Syracuse University SURFACE Dissertations - ALL SURFACE May 2015 Conceptual Design of Wind Farms Through Novel Multi-Objective Swarm Optimization Weiyang Tong Syracuse University Follow this and additional works at: https://surface.edu/etd Part of the Engineering Commons Recommended Citation Tong, Weiyang, "Conceptual Design of Wind Farms Through Novel Multi-Objective Swarm Optimization" (2015).edu/etd/243 This Dissertation is brought to you for free and open access by the SURFACE at SURFACE. It has been accepted for inclusion in Dissertations - ALL by an authorized administrator of SURFACE. For more information, please contact surface@syr. ABSTRACT Wind is one of the major sources of clean and renewable energy, and global wind energy has been experiencing a steady annual growth rate of more than 20% over the past decade.
energy market, although wind energy is one of the fastest increasing sources of electricity generation (by annual installed capacity addition), and is expected to play an important role in the future energy demographics of this country, it has also been plagued by project underperformance and concept-to-installation delays. There are various factors affecting the quality of a wind energy project, and most of these factors are strongly coupled in their influence on the socio-economic, production, and environmental objectives of a wind energy project. To develop wind farms that are profitable, reliable, and meet community acceptance, it is critical to accomplish balance between these objectives, and therefore a clean understanding of how different design and natural factors jointly impact these objectives is much needed. In this research, a Multi-objective Wind Farm Design (MOWFD) methodology is de- veloped, which analyzes and integrates the impact of various factors on the conceptual design of wind farms.
This methodology contributes three major advancements to the wind farm design paradigm: (I) provides a new understanding of the impact of key factors on the wind farm performance under the use of different wake models; (II) explores the crucial tradeoffs between energy production, cost of energy, and the quantitative role of land usage in wind farm layout optimization (WFLO); and (III) makes novel advancements on mixed-discrete particle swarm optimization algorithm through a multi-domain diversity preservation con- cept, to solve complex multi-objective optimization (MOO) problems. A comprehensive sensitivity analysis of the wind farm power generation is performed to understand and compare the impact of land configuration, installed capacity decisions, incoming wind speed, and ambient turbulence on the performance of conventional array layouts and optimized wind farm layouts. For array-like wind farms, the relative importance of each factor was found to vary significantly with the choice of wake models, i., appreciable differences in the sensitivity indices (of up to 70%) were observed across the different wake models. In contrast, for optimized wind farm layouts, the choice of wake models was observed to have no significant impact on the sensitivity indices.
The MOWFD methodology is designed to explore the tradeoffs between the concerned performance objectives and simultaneously optimize the location of turbines, the type of turbines, and the land usage. More importantly, it facilitates WFLO without prescribed conditions (e., fixed wind farm boundaries and number of turbines), thereby allowing a more flexible exploration of the feasible layout solutions than is possible with other existing WFLO methodologies. In addition, a novel parameterization of the Pareto is performed to quantitatively explore how the best tradeoffs between energy production and land usage vary with the installed capacity decisions. The key to the various complex MO-WFLOs performed here is the unique set of capabilities offered by the new Multi-Objective Mixed-Discrete Particle Swarm Optimization (MO-MDPSO) algorithm, developed, tested and extensively used in this dissertation.
The MO-MDPSO algorithm is capable of dealing with a plethora of problem com- plexities, namely: multiple highly nonlinear objectives, constraints, high design space di- mensionality, and a mixture of continuous and discrete design variables. Prior to applying MO-MDPSO to effectively solve complex WFLO problems, this new algorithm was tested on a large and diverse suite of popular benchmark problems; the convergence and Pareto cov- erage offered by this algorithm was found to be competitive with some of the most popular MOO algorithms (e. The unique potential of the MO-MDPSO algorithm is further established through application to the following complex practical engineering problems: (I) a disc brake design problem, (II) a multi-objective wind farm layout optimization problem, simultaneously optimizing the location of turbines, the selection of turbine types, and the site orientation, and (III) simultaneously minimizing land usage and maximizing capacity factors under varying land plot availability. CONCEPTUAL DESIGN OF WIND FARMS THROUGH NOVEL MULTI-OBJECTIVE SWARM OPTIMIZATION By Weiyang Tong B., Syracuse University, 2011 Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Mechanical Engineering in the Graduate School of Syracuse University Syracuse University Syracuse, New York May, 2015 c Copyright 2015 by Weiyang Tong All Rights Reserved DEDICATION I dedicate this thesis to my maternal grandmother, Yushan Wang, who raised me, loved me, and always believed in me.
