The linear induction machine (LIM) drive system can get direct thrust and linear motions without transmission, which enjoys strong climbing capability, high acceleration or deceleration ratio, and small mechanical losses. The LIM drive systems have been developed and commercialized in over 20 linear metro lines worldwide. However, due to the large air gap, end effects, and high-power, low-switching frequency drive, the LIM drive system in urban rail transit needs better efficiency. Although the existing efficiency optimization control strategies have improved machine efficiency, the parameter robustness and system efficiency still need to be addressed. This paper proposes a robust efficiency optimization strategy for three-level inverter-fed LIM systems under low switching frequency.

Firstly, the primary flux-based LIM loss model considering end effects is built, where the loss is expressed as a convex function of primary flux. Its parameter sensitivity and limitation are analyzed. Furthermore, combined with the gradient descent method, a hybrid optimal primary flux search method is proposed to eliminate the influence of parameter changes on optimal flux selection. Then, the cost function containing multiple objectives, such as primary flux control, switching frequency constraint, and neutral point voltage balance, is derived. A model-free predictive flux control based on the nonlinear-extended state observer is proposed to manipulate optimal flux flexibly under low switching frequency.

Finally, experimental comparisons with the existing methods on a 3 kW LIM confirm that efficiency and parameter robustness can be improved for the drive system under low switching frequency. The system efficiency with the proposed method can be improved by 1.22% and 0.64% compared with the mature control strategy and the existing efficiency optimization strategy under the working conditions of 8 m/s and 200 N.

The following conclusions can be drawn. (1) The proposed method takes the minimum DC-link current as the search objective, which considers the harmonic loss and inverter loss, thus improving the system’s efficiency. (2) Considering multiple objectives, such as the switching frequency constraint and neutral point voltage balance, a model-free predictive flux control with adaptive switching frequency regulation is developed. (3) By combining the hybrid optimal primary flux search method with model-free predictive flux control, the proposed method effectively avoids the influence of parameter changes and modeling errors on optimal flux selection and manipulation. In this way, the parameter robustness of the efficiency optimization control strategy is significantly enhanced.