Research Article
Artificial Neural Network-Based Control of Switched Reluctance Motor for Torque Ripple Reduction
Table 21
Comparison of the proposed technique with existing techniques.
| Parameters | Proposed technique | PI controller | Simplified model of SRM |
| MSE | MSE is very much reduced in ANN | MSE is higher as compared with ANN | MSE is of large value as compared with ANN and PI | Accuracy | Improved accuracy | Less accurate in comparison with ANN | Lack of accuracy | Conversions | Speedily conversions with reduction in torque ripple | Slow conversions as compared with ANN | Slow conversions as compared with ANN | Generalization | Good generalization when retrained improved performance with time | No generalization property has to be returned and no improvements with time | No generalization property | Simulation speed | Time taken for computation is less and fast in operation | Long computing time | Slow | Performance | Better performance for nonlinear characteristics of SRM | Performance is not up to the expected level | Poor performance | Regression | Regression is in the proximity of 1 which validate performance and fast change of parameters in ANN | No regression property, SRM demands a fast change of parameters, and the PI controller has a slow change of parameters | Nil | Complexity | Less complex because no mathematical model included | Complex because mathematical model included | Complex than ANN and PI controller | Stability | Good stability | Less stable than ANN | Less stable |
|
|