Design and meta-heuristic based optimization of axial-flux induction generator for variable speed wind turbines

Abstract

Renewable energy has inevitably become widespread due to the increasing world population, sustainability problems, and the green energy standards that are increasingly coming to the forefront. Therefore the demand for renewable energy resources such as geothermal, hydroelectric, solar, and wind energy is increasing day by day. Among these types of energy, wind energy stands out with features such as very high system efficiency and ease of application. Achieving much higher efficiencies in the electromechanical energy conversion of wind energy is made possible by innovative generator design approaches. At this point, axial-flux induction generators (AFIG) are one of the significant generator topologies that can meet the expectations from wind energy systems. AFIGs, in terms of their operating principle, are quite similar to the radial-flux induction machines frequently used in the industry. They contain three-phase AC windings in their stators and a rotor cage made of copper or aluminum material in their rotors. These machines do not contain rare earth magnets, and therefore, the sustainability problems in their manufacturing are less. Similar to many electrical machines, their core materials are chosen from soft ferromagnetic materials and subjected to lamination to minimize eddy current losses. Due to their use of axial-flux geometry, it is possible to design them with high power density and high efficiency. Additionally, the large outer diameter dimensions provided by the axial-flux geometry increase their moment of inertia. These features protect both the AFIGs and the grid from voltage and frequency fluctuations that may arise from sudden changes in wind speed. It is possible to derive different AFIG topologies using various stator and rotor combinations. The design of AFIGs is often subjected to multi-objective optimization, a common approach in engineering problems. This process involves assigning various geometric dimensions of the AFIG as optimization variables. Examples of these variables include the height and width of the stator and rotor slots, the length of the stator and rotor yokes, the inner and outer diameters of the machine, and the skew angles of the stator and rotor slots. By setting the objective functions to minimize total losses and core material volume, designs with high efficiency and high power density can be achieved. Contemporary designs that meet the high power density and efficiency requirements for wind turbine generators, and that can compete with various generator topologies, are attainable through this approach. Given the large number of variables and significant computational load, employing multi-objective metaheuristic optimization algorithms becomes necessary. Thanks to these algorithms, both the solution space will be obtained faster and it will be guaranteed to obtain results that meet the expected constraints. In this thesis, the designs and optimizations of two different AFIG topologies, namely double stator-single rotor axial-flux induction generator (DSSR-AFIG) and single stator-single rotor axial-flux induction generator (SSSR-AFIG), were performed using a metaheuristic algorithm type called Multi-Objective Grey Wolf Optimizer (MO-GWO). The established optimization algorithm uses the number of poles, conductor per slot value, and conductor diameter as local sweep variables. The global optimization variables are stator/rotor slot height and width, stator/rotor yoke height, inner and outer diameter lengths, and air-gap length. Different designs were examined by changing the weights of the objective functions of the optimization. First, a DSSR-AFIG design targeting 80% total loss reduction and 20% volume minimization, then a DSSR-AFIG design targeting 80% volume minimization and 20% total loss minimization, and finally a DSSR-AFIG design accepting both objective functions with equal weights were obtained. Lastly, the efficiencies and power density values of these designs were compared with the SSSR-AFIG topology with equal objective function weights. It was shown by both Finite Element Analysis (FEA) and the outputs of the developed algorithm that the DSSR-AFIG topology is superior in terms of both power density and efficiency. Additionally, the consistency and accuracy of the established algorithm were also proven by the performed FEAs.