Abstract: The application of Evolutionary Algorithms (EAs) to wind turbine blade design can be interesting, by reducing the number of aerodynamic-to-structural design loops in the conventional design process, hence reducing the design time and cost. Recent developments showed satisfactory results with this approach, mostly combining Genetic Algorithms (GAs) with the Blade Element Momentum (BEM) theory. The general objective of the present work is to define and evaluate a design methodology for the rotor blade geometry in order to maximize the energy production of wind turbines and minimize the mass of the blade itself, using for that purpose stochastic multi-objective optimization methods. Therefore, the multi-objective optimization problem and its constraints were formulated, and the vector representation of the optimization parameters was defined. An optimization benchmark problem was proposed, which represents the wind conditions and present wind turbine concepts found in Brazil. This problem was used as a test-bed for the performance comparison of several metaheuristics, and also for the validation of the defined design methodology. A variable speed pitch-controlled 2.5 MW Direct-Drive Synchronous Generator (DDSG) turbine with a rotor diameter of 120 m was chosen as concept. Five different Multi-objective Evolutionary Algorithms (MOEAs) were selected for evaluation in solving this benchmark problem: Non-dominated Sorting Genetic Algorithm version II (NSGA-II), Quantum-inspired Multi-objective Evolutionary Algorithm (QMEA), two approaches of the Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D), and Multi-objective Optimization Differential Evolution Algorithm (MODE). The results have shown that the two best performing techniques in this type of problem are NSGA-II and MOEA/D, one having more spread and evenly spaced solutions, and the other having a better convergence in the region of interest. QMEA was the worst MOEA in convergence and MODE the worst one in solutions distribution. But the differences in overall performance were slight, because the algorithms have alternated their positions in the evaluation rank of each metric. This was also evident by the fact that the known Pareto Front (PF) consisted of solutions from several techniques, with each dominating a different region of the objective space. Detailed analysis of the best blade design showed that the output of the design methodology is feasible in practice, given that flow conditions and operational features of the rotor were as desired, and also that the blade geometry is very smooth and easy to manufacture. Moreover, this geometry is easily exported to a Computer-Aided Design (CAD) or Computer-Aided Engineering (CAE) software. In this way, the design methodology defined by the present work was validated.
| Identifer | oai:union.ndltd.org:IBICT/oai:dspace.c3sl.ufpr.br:1884/30337 |
| Date | 05 June 2013 |
| Creators | Vianna Neto, Julio Xavier |
| Contributors | Coelho, Leandro dos Santos, Universidade Federal do Paraná. Setor de Tecnologia. Programa de Pós-Graduaçao em Engenharia Elétrica |
| Source Sets | IBICT Brazilian ETDs |
| Language | Inglês |
| Detected Language | English |
| Type | info:eu-repo/semantics/publishedVersion, info:eu-repo/semantics/masterThesis |
| Format | application/pdf |
| Source | reponame:Repositório Institucional da UFPR, instname:Universidade Federal do Paraná, instacron:UFPR |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0019 seconds