|
Abstract:
|
Statistics plays a crucial part in renewable energy research, e.g. building an efficient wind turbine calls for a comprehensive analyzes of the local wind speed distribution. This paper details outcomes of a year-long research by the faculty/student team at LaGuardia Community College, originally presented at the JSM 2017 conference. Our goal of designing a small horizontal axis wind turbine optimized for the wind pattern on the college roof requires a close statistical study of the wind. While collecting and validating short term wind speed measurements at our site for transforming (via. correlation method) long-term data from a nearby airport into a reliable local time series, we determine the best fitting probability distribution model for the airport data in order to gain insight into the wind pattern at our own (similar) location, fine-tune our measurements protocol and make a preliminary estimate of the potential energy output of our turbine. Surprisingly, Weibull distribution with parameters estimated by the Maximum Likelihood method provides an inadequate fit for the data. The Maximum Goodness of Fit and the Quantile methods both fare better in determining parameters of the Weibull model, while Gumbel and Logistic distributions provide a better fit. We apply continuity correction to simulate wind speed data that agrees with, but does not suffer from the deficiencies of, the raw data set. Consequently, Weibull model with parameters estimated by the Maximum Likelihood method provides the best fit. Airport data taken at ground level yields average wind speed estimate insufficient to justify the installation of the turbine. By rescaling the data for the height at which we propose to install the turbine we show sufficient wind speed average. Using the best fitting probability model, we predict the average annual energy output for a proposed turbine.
|
