Position Sensorless Drive With Online Parameters Estimation for Magnetic Suspension Centrifugal Compressor

Abstract

With the advantages of a high power-to-weight ratio and oil-free bearings, a magnetically Suspension centrifugal compressor (MSCC) equipped with active magnetic bearings (AMB) operates more efficiently than traditional mechanical compressors over a wide speed range by precisely controlling a high-speed permanent-magnet synchronous motor (PMSM). However, owing to the harsh working environment, the motor parameters, including the resistor, inductance and flux linkage, vary significantly. This degrades the rotor position observer performance, which is crucial for controlling a high-speed PMSM. Hence, this article proposes an improved sensorless control method combined with novel motor parameter identification. First, an improved position sliding mode observer using a novel super-twisting algorithm (STA) is designed to achieve higher precision and better transient performance. Second, the influence of parameter mismatch is analyzed. To overcome the problems of rank deficiency and system matrix uncertainties, a novel identification strategy based on recursive total least squares excitatory and inhibitory learning (RTLS EXIN) is developed to identify inductance and resistance in two-time scales. Meanwhile, the back electromotive force (EMF) in static coordinates is employed to estimate the flux linkage. Third, the results are adopted to update the control parameters immediately. Finally, the proposed method is tested on an MSCC prototype experimental platform, and the results verify its validity and feasibility.