Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Auto-calibration of position offset for PMSM represent the d- and q-axes stator currents and voltages, respectively;

drives with uncertain parameters Te is the output torque; P is the number of pole pairs; and wr is the
electrical rotor speed.
K. Choi, Y. Kim, S.-K. Kim✉ and K.-S. Kim The position offset defined by the difference between the position ur
and its measurement umeas
r can be expressed as
This Letter proposes a novel technique for automatic position offset uoff = ur − umeas
r = mwr + u0 , (4)
calibration for permanent-magnet synchronous machine (PMSM)
drives. Contrary to existing methods, the proposed technique estimates with the delay coefficient m and the initial rotor position u0 . The
and compensates for the position offset without using machine true relationship between the variables in the d–q frame and the biased
parameters or signal injection, which is possible under the condition d–q frame (i.e. d ′ –q′ frame), calculated using umeas
r , can be presented as
of zero-current control at high enough speed where parameter-     
dependent terms are eliminated. To achieve this condition using xd cos uoff sin uoff xd ′
= . (5)
uncertain parameters, speed control with absolute value feedback xq − sin uoff cos uoff xq′
and model-free current control are devised. Additionally, under this
condition, not only the misalignment of the position sensor but also The PMSM is described in the d ′ –q′ frame as follows:
the sensor delay can be considered as the position offset, which
d
improves the accuracy of the calibration. The proposed approach will Ld id′ = −Rs id ′ + wr Lq iq′ + vd ′ + Eex sin uoff , (6)
be verified by experimental results. dt
d
Ld iq′ = −Rs iq′ − wr Lq id ′ + vq′ − Eex cos uoff , (7)
Introduction: It is necessary to obtain accurate rotor position dt
information to enable the field-oriented control of permanent-magnet where Eex = wr (Ld − Lq )id + wr l pm − (Ld − Lq )(d/dt)iq .
synchronous machines (PMSMs). In the presence of errors in the
position value, a bias may occur in Park’s transformation, thereby Proposed auto-calibration method: Under zero-current control, the
generating ripples in the stator currents and output torque [1]. In PMSM models (6) and (7) are reduced to
addition, the stator currents cannot reach optimal values, such as the
maximum torque per ampere. vd ′ = −Eex sin uoff , (8)
In general, the rotor position is measured using a resolver or encoder vq′ = Eex cos uoff , (9)
attached to the PMSM. Position errors occur when the position offset,
which is the difference between the position reference points of the and the position offset is calculated as
PMSM and sensor, is not accurately identified. The position offset  
−vd ′
includes not only the initial rotor position (i.e. the misalignment uoff = tan−1 . (10)
between the sensor and a permanent magnet), but also the delay vq′
between the time of sensor measurement and control action [2]. The position offset can be expressed in terms of the parameter vector
To identify and compensate for the position offset, auto-calibration f as follows:
methods have been investigated. In [3], the voltage pulse method was
adopted, in which a voltage pulse is applied to each phase of the yk = Zk f, (11)
PMSM, and the initial rotor position is determined by examining a −1
  
