PMSM Drive System Efficiency Optimization Using a Modified Gradient Descent Algorithm With Discretized Search Space

This article investigates efficiency optimization in a permanent magnet synchronous machine (PMSM) drive system, including the motor and the power converter. At first, the motor drive efficiency model based on the input and output power is proposed. In this model, the input power is calculated from the dc-link voltage and current measurements, and the output power is modeled and computed from the dq-axis voltages and currents considering temperature rise, magnetic saturation, cross-coupling effect, and inverter nonlinearity. The proposed efficiency model includes both the inverter and motor losses without the need of loss models, so it simplifies the efficiency calculation significantly. Based on this efficiency model, a novel gradient descent algorithm-based approach with a discrete search space and proper constraints is proposed to optimize the PMSM drive system efficiency, ensure a fast searching speed, and reduce the influence from measurement uncertainties. Compared with the existing approaches, the proposed approach is computationally efficient, does not require loss models, and is noninvasive as no signal injection is involved. The proposed approach is evaluated on a laboratory PMSM drive system with extensive experimental tests.