Torque Ripple Minimization inaSensorless Switched Reluctance MotorBasedonFlexible Neural Networks
Changliang Xia · 2007
Theswitched reluctance motor(SRM)hasobtained greatpotential asanadjustable speedapplication duetoits outstanding merits. However, itsapplication islimited because of therotorposition sensors andtorque ripple. Thispaperproposes an approach totackle sensorless control andtorqueripple minimization ofSRM byusingflexible neural networks (FNN) whichhavemanygreat advantages, suchasless nervecells and quick learning speed. TwoFNN arebuilt: through measurement ofthephaseflux linkages andphasecurrents, thefirst oneisable toestimate therotorposition, thereby facilitating elimination of therotorposition sensor. Thesecond oneisfortheestimation of thereference currents witha desired torque, thenthereal currents inthearmatures areadjusted according tothereference values, therefore thetorqueripple generated bythenon-ideal currentwaveformsisminimized fora sensorless SRM. Simulation andexperimental results illustrate theimprovements oftheproposed methodcompared withtraditional controller.