Conventional model-free predictive current control (MFPCC) using an extended state observer (ESO) based on ultra-local model significantly decreases the sensitivity to partial motor parameters. However, the performance and stability of ESO will be degraded in the face of amplified high-frequency (HF) perturbation caused by inductance parameter mismatch. To solve the above problem, an enhanced ESO-based MFPCC method is proposed in this article. On the basis of constructing the discrete equation of ESO, an accurate and fast calculation of inductance parameter is proposed, which is real-time convergence to the actual value by deadbeat control principle, and fed back to the lumped disturbance and current observation calculation. It effectively eliminates the effects of HF perturbation on ESO. Simultaneously, the inductance and compensated lumped disturbance obtained from ESO are provided to the MFPCC command voltage calculation module, further improving the parameter robustness of MFPCC. The stability of the proposed ESO is ensured by the design of observer gain. Experiments are carried out on 1-kW surfaced permanent magnet synchronous machines test bench. Compared with the conventional ESO-based MFPCC method, the proposed method is superior in dynamic response, steady-state performance, and parameter robustness.

This article proposes a global delay compensation method to improve the sensorless control performance of square-wave voltage injection (SWVI) over low switching frequency applications. The cost function with delay information is preliminarily constructed by simple calculation of high-frequency (HF) αβ -axes current and estimated position; second, the cost function is minimized by the gradient descent method, so as to estimate the delay time and corresponding compensation angle. The compensation angle that estimated delay time multiplies rotor speed is used for reducing the lag angle in the estimated position, thus improves the current control performance. At the same time, the inverse of the cosine signal is proposed to substitute the conventional differential current demodulation of SWVI. In addition, the compensation angle is used to offset the phase shift in the cosine signal, thus suppressing the amplitude error of HF response current caused by time delay. Hence, the SWVI method with the proposed delay compensation has good position estimation performance. Finally, a 1.5 kW interior permanent magnet synchronous motor test bench is adopted to demonstrate the effectiveness of the proposed method

Fast and reliable initial rotor position detection is essential for re-starting sensorless permanent magnet synchronous motors (PMSMs) in free-running condition. In this paper, a fast initial rotor position estimation method for low speed free-running motor is proposed, which utilizes a combined sinusoidal current and square-wave voltage injection method. The sinusoidal current is imposed into the estimated d-axis to magnify magnetic saturation effect. The amplitudes of d-axis current caused by injected square-wave voltage are then accumulated. The large difference of the two integrated signals for positive and negative d-axes currents can be reliably used to identify the rotor polarity. Meanwhile, in low-speed free running stage, the change of saturation degrees introduced by the injected sinusoidal signal does not affect the position estimation accuracy. Moreover, even if the sinusoidal current signal is injected in the incorrect d-axis, the resultant torque is small and unexpected rotation of the rotor is prevented. Its influence on the free-running motor is negligible, due to the combined injection with continuously online updated estimated rotor position by high frequency (HF) square-wave voltage injection during the polarity identification process. Finally, the effectiveness of the proposed method is investigated on a 1.5 kW interior PMSM (IPMSM) test platform.

The conventional model-free deadbeat predictive current control method with extended state observer (MFDPCC-ESO) has been a promising method because of its great dynamic performance and significant parameter robustness. However, the conventional ESO still depends on inductance for designing controller gain, which inevitably reduces the parameter robustness of the method. In order to address this issue, an MFDPCC method with adaptive gain ESO (AGESO) is proposed in this paper. First, the impact of the conventional ESO input gain on control performance was analyzed, and it points out that controller gain mismatch will increase current ripple and even cause system instability. Then, the proposed method is based on a modified ultralocal model, in which the disturbance only includes the inductance term. Using disturbance as the objective function and combining finite-time gradient method (FGD) to estimate inductance parameters adaptively, the effect of controller gain mismatch is suppressed. Finally, experimental research was conducted on a 1kW surface permanent magnet synchronous motor (SPMSM) test bench to verify the steady-state performance and parameter robustness of the method.

This article proposes an optimized phase-locked loop (OPLL) based on Levenberg–Marquardt (LM) to improve the estimation accuracy of the sliding mode observer-based position estimator. The position estimation suffers from dc bias caused by digital control delay, and harmonics caused by inverter nonlinearity and flux spatial harmonics. The OPLL is constructed in two parts to overcome the above negative effects. The delay phase is preliminary calculated based on linear prediction, so as to compensate for the lag phase in the estimated back electromotive force (back-EMF); secondly, the objective function is built based on the reference and feedback position errors, the former of which contains an unknown harmonic. The objective function can be minimized by the LM to solve for the harmonic amplitude, and then the obtained harmonic is used to offset the harmonic components in the estimated position. Wherein, the required partial derivatives of the LM method can be deduced based on the transfer function of PLL and the delay-compensation back-EMF to achieve a stable update of harmonic amplitude. Finally, the corresponding comparative experiments in steady state and dynamic conditions proved the superiority of the proposed method applied to permanent magnet synchronous motors.

Accurate rotor position is significant for the back electromotive force (EMF)-based sensorless interior permanent-magnet synchronous motor (IPMSM) control. However, the ± (6 h±1)th harmonics will appear obviously in the estimated back-EMF due to the effect of the inverter nonlinearity and flux spatial harmonics. These harmonics will subsequently result in the ± (6 h)th harmonic pulsation in the estimated rotor position. In order to deal with this issue, a new complex-coefficient synchronous frequency filter (CCSFF)-based sliding mode observer combined with a quadrature phase-locked loop (PLL) is proposed to mitigate the back-EMF voltage distortion. Therefore, the performance of the position estimation is remarkably improved. The proposed CCSFF possesses both bandpass-filtering and frequency-adaptability characteristics. It can pass the fundamental component without magnitude attenuation and phase delay in different frequency scenarios. Moreover, the linearized model of the proposed CCSFF-PLL-based position estimation system is established and a systematic parameter design is presented to obtain promising dynamic performance. The effectiveness and feasibility of the proposed method are confirmed by experiments on a 1.5-kW IPMSM test platform.