Date of Degree

2009

Document Type

Master's thesis

Degree Name

MS (Master of Science)

Department

Industrial Engineering

First Advisor

Andrew Kusiak

Abstract

The large-scale wind energy industry is relatively new and is rapidly expanding. The ability of a wind turbine to extract power from the wind is a function of three main factors: the measured wind speed, the power curve of the turbine, and the ability of the machine to handle wind fluctuations. The key parameter determining wind turbine performance is wind speed and it is normally measured with an anemometer placed at the nacelle of a turbine.

The dynamic nature of wind speed, however, is a barrier for applying predictive engineering in wind energy. Traditional approaches based on physical science and mathematical modelings have limitations on wind power prediction models. Conventional approach based on dynamic modeling has disadvantage of power generation process modeling due to time-shift nature of the process.

Data mining is a promising approach for modeling wind energy, e.g., power prediction and optimization, wind speed forecasting, power curve monitoring and fault diagnosis. It involves a number of steps including data pre-processing, data sampling, feature selection, dimension reduction and, etc. This thesis focus on applying data mining to predictive engineering in wind industry, and ultimately builds wind speed prediction and wind farm power prediction models, develops turbine dynamic control and power optimization strategy, explores methodology for system level fault diagnosis. However the philosophy, methods and frameworks discussed in this research can also be applied to other industrial processes.

This thesis proposes a series of predictive models under the framework of data mining. Chapter 2 introduces a methodology for short term wind speed prediction based on wind farm layout information. Chapter 3 and Chapter 4 present prediction models for wind turbine parameters. Chapter 5 proposes strategies for dynamic control of wind turbines. Chapter 6 explores the fault diagnosis and prediction using SCADA data.

Pages

xii, 147

Bibliography

141-147

Copyright

Copyright 2009 Wenyan Li