I also dedicate this thesis to my paternal grandmother, Anqi Yang, who was a great woman of endless patience, eternal kindness, and boundless love. vi ACKNOWLEDGMENT I would like to express my deepest appreciation towards my advisor, Prof. Achille Messac, for his immense help throughout my doctoral studies. Messac provided me with invaluable advice and technical supervision that formed the foundation of the research presented in this dissertation and in the several well regarded journal articles that I have authored/co-authored thereof.
He also inculcated in me a spirit of professionalism that greatly contributed to my professional growth. I am thankful for his devotion to my future. Without his guidance, and persistent support, this dissertation would not have been possible. I would also like also to thank my co-advisor, Dr.
Souma has been a tremendous mentor, an excellent colleague, and a great friend. I am thankful for the superb example he set as an outstanding scholar and former student of Prof. The enthusiasm, inspiration, and sharp insights he has on research will always remain an excellent source of motivation for me in my future career. I would like to thank my doctoral committee members, Prof.
Utpal Roy, Prof. John Dannenhoffer, Prof. Jeongmin Ahn, Prof. Benjamin Akih-Kumgeh, and my committee chair Prof.
Can Isik, for their valuable advice and comments, as well as their willingness to serve on my committee. Special thanks to Prof. Roy and Prof. Dannenhoffer, who have been supportive in many ways within the MAE department in Syracuse University.
I wish to extend my warmest thanks to my former and present colleagues, Dr. Jie Zhang, Dr. Junqiang Zhang, Samuel Notaro, and my dear friend Ali Mehmani, who have helped me in many different ways at the Multidisciplinary Design and Optimization Labo- ratory. I greatly appreciate their friendship, and their contributions to my research and this vii dissertation.
I am also grateful to my closest friends, Xiaomeng Li, Ang Gao, Jia Li, Xu Meng, Zi Wang, Bensong Yu, and Zhen Liu, who cheered me up even in the worst of times and made my life (far away from home) fun. Special thanks to “Fly Empire” and “Starkville Soccer Group”, you gave me a lot of happiness and helped through the tough times. Sponsorship of this work by the National Science Foundation awards CMMI-1100948, and CMMI-1437746 is also gratefully acknowledged. These acknowledgements would not be complete without thanking the wonderful staff at Syracuse University, including Kathleen Datthyn-Madigan, Kimberly Drumm-Underwood, Kristin Shapiro, Linda Manzano, and Deborah Brown at the MAE department, and Cathy Mentor at the Sluztker Center, for all their efforts.
Finally, my deepest thanks go to my family; I would like to express my sincere gratitude to my parents, Lian Yu and Jun Tong, whose love and encouragement have always been my greatest strength; and I am also grateful to my uncle Fu Tong and my aunt Xiulian Zheng, who have always been supportive of all my academic endeavors. viii CONTENTS DEDICATION. vii LIST OF TABLES. xiii LIST OF FIGURES.
xiv LIST OF ACRONYMS. xvi I Technical Preliminaries xviii 1. Research Motivation and Objective .1 Overview of Wind Farm Development .2 Conceptual Design of Wind Farms .1 Wind Farm Development Process .2 Role of Land Resource .3 Multi-Objective Mixed-Discrete Optimization Problems .1 Swarm-based Algorithms .4 Research Goals and Impact .1 Analyzing the Sensitivity of Wind Farm Power Output to Key Factors .2 Multi-Objective Wind Farm Design Framework .3 Land Use Related Considerations .4 Multi-Objective Mixed-Discrete Particle Swarm Optimization 18 1.1 The Wake Effects .1 The Role of Wake Effects in Wind Farm Power Estimation .2 Analytical Wake Models .2 The State of the Art in Wind Farm Layout Optimization .1 Overview of Wind Farm Layout Optimization Frameworks .2 Performance Criteria in WFLO .3 Optimization Algorithms in WFLO .2 Particle Swarm Optimization Algorithms .3 Simulated Annealing Algorithm .3 Multi-Objective Particle Swarm Optimization (MOPSO) .1 Overview of MOPSO .2 Search Strategies in MOPSO .4 Research Observations and Needs .1 Research Needs in Wind Farm Power Estimation .2 Research Needs in Wind Farm Design .3 Research Needs in the Multi-Objective Optimization Solver 45 II A Novel Approach to the Conceptual Design of Wind Farms 46 3. Primary Performance Objectives in Wind Farm Design .1 Annual Energy Production .2 Wind Farm Cost of Energy.
