2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

Sensorless Control for PMSM with Novel Back

EMF Observer Based on Quasi-PR Controller
Lin Ji Yongxiang Xu Jibin Zou Minghui Wang
Department of Electrical Department of Electrical Department of Electrical Department of Electrical
Engineering Engineering Engineering Engineering
Harbin Institute of Harbin Institute of Harbin Institute of Harbin Institute of
Technology Technology Technology Technology
Harbin, China Harbin, China Harbin, China Harbin, China
[email protected] [email protected] [email protected] [email protected]

Abstract—Sensorless Control of Permanent Magnet sensorless control in medium and high speed range, when the
Synchronous Motor (PMSM) is a hot research topic nowadays. back EMF is large enough to be estimated. For sensorless
The main middle and high speed sensorless control methods are control technology based on motor model, MRAS may have
based on the back EMF model of the motor. Therefore, the back difficulties in model parameter adaptation. For example, on-
EMF observer plays an important role in the sensorless control line identification of motor parameters can improve the
system. The contribution of this paper is to propose a novel back performance of sensorless control based on MRAS[12].
EMF observer based on quasi-PR controller. The proposed Sensorless technology based on EKF works well in noise
quasi-PR observer has a simple structure similar to the sliding environment. However, the requirement of hardware for
mode observer and eliminates the chattering phenomenon at the
complex computing is very high [7]. The main idea of sliding
same time. In this paper, a quasi-PR observer is proposed to
estimate the back electromotive forces (BEMFs), and a phase
mode control is to select an ideal synovial surface, and then
locked loop (PLL) is utilized to estimate the electrical speed and make the state variables move along the synovial surface by
rotor position.The effectiveness is demonstrated by simulation dynamically changing the system structure. Sliding mode
and experimental results. control has discontinuity similar to switching characteristics.
In fact, the system is moving along the synovial surface at a
Keywords—Permanent Magnet Synchronous Motor (PMSM), small amplitude and a high frequency[45]. In sensorless control
sensorless, observer, quasi-PR controller, PLL of permanent magnet synchronous motor (PMSM), sliding
mode controller is used to observe the rotor position, so it is
I. INTRODUCTION called sliding mode observer. The error between the observed
In recent years, permanent magnet synchronous motor current and the actual current in the stationary two-phase
(PMSM) has been widely used because of its simple structure, coordinate system is taken as the sliding surface. The back
small size, high power density, high reliability, good speed EMF is reconstructed from the current error, and then the
control performance and fast dynamic response. In traditional position information contained in the back EMF is extracted.
permanent magnet synchronous motor (PMSM) control The sliding mode observer is insensitive to external
system, position sensors are used to obtain rotor position and disturbances and has good robustness. However, the inherent
speed information. However, the installation of position discreteness of the sliding mode algorithm makes it difficult
sensor increases the cost and complexity of the system, to solve the chattering problem of the system fundamentally.
reduces the mechanical strength, reduces the anti-interference
II. QUASI-PR BACK EMF OBSERVER
ability of the motor control system to electromagnetic noise,
mechanical vibration and temperature, and reduces the The mathematical model of surface mounted permanent
reliability of the motor control system. To solve these magnet synchronous motor in static two-phase coordinate
problems, sensorless control has become a research hotspot in system is as follows.
the field of permanent magnet synchronous motor control.
In order to improve reliability and to reduce cost, u   Rs  PLs 0   i    sin( r ) 
sensorless control of permanent magnet synchronous motor u     i   r f    
(PMSM) has become a research hotspot in the field of electric    0 Rs  PLs      cos( r ) 
drive since 1980s. The sensorless control technology of
permanent magnet synchronous motor (PMSM) can be where uαβ is motor voltage; iαβ is motor current; eαβ is
divided into two categories: one is high frequency injection BEMF; Ls and Rs are motor inductance and resistance,
method based on saliency tracking of PMSM, the other is back respectively.
EMF position extraction method based on mathematical The state equation with current as state variable is
model of PMSM. Research on sensorless control technology organized as follows.
of permanent magnet synchronous motor mainly focuses on
three directions: rotor initial position detection, control in low
and zero speed range and control in medium and high speed  di Rs 1 1
range.  dt   L i  L u  L e
 s s s
   
