Deterministic and artificial intelligence approaches in optimizing permanent magnet generators for wind power applications

This paper presents the design procedure of a permanent magnet generator for a 20 kW wind turbine prototype. This work has been developed in the frame of a research project funded by the General Secretariat for Research and Technology of Greece concerning the design and construction of a gear-less wind turbine for both autonomous and interconnected operation with the electrical grid. In wind power applications, multipole permanent magnet generators have become very attractive especially in small ratings. Developments in magnetic materials, on the other hand, enabled remanence levels up to 1.2 T by using convenient Nd–Fe–B magnets resulting in competitive design of “peripheral” type generator structures. Such a design tends to replace the traditional “internal” configurations involving ferrites. The magnet cost remains the most important concern in medium scale applications that is why the present paper is mainly devoted to the permanent magnet material optimization. The method involves a preliminary design stage based on standard formulae in a first step and the introduction of a two-dimensional finite element code into an optimization procedure in a second step, in order to investigate the optimal shape design problem. A neural network approach has been applied in a third step in order to be compared with the deterministic technique.