Since electric drives for propulsion can eliminate the need for shafts and gearing while increasing vehicle stealth and power system flexibility, there has been a large number of advances on permanent magnet synchronous motors (PMSMs) for vehicle propulsion in power semiconductors, magnetic materials, and energy storage systems (Liu et al., 2018; Nguyen et al., 2018; Cui et al., 2020; Liu et al., 2021; Zhang et al., 2021; Fu et al., 2022). Novel propulsion configurations are pushing the need for the PMSM to operate at the location where torque is required, which in many propulsion systems is a harsh environment due to severe moisture, humidity, vibration, or temperature. In order to achieve accurate, fast, and small overshoot control of PMSM, the key part is to determine the position information and speed of the rotor magnetic pole. Generally speaking, mechanical sensors are installed to collect the specific parameter information of the signal through the photoelectric encoder and other sensors (Bourogaoui et al., 2017; Cao et al., 2020; Xu et al., 2021; Xiong et al., 2022). However, these sensors require the additional mounting space, increase the cost, and reduce the system reliability in environmental constraint. Hence, the sensorless methods have been developed to simplify the structure and reduce the manufacturing cost of the motor.

Currently, a lot of research has been done on the speed sensorless control of PMSM. In a sensorless permanent magnet synchronous motor, we extract the parameters related to speed and position from the stator voltage and current, which are easy to measure at the stator edge, to replace the mechanical sensor, and to realize the motor closed-loop control (Xiong et al., 2016a; Xiong et al., 2016b; Yi et al., 2020). At present, the sensorless control technology of a permanent magnet synchronous motor is mainly divided into three categories: high-frequency signal injection method (Li and Wang, 2018; Xiong et al., 2020; Yu et al., 2021), motor model method (Chen and Pi, 2016; Zhong and Lin, 2017; Luo et al., 2019; Sun et al., 2019; He et al., 2020; Rongyn et al., 2020), and intelligent algorithm (Urbanski and Janiszewski, 2019; Lu et al., 2020; Ma et al., 2020). Since high-frequency signal injection methods may affect the current signal and many artificial intelligence methods are still in the stage of simulation and theoretical exploration, motor model-based methods are widely used and they have received considerable research attention in the past few years.

The sliding mode observer (SMO) method is one of the algorithms based on the motor model. Compared with other methods, the SMO method is not sensitive to system parameter changes and external disturbances, so it has attracted extensive attention of experts and scholars, see Jin et al. (2014), Lin et al. (2014), Li et al. (2015), Zhang et al. (2016), Zhang et al. (2019), and Hu et al. (2020) and references therein. Due to the inertia of traditional SMO, the equivalent back EMF value observed by SMO contains high-frequency components, and the estimated value has the problem of chattering, which affects the application of traditional SMO in high-precision occasions, and the angle compensation after low-pass filtering also affects the observation accuracy in practical applications. Aiming at the chattering problem in the estimation of traditional SMO, a rotor position identification method of PLL is proposed based on the effective flux model (Zheng et al., 2015; Zhan et al., 2020; Zhao et al., 2020). The comparison link between the actual rotor position and the estimated rotor position of this method is equivalent to the phase detector in the phase-locked loop model. The one-time integration from the speed to the rotor position is equivalent to the voltage-controlled oscillation in the phase-locked loop model. The speed of the motor is adjusted through the error between the system input signal and the output signal.

This article focuses on eliminating the chattering phenomenon of traditional SMO in estimating rotor position and speed. Distinguished from traditional SMO, a sensorless control of the permanent magnet synchronous motor based on SMO with PLL has been designed to improve the accuracy of rotor speed and position estimation. Compared with the literature (Li and Wang, 2018; Yi et al., 2020; Yu et al., 2021), the main contributions of this article can be outlined as follows:

(1) The system and measurement have been modeled to formulate the state estimation problem for PMSM. The research of this problem has great significance for establishing an accurate PMSM mathematical model and selecting which algorithm to estimate the speed and position of the motor rotor in the next step.

The rest of this article is organized as follows: in Section 2, we present the mathematical model of PMSM. In Section 3, we analyze the principle of the SMO with the PLL control scheme and present main results. A simulation result is used to verify the proposed method in Section 4. Finally, Section 5 concludes this article.

