A Surrogate Assisted Hooke-Jeeves Algorithm to Optimize the Control System of a PMSM Drive

This paper presents the optimization of a vector controlled Permanent Magnet Synchronous Motor (PMSM). The optimization is carried out by a Surrogate Assisted Hooke-Jeeves Algorithm (SAHJA). The SAHJA is a local searcher having a steepest descent pivot rule which employs both the real fitness and an approximated model (surrogate), in a cooperative way, in order to perform the optimization saving calculation time. The real fitness of each set of control parameters is evaluated by means of a simulated test which takes a few seconds to be executed. In a combined way, a computationally cheap approximated function, generated by means of the least square method, is employed. The numerical results show that the usage of the SAHJA leads to a significant reduction in terms of computational cost with respect to the classical Hooke-Jeeves algorithm, still maintaining high performance in terms of reliability.