Sensorless control of permanent magnet synchronous motor for exhaust energy recovery of internal combustion engine: a comparison between Kalman filter and MRAS observer

This paper deals with the sensorless control of a Permanent Magnet Synchronous Motor (PMSM) to be employed on a turbo-generator to recover the exhaust gas energy of an Internal Combustion Engine (ICE). The main problem with this application is the high rotational speed required by the exhaust gas turbo-generator. Indeed, this implies the difficulty of installing an encoder for speed measurement and control. For this reason, sensorless control is highly recommendable, but, on the other hand, the main sensorless techniques could fail in this peculiar situation because of the high computational effort required compared with the other sensorless applications. Moreover, a good knowledge of the model parameters is necessary. Therefore, the main goal of this paper is to compare two main sensorless techniques for rotational speed measurement, both in the deterministic framework (Model Reference Adaptive System (MRAS) observer) and stochastic framework (Extended Kalman Filter (EKF)), to verify their effectiveness, their computational load, their sensitivity against parameter variation and noisy measurements, and, finally, to present the best solution for the exhaust gas energy recovery from the internal combustion engine.