PMSM is a multivariable, strong-coupling, and nonlinear system. It needs to be simplified in mathematical modeling. Now, the following idealized assumptions are made:

(1) Idealization of the air gap magnetic field: the magnetic field generated by a permanent magnet and three-phase winding is sinusoidal in the air gap, regardless of the end effect and the influence of primary cogging on the magnetic field.

Through the above assumptions, in order to simplify the mathematical model of PMSM and realize sensorless closed-loop control, correlation transformation must be carried out. Before transformation, the coordinate systems used in the PMSM control principle include the ABC three-phase coordinate system, the αβ two-phase stationary coordinate system, and the dq two-phase rotating coordinate system (Kawabata et al., 2014). The specific positional relationship of the three-coordinate systems is shown in Figure 1. θ is the electrical angle between the rotor shaft (d-axis) and the stator shaft (a-axis), and ω is the synchronous speed.

The voltage equation of PMSM in the αβ axis rotating coordinate system is as follows: u α = R s i α + d ψ α d t ; u β = R s i β + d ψ β d t . (1) where u _α and u _β are the voltages of the α and β axes, respectively; i _α and i _β are the currents of the α and β axes, respectively; ψ _α and ψ _β are the fluxes of the α and β axes, respectively; and R _s is the stator resistance.

The flux linkage equation is ψ α = L i α + ψ f ⁡ cos ⁡ θ ; ψ β = L i β + ψ f ⁡ sin ⁡ θ . (2) where L _α and L _β are the inductances of the α and β axes, respectively, in the surface-mounted PMSM, L _α = L _β = L. ψ _f is the rotor flux, and θ is the rotor position.

Combining Eqs 1, 2, we can obtain that u α = R s i α + d ψ α d t = R s i α + L d i α d t + ψ f d ⁡ cos ⁡ θ d t = R s i α + L d i α d t − ω ψ f ⁡ sin ⁡ θ ; u β = R s i β + d ψ β d t = R s i β + L d i β d t + ψ f d ⁡ sin ⁡ θ d t = R s i β + L d i β d t + ω ψ f ⁡ cos ⁡ θ . (3) where ω is the rotor angular speed, d θ d t = ω .

Arranging Formula 3, we obtain d i α d t = 1 L u α + 1 L ω ψ f ⁡ sin ⁡ θ − R s L i α ; d i β d t = 1 L u β − 1 L ω ψ f ⁡ cos ⁡ θ − R s L i β . (4) when e α = − ψ f ω ⁡ sin ⁡ θ ; e β = ψ f ω ⁡ cos ⁡ θ . (5)

The PMSM mathematical model of Eq. 4 can be written as d i α d t = u α − e α L − R s L i α ; d i β d t = u β − e β L − R s L i β . (6) where e _α and e _β are the back electromotive forces (EMFs) of the α and β axes, respectively. It can be seen from Eq. 5 that the back EMFs e _α and e _β contain all the information of rotor position. Therefore, only by accurately obtaining the back EMF, the speed and position information of the motor be calculated.

Through the derivation of the mathematical model, we can see that the back EMF of PMSM has the sinusoidal waveform and cosine waveform, which contains the rotation angle, speed, and flux linkage information of the motor. It can be seen from Eq. 5 that the phase of the back EMF is related to the rotation angle of the motor, and the amplitude of the back EMF is related to the speed of the motor. The position and speed of the rotor can be obtained as θ = − arctan e α e β ; ω = e α 2 + e β 2 ψ f . (7)

In this article, the sliding mode observer (SMO) is used to estimate the back EMF of the motor. The basic principle is to estimate the back EMF e _α , e _β of the PMSM by using the SMO, then filter the high-frequency harmonics through the low-pass filter to obtain e ̂ α , e ̂ β , and then estimate the rotor speed ω and position θ of the motor by using the phase-locked loop (PLL) technology.

As shown in Figure 6, the system structure block diagram of PMSM sensorless control based on SMO and PLL is constructed.