Identifying Key Factors Influencing Wind Farm Performance .1 Impact of Different Analytical Wake Models on Wind Farm Power Estimation 56 4.2 Single Wake Analysis .3 Wind Farm Power Generation Analysis .1 Power Variation with the Land Area per Turbine .2 Power Variation with the Incoming Wind Speed .2 Sensitivity Analysis of Wind Farm Power Output .1 Overview of the Extended Fourier Amplitude Sensitivity Test .2 Upper and Lower Bounds of Input Parameters .3 Numerical Experiment I: Sensitivity Analysis of the Power Output of Wind Farms with Array-Like Layouts .4 Numerical Experiment II: Sensitivity Analysis on Maximized Farm Output with Optimal Layouts. Developing the Multi-Objective Wind Farm Design Methodology .1 Implementation of MOWFD Methodology .2 Case Study: Multi-Objective Wind Farm Design .1 Pareto Shifting Technique .2 Result and Discussion. Multi-Objective Wind Farm Design Considering Land Usage .1 Developing a Consolidated Visualization Platform for Co-operative Decision- Making in Wind Farm planning .1 Description and Settings .2 Results and Discussion. 95 III Development of Multi-objective Mixed-Discrete Optimiza- tion Solver 98 7.
Development of the Multi-Objective Mixed-Discrete Particle Swarm Optimization Algorithm .1 Overview of the Single-Objective Mixed-Discrete Particle Swarm Optimiza- tion Algorithm .1 Overview of Single Objective MDPSO .2 Introducing the Multi-Objective Capability to Mixed-Discrete PSO .3 The Multi-domain Diversity Preservation in Multi-Objective Mixed- Discrete Particle Swarm Optimization (MO-MDPSO) .4 Roles of Diversity Preservation Coefficients .1 Numerical Experiments with Continuous Benchmark Problems .3 Results and Discussion .1 Class I: Unconstrained Continuous Bi-objective Optimization Problems .2 Class II: Constrained Continuous Bi-objective Optimization Problems .3 Numerical Experiment with Mixed Integer and Practical Multi-Objective Op- timization (MOO) Problems .1 Results of Mixed-Integer MOO Problems. Practical Application using the Multi-Objective Mixed-Discrete Particle Swarm Optimization Algorithm .1 Disc Brake Design .2 Multi-Objective Wind Farm Layout Optimization .3 Multi-Objective Wind Farm Optimization Considering Different Land Plot Availability. Conclusion and Future Work .1 Multi-Objective Wind Farm Design .2 Consideration of Land Configuration .3 Parameterization of Key Tradeoffs in Wind Farm Design .4 Multi-Objective Mixed-Discrete Particle Swarm Optimization .1 Quantification of Wind Farm Performance .2 Implementation of Parameterization of Tradeoffs .3 Multi-Domain Diversity Preservation in MO-MDPSO. 148 xii LIST OF TABLES 1.1 Capital Cost Breakdown for Typical Onshore/Offshor Wind Energy Projects in 2011 [11] .1 Comparison of computation time of wake simulation for two turbines in line [32] 25 4.1 Analytical wake model inputs .2 Specifications of “GE 1.5 MW xle” turbine [122] .3 Upper and lower bounds of natural factors .4 Upper and lower bounds of design factors .1 User-defined parameters in MDPSO .2 Parameterization of CF-LAMI Tradeoff .5 MW xle Turbine [122] .1 User-defined parameters in MO-MDPSO .2 Continuous unconstrained bi-objective optimization problems .3 Continuous constrained bi-objective optimization problems .4 Accuracy (Γ) metric for test problems in Class I .5 Uniformity (∆) metric for test problems in Class I .6 Performance indicators for Class II .