combination of changes in each phase current. Although this method where yk = tan −vd ′ ,k /vq′ ,k , Zk = wr, k 1 , and f = [m u0 ]T ,
is simple, the pulse duration must be tuned, and this approach is valid and the parameter vector can be estimated by applying recursive least
only for PMSMs with a large saliency. A current pulse method was squares techniques:
proposed in [4], which does not depend on the saliency; however, 1
direct phase voltage measurement is required. Pk = (Pk−1 − Pk−1 Z Tk (m + Zk Pk−1 Z Tk )−1 Zk Pk−1 ), (12)
m
Furthermore, methods for the sensorless control have been applied
to estimate the position offset. The signal injection method [5, 6], in gk = Pk Z Tk , (13)
combination with advanced signal processing techniques, exhibited a
high accuracy for estimating the initial rotor position. However, this f̂ k = f̂ k−1 + gk (yk − Zk f̂ k−1 ), (14)
method also depends on the saliency, and thus, it is not a general
solution. A model-based approach can be used to estimate the initial where m is the forgetting factor, and Pk and gk represent the gain
rotor position in a simple manner that does not depend on the saliency matrices.
[7]. However, the accuracy of the model-based approach depends on the Considering this result, the offline auto-calibration method is pro-
parameters of the PMSM, the exact values of which are difficult to posed in Fig. 1. First, the mechanical speed wm , defined by wr /P, is
identify at the stage of the calibration. controlled to attain a target value w∗m , which is set to be sufficiently
Considering this background, this Letter presents an auto-calibration large (e.g. the rated speed) to obtain the information regarding the
method for the position offset of PMSM drives, that does not depend on sensor delay (i.e. mwr ). When the magnitude of the speed error is
the saliency and true parameters of the PMSM. This is possible under smaller than the threshold 1wm , the zero-current control is enabled.
the condition of zero-current control at high enough speed where Under the condition where the currents id ′ and iq′ are less than the
parameter-dependent terms are eliminated. To achieve this condition threshold 1i , the parameters m and u0 are estimated using (12)–(14)
using uncertain parameters, speed control with absolute value feedback until the speed naturally decays to wmin m (by mechanical loss), which
and model-free current control are devised. An additional advantage of can be selected to be about 30–50% of w∗m . Finally, the calibrated
using this condition is that the sensor delay, which was not considered in rotor position ucal
r is obtained by adding the estimated offset to the
the existing methods, can also be included in the position offset, which measurement.
can improve the accuracy of the calibration. Experiments are performed To implement this algorithm, speed and current controllers that do not
on a 12 kW PMSM drive to demonstrate the effectiveness of the rely on the accurate values of the position offset and machine parameters
proposed method. are required. However, in general, controllers using uncertain model
parameters do not guarantee their closed-loop stability. Thus, in
Preliminaries: The PMSM is modelled in the d–q frame as follows: the following, absolute value feedback and model-free control are
d introduced as means to ensure the stability of speed and current controls,
Ld id = −Rs id + wr Lq iq + vd , (1) respectively.
dt
d Speed controller with absolute value feedback: The speed dynamics of
Lq iq = −Rs iq − wr (Ld id + l pm ) + vq , (2)
dt the PMSM is represented as
Te = 1.5P(l pm + (Ld − Lq )id )iq , (3) J ẇm + Bwm = Te − TL , (15)
where Ld , Lq , l pm , and Rs denote the d- and q-axes inductances, flux where J and B denote the moment of inertia and damping coefficient
linkage, and stator resistance, respectively; id and iq , and, vd and vq of the PMSM, respectively, and TL is the load torque. To analyse the

ELECTRONICS LETTERS 30th September 2020 Vol. 56 No. 20 pp. 1048–1051


1350911x, 2020, 20, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/el.2020.1669 by Yuan Ze University, Wiley Online Library on [11/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
stability of speed control in the d ′ –q′ frame, it is assumed that id ′ = 0 with the parameter l . 0, the current iy exponentially converges to its
without loss of generality. The output torque (3) in the biased frame reference i∗y with the bandwidth of K pi as ḋ y  0 exponentially.
is rewritten as follows: The proof can be derived using the result in [8]. The current control
bandwidth K pi should be about ten times higher than the speed
Te = 1.5P(l pm + (Ld − Lq ) sin uoff iq′ ) cos uoff iq′ . (16) control bandwidth K. The control laws (20)–(22) are used for the
This equation indicates that the sign of the output torque can vary with block ‘Current controller’ in Fig. 1.
the position offset even for the same current; thus, the stability of the The parameters of the position offset can be changed for long term
speed control cannot be ensured. use. This aspect corresponds to the importance of online auto-calibration
[9]. The online auto-calibration method can be implemented in a similar
manner to the principle of the offline auto-calibration method, by using
w*m the condition of zero-current at high enough speed during the PMSM
operation, as shown in Fig. 2. When the estimated parameters deviate
speed control more than a certain tolerance 1j (j = m, u) from the values obtained in
set i *d ′ i *q ′ s.t. |wm| → w*m the offline calibration, the values are replaced by the estimates.

