Technical notePlacement of wind turbines using genetic algorithms
Introduction
Two main reasons for interest in wind as an energy source are diminishing fossil fuel resources and the effect use of fossil fuel sources has on the environment. Wind energy, a widely available derivative of solar energy that has been captured by the earth’s atmosphere, is receiving considerable attention as an emission-free, low cost alternative to traditional energy sources. Considerable development has taken place in the design of wind energy conversion systems. Modern wind turbines are highly sophisticated machines built on the aerodynamic principles developed in the aerospace industry. Advanced materials and electronics have been incorporated into wind turbines designed to deliver energy across a wide range of wind speeds.
As rule of thumb, 10 ha/MW can be taken as the land requirement of wind farms, including infrastructure [1]. The spacing of a cluster of machines in a wind farm depends on the terrain, the wind direction and speed, and the turbine size. According to Patel, the optimal spacing is found in rows 8–12 rotor diameters apart in the windward direction, and 1.5–3 rotor diameters apart in the crosswind direction [2]. Ammara et al. [3] contended that this intuitive spacing scheme resulted in sparse wind farms that were inefficiently using the wind energy potential of the site. A dense, staggered siting scheme was proposed that would yield production similar to the sparse scheme, but would use less land. While this approach successfully reduced the land mass required for a given amount of wind turbines, the method of placement was still intuitive.
Mosetti et al. proposed a position optimization scheme based on genetic algorithms [4]. In this research, algorithms were developed for wind farm performance evaluation and optimization. The investment cost and the total power extracted were the variables optimized. The wind and cost models chosen were incomplex for the purposes of demonstrating the effectiveness of the optimization algorithm. While the power and efficiency calculations of the optimally placed wind turbines compare favorably with a greater number of randomly placed turbines, the optimal configurations presented do not yield even the simplest empirical placement schemes. This study seeks to determine the effectiveness of the genetic algorithm optimization procedure in identifying optimum configurations.
Section snippets
Wake and cost modeling
As in the Mosetti study, a wake model similar to the Jensen analysis is used for simplification of the wind field calculations [4], [5], [6], [7]. This wake analysis is based on the assumption that momentum is conserved inside the wake. In the analysis of a single wake, the near field behind the wind turbine is neglected making it possible to model the resulting wake as a turbulent wake or a negative jet. At the turbine, the wake has a radius equal to the turbine radius, rr. As the wake
Optimization
Genetic algorithms are probabilistic search algorithms combining the mechanics of natural selection and survival of the fittest. These algorithms are capable of efficiently finding an optimal solution for complex problems without necessitating reformulation for the evaluation of individual solution candidates. Unlike calculus-based methods, genetic algorithms are robust, global, and do not require the existence of derivatives for search. Enumerative schemes are lacking in robustness due to
Numerical procedure
A square grid divided into 100 possible turbine locations was used as the computational domain. The width of each cell, in the center of which a turbine would be placed, is equal to five rotor diameters, 5D, or 200 m, giving the domain dimensions of 50D×50D. Based on the dimensions of the computational domain, the maximum radius of the wake from a single turbine placed in the position (x, y)=(100 m, 100 m) (see Fig. 3 as a reference) is 189.9 m. The width of each cell, in the center of which a
Case (a)
The optimal solution for Case (a) can be derived heuristically. The domain under consideration for this study uses a wake model that increases in diameter only as a function of downstream distance. A simple optimization of one 10-cell column in the computational domain can therefore be projected across the entire domain in order to find the optimal solution for this simple wind scenario. This optimization produced an optimal configuration of three turbines in positions 1, 6 and 10 as shown in
Conclusions
The results of the present study demonstrate that genetic algorithms could accurately predict optimal wind farm configurations. In the simple case of uniform unidirectional wind, heuristic arguments verify the optimal configuration of the present study, which is at odds with that of Mosetti et al. It is possible that their work did not run enough individuals for sufficient number of generations to achieve convergence. Although the genetic algorithm is an effective global search method, it can
References (9)
- et al.
On some of the design aspects of wind energy conversion systems
Energy Convers Manag
(2002) - et al.
Optimization of wind turbine positioning in large wind farms by means of a genetic algorithm
J Wind Eng Ind Aerodyn
(1994) Wind and Power Solar Systems
(1999)- et al.
A viscous three-dimensional differential/actuator-disk method for the aerodynamic analysis of wind farms
J Sol Energy Eng
(2002)
Cited by (663)
Multi-objective optimization of clustered wind farms based on potential game approach
2024, Ocean EngineeringOptimal placement of fixed hub height wind turbines in a wind farm using twin archive guided decomposition based multi-objective evolutionary algorithm
2024, Engineering Applications of Artificial IntelligenceMulti-objective turbine allocation on a wind farm site
2024, Applied EnergyAn optimization framework for wind farm layout design using CFD-based Kriging model
2024, Ocean EngineeringOptimal floating offshore wind farms for Mediterranean islands
2024, Renewable Energy