Back EMF observer method based on PMSM fundamental di
    Rs i  1 u  1 e
frequency mathematical model, including model reference  dt Ls

Ls

Ls


adaptive system (MRAS)[6], extended Kalman filter (EKF)[7]

and sliding mode observer (SMO)[8-11], is suitable for
Among them, the back EMF equation is:

978-1-7281-3398-0/19/$31.00 ©2019 IEEE

Authorized licensed use limited to: Karunya Institute of Technology and Sciences. Downloaded on February 13,2025 at 16:12:10 UTC from IEEE Xplore. Restrictions apply.
978-1-7281-3398-0/19/$31.00 ©2019 IEEE
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

 e   f  r sin( r ) is the integral coefficient.

Ki
    PI controller has a zero in the left half plane and a pole at
 e   f  r cos( r ) the origin. The gain of PI controller at fundamental wave angle
frequency  0 is as follows:
Obviously, back EMFs contain the position information of
the rotor. We can firstly get the back EMFs of the motor, and
2
then extract the position information contained in it. Formula K 
1 shows that since the voltage and current of the motor are  API ( 0 )  K p2   i   
known values, then the back EMF of the motor can be  0 
obtained directly by subtraction calculation. However, it is
difficult to achieve effective control through this method in Apply PI controller to the system in Fig. 1.. When Ki and
practice for the calculating formula contains a differential Kp are finite, the gain of PI controller at fundamental angle
operation of the current. The differential operation may frequency is finite. Bring this gain into equation(4).
amplify the detection error of the current, resulting in a violent
oscillation of the calculated back EMF of the motor, which Gc ( s )
 iˆ 、   i 、   
contains a great quantity of high-frequency components， Ls  s  R s  G c ( s )
from which it is impossible to extract effective position
information of motor rotor.
Gc ( s)
In order to avoid amplifying detection error, the  1  
Ls  s  Rs  Gc ( s)
differential of motor current is taken as the main variable, the
differential operation is replaced by integral operation, and the
back EMF is expressed as a function of current estimation The gain of the first part of the formula is less than 1,
error. The system structure diagram of the back EMF observer which results in a steady-state error in tracking the current at
is shown in figure. 1. the fundamental angle frequency of the system.

𝑢𝛼 、𝛽 1
𝑖𝛼 、𝛽  iˆ、  u、  
+
1 1 𝑖̇𝛼 、𝛽 + Ls  s  Rs  Gc ( s )
− 𝐿𝑠 𝑠 −
− 1
Rs
𝑖̇𝛼 、𝛽
 0 
Ls  s  Rs  Gc ( s )
𝑒̂ 𝛼 、𝛽 𝑖̇𝛼 、𝛽
Gc(s)
The gain of the second part of the formula is greater than
Fig. 1. Block diagram of the proposed quasi-PR observer
0, which means that the disturbance of motor voltage to
current estimation cannot be neglected.
where quantities with superscripts represents observed Therefore, the PI controller cannot eliminate the steady-
physical quantity. Gc(s) represents the transfer function of the state error, suppress the low-frequency harmonics, and has
controller. We can obtain the closed-loop transfer function of poor anti-interference ability to the voltage. PI control strategy
the observer from Fig. 1.. is not suitable for this system.

Gc  s  From the above analysis, it can be concluded that

 1
 i 、 = i、 + u、   increasing the gain of the regulator at the fundamental
Ls  s  Rs  Gc  s  Ls  s  Rs  Gc  s  frequency can make the steady-state transfer gain closer to 1
and the output gain against motor voltage closer to 0.
The transfer function of the system is divided into two Especially, when Ac(ω0)=∞, we can obtain that:
parts, The first part is the transfer function against actual
current, which represents the steady-state performance of the Gc ( s)
system; while the second part is the transfer function against  1  
motor voltage, which represents the anti-interference ability Ls  s  Rs  Gc ( s )
of the system. Different controllers will affect the performance
of the system in different ways. We will choose the 1
appropriate controller by comparing the performances  0 
Ls  s  R s  G c ( s )
respectively.
PI controller, PR controller and quasi PR controller are At this time, the steady-state error of current estimation at
commonly used controllers in practice of engineering. Among fundamental angle frequency is 0, and it is completely free
them, PI controller is the most widely used one. The transfer from the interference of motor voltage. Therefore, one of the
function of PI controller is as follows: requirements for the controller is to have a large enough gain
at the fundamental angle frequency.
Ki The transfer function of PR controller is：
 GPI ( s)  K p   
s