The system adopts a double-loop structure; the inner loop is the current loop and the outer loop is the speed loop. The closed-loop control is completed by using rotor field oriented vector control with i _d = 0, speed loop PI control, current loop PI control, and the rotor position estimation method based on SMO with PLL. Similar to the traditional SMO, the voltage in the αβ coordinate system is controlled by the SMO and acts with the current, and after the back EMF signal passes through the switching control function and low-pass filter, the rotor position and speed information are obtained by phase-locked loop and phase compensation. The feedback link is used to return the back EMF signal output by the control function to the SMO to realize the closed-loop structure as a whole. This method can realize a stable control system.

In order to verify the feasibility of the above control system, the simulation system model is built in the MATLAB/Simulink environment, and the simulation analysis is carried out. The following are the detailed parameter settings of the simulation. Figure 7 shows the comparison of rotor speed waveforms corresponding to the two methods. It can be seen that the estimated rotor speed corresponding to the two methods is basically consistent with the actual rotor speed. It can be seen from Figure 7A that the given target speed of the motor is 1,000 rad/min. The traditional SMO control can track the expected speed. The estimated speed is almost equal to the actual speed when it is stable, but there will be large chattering in the speed estimation at the moment of low speed and speed change, and there will be large deviation in the speed estimation. It can be seen from Figure 7B that although there is a certain overshoot during startup, there is no chattering and tends to be stable soon. The corresponding speed waveform curve is much smoother. The estimated speed can quickly follow the set speed and converge to the actual speed, which shows that the system has good dynamic and static performance.

Permanent magnet motor vector control is adopted. The simulation conditions are as follows: DC side voltage U _dc = 311 V, the pulse width modulation (PWM) switching frequency is 10 kHz, the given speed is 1,000 rad/min; start with no load first, and mainly observe the estimated differences of speed and rotor position in SMO and SMO with PLL.

Using the traditional SMO and the SMO with PLL method proposed in this article, the speed and position simulation diagrams of the traditional SMO control and SMO with PLL are given, respectively.

Figure 8 is a waveform of the deviation between the estimated speed and the actual speed. It can be seen from Figure 8A that the maximum speed estimation error of the traditional synovial observer is 16 rad/min and finally stabilized at 10 rad/min. That is, the steady-state error is 10 rad/min. It can be seen from Figure 8B that the SPM and PLL speed fluctuation has been greatly improved, which not only ensures the real-time tracking of rotor speed but also optimizes the steady-state error of speed. The maximum speed error is only 12 rad/min after stabilization, and the error is basically 0. The steady-state error value is further weakened, which shows that this method has a good filtering effect, strong tracking performance, and small steady-state velocity fluctuation.

Figure 9 is the motor rotor position curve. It can be seen that the estimated rotor position can maintain good consistency with the actual rotor position, which shows that the observer has a good estimation effect. As can be seen from Figure 9A, due to the high-frequency chattering component, the estimated position of the rotor lags behind the actual position, and there is a fixed error, which cannot be corrected and improved and will affect the dynamic response performance of the system. Figure 9B shows that after the PLL is introduced, the output frequency is controlled by the voltage difference between the phase of the input signal and the phase of the output signal to reach the desired frequency value, maintain a fixed phase error, and be in the phase locking state. There is neither phase lag nor amplitude attenuation. The estimated angle is basically the same as the actual angle, which effectively weakens the high-frequency noise and harmonics in the back EMF and makes the output stable, and the control effect more meets the performance control requirements of PMSM.

Figure 10 is the motor rotor position error curve. The maximum position error obtained by the traditional SMO shown in Figure 10A is 6.1 rad/min, while after the introduction of PLL, the maximum error between the actual position and the estimated position of the motor is .75 rad/min. It shows that the accuracy based on SMO with PLL is improved.

This article has studied a chattering suppression method based on the combination of SMO and PLL technologies, and the chattering problem of traditional SMO in PMSM sensorless control has been solved. In this method, the SMO has been used to estimate the back EMF, and the phase-locked loop has been employed to estimate the speed and position of PMSM. The design principle of PLL has been analyzed theoretically, and the derivation and verification have been given. Through the simulation of the design scheme, the results show that the SMO with PLL improves the chattering problem of the traditional SMO, has high estimation accuracy of motor speed and rotor position, has a small amount of calculation, and has strong robustness to external interference.