Employing magnet-less motors such as switched reluctance motors (SRMs) in electric and hybrid electric vehicles (EV/HEVs) has gained attention and popularity in industry and academia, due to lower cost and more robust performance. In order to achieve the highest performance with the lowest manufacturing cost, optimal design of a SRM to meet specific characteristics for certain application involves optimization of both motor geometry and control strategy simultaneously. The most important control feature of a SRM that distinguishes it from other motors is its controllable firing angles. Inappropriate firing angles can lead to non-optimal operation of the SRM which eventually results in lower average output torque and higher torque ripple. In this situation, a common solution is to increase the motor stack length to meet the required average torque which will make the SRM oversized and costly. In this presentation, a constrained multi-objective optimization framework for design and control of a SRM, based on a non-dominated sorting genetic algorithm (NSGA-II) will be presented. The proposed optimization method optimally selects the firing angles to avoid oversizing the motor for specific torque characteristics. MotorSolve is an enabler for this optimization process which links MATLAB with Finite Element Analysis (FEA) by providing an environment to write the optimization code in MATLAB which leads to expedite the design process. The motor control and geometry variables are defined in MATLAB script, so that, for each iteration of the optimization, the motor will be plotted and simulated in MotorSolve.. All steps of programming in MATLAB and designing the SRM in MotorSolve will be presented.
What problems were resolved?
- MotorSolve simulates the SRM very quicker than the other FEA packages and the output characteristics is so close to 3D simulation.
- The feature of linking MATLAB with FEA makes it possible to enter the motor geometry and control variables in MATLAB.
- The SRM output characteristics can be imported into MATLAB which can be stored and analysed easier.
Bahar Anvari, Texas A&M University