where Kp is the proportion coefficient of the controller, while

Authorized licensed use limited to: Karunya Institute of Technology and Sciences. Downloaded on February 13,2025 at 16:12:10 UTC from IEEE Xplore. Restrictions apply.
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

2Ki s which can meet the requirements of the system for steady-state
 Gc ( s)  GPR ( s)  K p    performance and anti-jamming performance. At the same time,
s  0 2
2
the gain far from the fundamental frequency is small, which
can suppress the high-frequency and low-frequency clutter,
where K p is the proportion coefficient of the controller, and meet the system design requirements.
while K i is the resonance coefficient, and  0 is the At the same time, the quasi-PR controller has a relatively
fundamental angle frequency. wide bandwidth and maintains a high gain near the
fundamental frequency. When the fundamental frequency of
The gain of PR controller at fundamental wave angle the controlled variable is slightly offset, it has little effect on
frequency is calculated as follows: the control effect.

2
III. PLL
 2K   After obtaining the back EMF of the motor by using the
 APR (0 )  K p 2
 2 i 0 2   
 0  0  quasi PR observer, the speed and position information of the
motor can be extracted from it.[1]
Obviously, the PR controller satisfies the requirement of Park transformation of the motor back EMF under the axes
Ac(ω0)=∞, which means the steady-state error of current of α and β is carried out to obtain the motor back EMF under
estimation at fundamental angle frequency is 0, and it is the axes of d and q:
completely free from the interference of motor voltage.
Draw the Bode diagram of PR controller: ed   cos r  sin r   e 
      
eq   sin r  cos r   e 

Bring formula (3) to formula (16), we can obtain that:

ed   0 
 e       
 q  f r
Fig. 2. Bode diagram of PR controller
The d-axis component of estimated back EMF can be
obtained by PARK transformation of the estimated back EMF
As shown in Fig. 2. PR controller has an extremely high
under the axes of α and β.
gain at the fundamental angle frequency, but on both sides of
the fundamental angle frequency, the gain decreases sharply.
On the one hand, it is beneficial for the system to filter low-
frequency and high-frequency clutter. On the other hand,
    
eˆd  cos ˆr eˆ  sin ˆr eˆ  
when the fundamental frequency of the input waveform
fluctuates and shifts, the control performance of PR controller When the estimated back EMF converges to the actual
will deteriorate due to the narrow bandwidth. The steady-state value：
error and voltage interference cannot be ignored. In order to
solve this problem, we take quasi-PR controller into  eˆd  cos ˆr  e  sin ˆr  e   f r sin(ˆr  r )   f r (ˆr   r ) 
consideration.
The transfer function of the quasi-PR controller is: The above formula shows that the observation angle tends
to the actual angle when the back EMF of d-axis observation
2 K ic s approaches zero. In order to obtain accurate rotor angle
 GQPR ( s )  K p    estimation, PI controller is used to stabilize the back EMF of
s  2c s  0 2
2
d-axis observation at 0.[2] The structure of the angle extraction
system is shown in Fig. 4..
The Bode diagram of quasi-PR controller is shown as
follows: 𝜓𝑓 𝜔𝑟 𝜔𝑟

1 𝜃𝑟
𝜃𝑟 +
𝜓𝑓 𝜔𝑟 ÷ PI
𝑠
−
𝜃𝑟

Fig. 4. The structure of the angle extraction system

The output of PI controller has the characteristic of

constant value. It can be used as the observation velocity.
Fig. 3. Bode diagram of quasi-PR controller
Then the observation angle can be obtained by integrating the
From the graph, it can be seen that the quasi PR controller observation velocity, and the observation angle can be used as
has a higher gain at the fundamental frequency and its vicinity, the input of Park transform angle to form a closed loop. This

