Robust Design of an Axial-Flux Permanent Magnet Synchronous Generator Based on Many-Objective Optimization Approach

A new methodology is described for designing a robust optimum TORUS topology-based 10 kW axial-flux permanent magnet (NdFeB) synchronous generator for direct-coupled wind turbines. This methodology uses a modified version of the many-objective evolutionary algorithm non-dominated sorting genetic algorithm III (NSGA-III) in order to: 1) handle a mixed variables model; 2) maximize efficiency, minimize active material cost, weight, and outer diameter; and 3) make efficiency robust under uncertainties. Since the design procedure is correlated with dimensional and electromagnetic parameters, uncertainties that depreciate the generator efficiency may be introduced due to inaccuracies in construction and in field calculation. These likely uncertainties were taken into account in optimization by an additional objective function, where the search was guided by a robustness metric that minimizes the efficiency maximum estimated deviation. Thus, the novelty of this paper is the design of a robust optimum axial-flux generator using a tailored many-objective robust optimization method for mixed variables problems with parameters uncertainties.