I. Introduction
Aiming to control the stator current in surface permanentmagnet synchronous motor (SPMSM) Drives, deadbeat predictive current control (DPCC) has got a wide attention. Based on the discrete mathematical model of a system, DPCC can predict the future current behavior. However, DPCC can be negatively affected by the parameter mismatch of the controlled system [1]-[3]. For example, SPMSM stator inductance decreases as the stator current rises [4]-[5]. Many methods have been developed to reject the inductance disturbance. A sliding mode observer with a novel incremental machine model is proposed to estimate the inductance [6]. An additional signal injection method is proposed to compensate for the disturbance in [7] but it will inevitably deteriorate the control performance. In [8], an adaptive observer with recursive identification algorithm is proposed to estimate the d-axis and q-axis inductances simultaneously. Apart from the inductance disturbance, the value of SPMSM rotor flux linkage and resistance parameters can change with temperature varies. Authors of [9] proposed a full order observer to reject the disturbance. Based on DPCC, A stator current and disturbance observer based on adaptive sliding control is proposed to estimate the actual disturbances caused by SPMSM inductance, resistance and flux linkage parameter mismatch in DPCC [10]. Without utilizing observers, a model reference adaptive control method has been introduced in [11], where an adaptive law derived by the Lyapunov stability is employed to reject the disturbance. Recently, a nonparametric predictive current control was proposed [12]. This method predicts the next instant current based on measured currents without any observers. To reject the torque ripple caused by the undesired voltage vector, the authors of [13] proposed an improved nonparametric predictive current control with an anti-stagnation detection. The basic procedure is that the switching state will be forced to be conducted in an inverter if a certain switching state did not occur within 50 successive control periods. Although the stagnant current is avoided, this detection can have an adverse impact on current performance due to the undesired switching state involving in the system. In order to improve the current and torque ripple performance further, [14] proposed a novel nonparametric predictive current control with a novel current variations update mechanism. However, the three methods in [12]-[14] are only applied in finite-control set model predictive current control instead of DPCC.