Non-Linear Inductance Modeling of Switched Reluctance Machine Using Multivariate Non-Linear Regression Technique and Adaptive Neuro Fuzzy Inference System

This paper presents two different methods of real time applicable modeling techniques for the Non-linear inductance model of a Switched reluctance Machine (SRM). These methods are based on Multivariate nonlinear regression technique (MVNLRT) and Adaptive Neuro fuzzy inference system (ANFIS). These techniques are applied for the nonlinear inductance calculation by using the magnetization characteristics of SRM. MVNLRT is an excellent solution for nonlinear modeling and its real time implementations. Similarly ANFIS also has a strong nonlinear approximation ability which could be used for nonlinear modeling and its real time implementations. In this paper, the best features of MVNLRT and ANFIS are utilized to develop the computationally efficient inductance model for SRM. Mathematical models for the phase Inductance L(I,θ) using MVNLRT and ANFIS have been successfully arrived, tested and presented for various values of phase currents(Iph) and rotor positions(θ) of a non linear SRM. It is observed that MVNLRT and ANFIS are highly suitable for Inductance L(I,θ) modeling of SRM which is found to be in good agreement with the training data used for modeling.