Gradient based optimization of Permanent Magnet generator design for a tidal turbine

An optimization of an analytical problem with nine variables is executed to find the optimal Permanent Magnet (PM) generator for a tidal turbine. A gradient based solver is used to find the minimum cost of active materials for the given design specifications. The MATLAB function fmincon is used, and the possible minimization algorithms available for this function are compared. As these solvers are only able to find a local minimum, a search is performed trying to find other minimas, both using a MultiStart procedure and using a Genetic Algorithm (GA). Losses are calculated for windings, stator laminations and rotor magnets and solid steel, and a constraint is put on efficiency. The cost effect of varying this constraint is investigated. Optimizations are done with both weight and material cost as objective function, and the different resulting designs are presented.