Feature Extraction and Pattern Identification for Anemometer Condition Diagnosis

Abstract

Cup anemometers are commonly used for wind speed measurement in the wind industry. Anemometer malfunctions lead to excessive errors in measurement and directly influencethe wind energy development for a proposed wind farm site. In the PHM 2011 Data Challenge Competition, two types of data need to be processed for anemometer condition diagnosis: paired data consisting of wind data from paired anemometers, and shear data composed of measurements from an array of anemometers at different heights. Since the accuracy of anemometers can be severely affected by the environmental factors such as icing and the tubular tower itself, in order to distinguish the cause due to anemometer failures from these factors, our methodologies start with eliminating irregular data (outliers) under the influence of environmental factors. For paired data, the relation between the normalized wind speed difference and the wind direction is extracted as an important feature to reflect normal or abnormal behaviors of paired anemometers. Decisions regarding the condition of paired anemometers are made by comparing the features extracted from training and test data. For shear data, a power law model is fitted using the preprocessed and normalized data, and the sum of the squared residuals (SSR) is used to measure the health of an array of anemometers. Decisions are made by comparing the SSRs of training and test data. The performance of our proposed methods is evaluated through the competition website. As a final result, our team ranked the second place overall in both student and professional categories in this competition.