A comparative study of techniques utilized in analysis of wind turbine data

Power produced by a wind turbine is dependent on many factors with different importance degrees. Main factors can be found by a thorough analysis of all the factors and their correlation and impact on the main output. Therefore, it is important to monitor the performance of the wind turbines in order to minimize the operation and maintenance costs by pointing out abnormalities. This paper analyzes the main factors affecting active output power of a wind turbine which are Gearbox Temperature, Pitch Angle, Rotor Speed and Wind Speed. The data are measured over a 12-month period. Several techniques, Kohonen Maps, Multilayer Perceptron, Decision Trees and Rough Sets, are applied to these data. The objective is to show a comparison of different techniques, their positive and negative points and give the reader the ability to choose the best technique for the study based on the their advantages and disadvantages. For the assessment of data, MATLAB and WEKA software are utilized. Each study presents its accuracy based on the output error.

I. Introduction

WIND energy has great potential to satisfy a significant part of energy demanded from society. The numerous numbers of research performed in this area confirm this assertion. However, in order to reach the expected availability of wind farms, their reliability and applied maintenance should receive special attention from the management perspective [1]. Particularly it is important to detect whether an anomaly is present in a wind turbine in order to prevent anomaly's impact on the wind turbine's availability.