Statistical Models for Characterizing Heterogeneous Wake Effects in a Wind Farm

Wind energy;Wake effects;Spatial modeling;Heterogeneity;Industrial and Operations Engineering;Engineering;Industrial & Operations Engineering

You, MingdiJin, Judy ;

University of Michigan

关键词: Wind energy;    Wake effects;    Spatial modeling;    Heterogeneity;    Industrial and Operations Engineering;    Engineering;    Industrial & Operations Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/138788/mingdyou_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Wind energy is becoming one of the most promising renewable sources. With the rapid growth of wind energy, modern utility-scale wind farms consist of a large number of wind turbines. In order to improve the power generation efficiency of wind turbines, an accurate quantification of power generations from multiple turbines is critical in both wind farm design and operational controls. One challenging issue is that the power outputs from turbines are different from a stand-alone turbinebecause of complex interactions among turbines, known as the wake effects. In general, upstream turbines absorb kinetic energy from wind.Therefore, downstream turbines tend to produce less power than upstream turbines. Moreover, depending onweather conditions, the power deficits of downstream turbines exhibit heterogeneous patterns.In order to address these challenges, this dissertation study initiates two major ideas: (1) to analyze the stochastic nature of generating wind energy, this study avoids the traditional approach which focuses on wind field modeling within a wind farm.Instead, this study proposes new statistical approaches that directly model the relationship between free-flow wind conditions and power generations from multiple turbines; (2) to analyze the physical interactions among wind turbines, this dissertation proposes data-driven approaches in a comprehensive framework.The objective of this research is to provide a new integrativemethodology to characterize multi-turbines;; heterogeneous performance at a wind farm scale. Specifically, this dissertation develops: 1. a new statistical model for characterizing heterogeneous wake effects under the dominant wind direction;2. a canonical model-based approach to handle wake effects under different wind directions; and 3. a new method to quantify the improvement of power productions through retrofitting, e.g., the vortex generator VG installation.In a wind farm, interactions among turbines alter the power generation efficiency of turbines. Moreover the power deficits of downstream turbines in a wind farm exhibit heterogeneous patterns, depending on wind conditions. This study first characterizes heterogeneous wake effects under a dominant wind direction. The proposed model decomposes the power outputs into the average pattern commonly exhibited by all turbines and the turbine-to-turbine variability caused by multi-turbine interactions. To capture the interactions among turbines, turbine-specific regression parameters are modeled using a Gaussian Markov random field (GMRF).Second, the power curve of each turbine becomes heterogeneous when changes in wind directions cause some upstream turbines to become downstream turbines. This dissertationproposesan integrative methodology that quantifies the heterogeneous wake effects over a range of wind directions by utilizing the concept of canonical models and similarity functions.The direction-dependent multi-turbine power curves are modeled in a Bayesian hierarchical framework.Lastly, based on the model quantifying the wake effects proposed in this dissertation, a new method is introduced to quantify the retrofitting effect on the power generation performance. The result can help practitioners perform the cost and benefit analysis when they consider the retrofitting for existing turbines.The advantages of all proposed approaches are demonstrated with the data collected from operational wind farms. The results of case studies validate that the proposed models successfully resolve some issues observed from the real world data.