Two-dimensional least squares method for improved real-time estimation of switched reluctance machine parameters

A two-dimensional least squares (2DLS) method is considered for modeling the magnetization characteristics of switched reluctance machines (SRM). This is motivated by needs for reducing torque ripple during operation that leads to audible noise and mechanical vibrations. By creating an accurate model suitable for real-time control, the highly nonlinear ' flux linkage characteristics and be adjusted using existing microcontroller architectures. This research examines a commercially available SRM in detail. Results from a detailed magnetic finite element analysis are used to develop and validate the 2DLS method for estimating the magnetic flux levels in the SRM. From the results it has been concluded that a trade-off in computational complexity versus memory requirements always exists. Increased accuracy could be achieved using least squares approximation theory by increasing the polynomial order, but this comes at the penalty of increased computational expense for an embedded microcontroller.