Achilleas Tsitsimelis: Design and optimization of AFPM generators with the use of evolutionary algorithms for grid-tied small wind turbines
As AFPM generators are appropriate for locally manufactured small wind turbines, they are commonly used in rural electrification applications. Due to that fact that in such cases the access to financial resources is limited, one basic issue is the optimization of the generator’s cost (and specifically its minimization). For this reason, the goal of this work is to extract design trends for low cost small wind applications with the help of optimization algorithms considering this constraint. The results that are presented in this presentation are be both qualitative and quantitative, based on simulations.
The optimization procedure of an electric machine’s initial design is a non linear complex problem with constraints and so, the evolutionary computation algorithms Particle Swarm Optimization (PSO) and HPSO that are efficient in converging to the total optimal solution are being used. In the second case, the Pareto Front plot provides a set of optimal solutions and gives information on trade-offs between performance parameters.
The performance indexes of the generator that are examined in the optimization procedure are the cost, mass, volume, efficiency and their combinations. The variables that are optimized in this work are the pole arc to pole pitch ratio, ai and the inner radius Rin to outer radius ratio Rout, kd. The AFPM generator is a double rotor single stator coreless machine. The rotor consists of two steel discs where Neodymium NdFeB permanent magnets are mounted. Permanent magnets are the most expensive material in this kind of construction and so the most important material to reduce the cost of. Two different magnet grades of NdFeB are examined, specifically, N40 and N45. Also some constraints are introduced, which in this specific case are construction constraints. The range of nominal powers from 3 kW to 6 kW are studied with a step of 500W.
Regarding the procedure, in MATLAB software, optimization algorithms call the initial design routine, which returns the value of the objective function. In this way the algorithms iterations converge to a point that the values of the optimized variables are minimized in the objective function, without satisfying the construction constraints penalties.