Surrogate-Based Optimization of Firing Angles for Switched Reluctance Motor

In this work, optimal firing angles of Switched Reluctance Motors (SRMs) are explored by surrogate-based optimization in order to minimize the torque ripple. Surrogate-based optimization is facilitated via Neural Networks (NNs) which are regression tools capable of learning complex multi-variate functions. Flux and torque calculations of a nonlinear 16/20 SRM are evaluated with a NN, and consequently the computation time is expedited by replacing the look-up tables of flux and torque with the surrogate NN model. An optimization algorithm is proposed to discover optimal firing angle objects to minimize the 16/20 SRM torque ripple for a certain electrical load requirement. The resulting optimal firing angles are also represented by simple NN models to expedite online control. Comprehensive simulation and experimental results are provided to validate the theoretical findings.