by icing.