Authorized licensed use limited to: Karunya Institute of Technology and Sciences. Downloaded on February 13,2025 at 16:12:10 UTC from IEEE Xplore. Restrictions apply.
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

closed-loop angle observation method avoids the enlargement 1.0

speed estimation error(rad/s)

of the division error in the open-loop method, improves the
observation accuracy and greatly reduces the position
observation error. 0.5

IV. SIMULATION RESULTS

0.0
Under the condition of traditional double closed-loop
control of speed and current, the position observer based on
quasi-PR Regulator designed in the previous section is used to -0.5
observe the rotor position, and the observed results are 0.45 0.50 0.55 0.60 0.65 0.70
time(s)
compared with the actual position signal. In order to prove that
Fig. 8. Speed estimation error
the position observer based on quasi PR controller can
accurately track the rotor position under various speed and 0.022
load conditions, the steady-state performance is compared

position estimation error(rad)

under different speed and load conditions.
0.021
The following waveforms can be obtained when the speed
is 1000r/min and the load is 2Nm.
4 0.020
Actual current
Estimate current
Current under α axis(A)

2
0.019
0.45 0.50 0.55 0.60 0.65 0.70
0 time(s)

-2 Fig. 9. Position estimation error

-4 The control system has good steady-state performance.

0.35 0.40 0.45 0.50
time(s) 110
estimated speed
Fig. 5. Actual current and estimated current actual speed

100 105
speed(rad/s)

α axis
β axis
Estimated Back EMF(V)

50
100
0

-50 95
0.55 0.60 0.65 0.70 0.75 0.80
time(s)
-100
0.35 0.40 0.45 0.50
time(s) Fig. 10. Speed response curve when load is increased
Fig. 6. Estimated back EMF
135
estimated speed
As can be seen from the figure, the estimated current 130 actual speed
follows the actual current, and the estimated back EMF does 125
speed(rad/s)

not contain high frequency components. It has a good

120
sinusoidal degree and does not need filtering,which will result
in phase lag. 115
110
At 0.5s, the observed position and rotational speed
information are used to replace the original actual position and 105
rotational speed information. The following waveforms are 100
obtained when the speed is 1000r/min and the load is 2Nm. 0.55 0.60 0.65 0.70 0.75 0.80
time(s)
105.00
Fig. 11. Speed response curve when speed is increased

104.75 The simulation results show that the new control algorithm
speed(rad/s)

has good steady-state performance and dynamic performance.

104.50 V. CONCLUSION
estimated speed
A new quasi PR back EMF observer is proposed in this
actual speed paper. In theory, this new type of back-EMF observer can
104.25
0.45 0.50 0.55 0.60 0.65 0.70 track the back-EMF without steady-state static error. At the
time(s) same time, it can keep good observation effect when the
Fig. 7. Actual speed and estimated speed observed electric frequency is slightly offset. Combined with
orthogonal phase-locked loop, a novel back-EMF observer is

Authorized licensed use limited to: Karunya Institute of Technology and Sciences. Downloaded on February 13,2025 at 16:12:10 UTC from IEEE Xplore. Restrictions apply.
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

applied to sensorless control of permanent magnet International Conference on Electrical Machines and Systems
synchronous motor. The simulation results show that the (ICEMS), Jeju, pp. 1644-1648 , 2018.
system has good steady-state dynamic performance. [2] O. Saadaoui, A. Khlaief, M. Abassi, A. Chaari and M. Boussak,
"Position sensorless vector control of PMSM drives based on SMO,"
2015 16th International Conference on Sciences and Techniques of
REFERENCES Automatic Control and Computer Engineering (STA), Monastir, pp.
[1] M. Wang, Y. Xu and J. Zou, "Sensorless Control for PMSM Connected 545-550 , 2015.
with LC Filter Based on Extended State Observer," 2018 21st

Authorized licensed use limited to: Karunya Institute of Technology and Sciences. Downloaded on February 13,2025 at 16:12:10 UTC from IEEE Xplore. Restrictions apply.