In this manuscript, a hybrid technique is proposed for Torque Ripple (TR) minimization of Switched Reluctance Motor (SRM). The proposed technique is the consolidation of Wingsuit flying search (WFS) optimization and Gradient Boosting Decision Tree (GBDT) algorithm, hence it is known as WFS-GBDT technique. The control mechanisms consists of fractional order proportional integral derivative (FOPID) speed controller on external loop as well as current controller on internal loop with controlling turn activate and deactivate angles for SRM. The complexity of acquiring the ideal evaluation of proportional, integral and derivative gains for speed and current controller including turn activate and deactivate angles are deemed as a multi-objective optimization issue. Here, the WFS optimize the gain parameters of external speed loop along internal current loop with commutation angles for turn activate and deactivate switches. The WFS optimization processing is used to productive machine learning dataset under the types of SRM parameter. By using the satisfied dataset, the GBDT is predicted and mandatory forecasting is implemented in the entire machine operating stage. The optimized gain parameters based, the fractional order proportional integral derivative controller is tuned exactly. The proposed WFS-GBDT control technique lessens the torque ripple and quick settling time with this proper control, because of its systematic random search capabilities, thereby enhancing the dynamic execution of SRM drive. Finally, the proposed technique is activated in MATLAB/Simulink site, its performance is analyzed with existing techniques, like Base, ALO and WFS. The best, worst, mean, standard deviation for ISEspeed using proposed technique attains 230.5364, 231.5934, 230.952 and 0.05314. The best, worst, mean and standard deviation for torque ripple using proposed technique attains 0.4571, 0.6548, 0.585 and 0.472. The best, worst, mean, standard deviation for ISEcurrent using proposed technique attains 3.1257, 3.9754, 3.5783 and 0.0472.
Read full abstract