A Frequency-Adaptive Delay Signal Cancelation Based Filter to Reduce Position Estimation Error for Sensorless IPMSM Drives

In a sliding mode observer-based sensorless drive system, the estimated back electromotive force (BEMF) always contains abundant low-order harmonics caused by nonideal factors, e.g., flux spatial harmonics, inverter nonlinearity, and current measurement error, resulting in large estimated position errors. In previous studies, adaptive notch filters are usually adopted to suppress the BEMF harmonics. However, these filters suffer from enormous computation and increase system complexity, especially facing numerous low-order harmonics. In this article, a new frequency-adaptive notch filter employing delay signal cancelation (DSC) is proposed, which can be tailored to eliminate multiple harmonics with a simple structure. First, Lagrange interpolation is applied to realize fractional-order delay in the frequency-adaptive DSC (FADSC). Then, a frequency-dividing record method is designed to shorten the recording array length for the FADSC at low speed. Next, a switching procedure is introduced to change the frequency-dividing factor smoothly at different speeds. Finally, the transformation matrix in the DSC is changed in the FADSC by adding the rotation direction to make it suitable to forward and reverse rotations of an interior permanent magnet synchronous machine (IPMSM). The accuracy of the estimated rotor position can be improved by FADSC, which is verified by experiments on a prototype IPMSM.