Comprehensive evaluation of wind speed distribution models: A case study for North Dakota sites

Accurate analysis of long term wind data is critical to the estimation of wind energy potential for a candidate location and its nearby area. Investigating the wind speed distribution is one critical task for this purpose. This paper presents a comprehensive evaluation on probability density functions for the wind speed data from five representative sites in North Dakota. Besides the popular Weibull and Rayleigh distributions, we also include other distributions such as gamma, lognormal, inverse Gaussian, and maximum entropy principle (MEP) derived probability density functions (PDFs). Six goodness-of-fit (GOF) statistics are used to determine the appropriate distributions for the wind speed data for each site. It is found that no particular distribution outperforms others for all five sites, while Rayleigh distribution performs poorly for most of the sites. Similar to other models, the performances of MEP-derived PDFs in fitting wind speed data varies from site to site. Also, the results demonstrate that MEP-derived PDFs are flexible and have the potential to capture other possible distribution patterns of wind speed data. Meanwhile, different GOF statistics may generate inconsistent ranking orders of fit performance among the candidate PDFs. In addition, one comprehensive metric that combines all individual statistics is proposed to rank the overall performance for the chosen statistical distributions.