Efficiency Optimization of New Stator-doubly-fed Doubly Salient Motor by Using Extreme Learning Machine
Yagang Shu · Proceedings of the CSEE · 2009
Satisfy the demands of the high efficiency of electric vehicles (EVs), the stator-doubly-fed doubly salient (SDFDS) motor for EVs can run with high efficiency in the whole velocity range by the coordinate control of field current and armature current. But the nonlinear relationship among field current, armature current, torque and speed makes the realization of the coordinate control difficult. As a new learning algorithm with fast speed and good generalization, the extreme learning machine for single-hidden layer feed forward neural networks (SLFNs) can solve the nonlinear relationship effectively. Thus, a coordinated controller of field current and armature current based on extreme learning machine is proposed, in which the SLFNs are trained off-line by the experimental data of highest efficiency, which composes neural networks with single hidden layer with two inputs and two outputs. And the convergence results of training, which are nodes and output weights, are applied for on-line control afterward. The experimental results show that with the proposed scheme the SDFDS motor could obtain apparently efficiency optimization in the whole velocity range.