||wm | – w *m| ≤ εwm wm speed


N calculator
Y
zerocurrent control |id ′| ≤ εi & |iq ′| ≤ εi
θrmeas
& |wm| > wmmin
set i d* ′ → 0,i*q ′ → 0 position
sensor Y

vd ′ offset estimation
|id ′| ≤ εi & |iq ′| ≤ εi i d* ′ current
get m k ,θ 0,k w/(12) – (14)
controller vq ′
Y i q* ′
offset estimation
current
get m k ,θ 0,k w/(12) – (14) sensors
|m – m k | > εm |θ 0 – θ 0,k | > εθ
ia,ic N N
id ′,iq ′ Y Y
|wm| ≤ wmmin dq/abc
N m k → m θ 0,k → θ 0
Y
offset calibration
Fig. 2 Proposed online auto-calibration algorithm
set m k → m ,θ 0,k → θ 0
cal meas
θr = θr + (m wr + θ 0)

θr
cal Experimental results: An experiment was performed to verify the
offline auto-calibration method, employing a 12 kW PMSM drive.
The specifications of the PMSM drive are presented in Table 1. A
Fig. 1 Proposed offline auto-calibration algorithm
digital signal processor (TMS320F28377D) was used to implement
the proposed algorithm. The algorithm parameters were set as follows:
To address the stability issue, the absolute value feedback is used for
w∗m = 1100 RPM, wmin ∗ ∗
m = 0.5wm , 1wm = 0.08wm , 1i = 0.1, and
the speed control, in which the error is defined as ew := w∗m − |wm |.
m = 0.999. The control gains were selected as K pw = 0.2,
Using this approach, the speed can be controlled to w∗m or at least
K pi = l = 628 for the nominal parameter of Ld, 0 = 0.95Ld , so that
−w∗m , which is acceptable for the purpose of position offset calibration.
the bandwidth of the current control is at least ten times that of the
Then, the speed control law can be defined as
speed controller. Two values were used for the initial rotor position
i∗d ′ = 0, i∗q′ = K pw ew , (17) 5
u0 : the true value x (unknown), and x − p, which is obtained by
6
which yields the following closed-loop dynamics assuming that the 5
deliberately adding p to the position measurement umeas
r .
currents rapidly converge to their references: 6

ėw + Kew − (+cos uoff Bw∗m + TL )/J = 0, (18) Table 1: Specifications of the PMSM drive
where rated speed 1200 RPM
DC-link voltage 100 V
K = B/J + 1.5P(l pm + (Ld − Lq ) sin uoff K pw ew )K pw |cos uoff |,
switching frequency 10 kHz
and +cos uoff denotes the sign of cos uoff . With a gain K pw that makes K number of pole pairs (P) 8
positive, ew converges to a value in an almost exponential manner. d-axis inductance (Ld ) 0.55 mH
For a negligible damping effect and load torque such that q-axis inductance (Lq ) 0.80 mH
+cos uoff Bw∗m + TL ≪ J , ew  0. The control law (17) is used for flux linkage (l pm ) 0.056 Wb
the block ‘Speed control’ in Fig. 1. stator resistance (Rs ) 25 mV
Using an integrator is considerable to eliminate the steady-state error, moment of inertia (J ) 0.0001 kg · m2
but this may adversely affect the stability. Therefore, it is desirable to damping coefficient (B) 0.001 Nm · s/rad
reduce the steady-state error to a desired level by tuning the gain K pw .

Model-free current controller: For model-free control, the current


dynamics (6) and (7) can be rewritten as follows: The experimental results are presented in Fig. 3. It can be seen that
vd ′ = 0 (i.e. uoff = 0) after calibrating the position measurement with
d 1 the estimated parameters, implying that the calibrations were successful.
iy = v y + dy , (19)
dt Ld, 0 Comparing Figs. 3a and b, the deviation of m̂ can be noted to be  3%,
5
where y = d ′ , q′ , Ld, 0 is the nominal value of Ld , and dy is the lumped and the difference in û0 is 2.63 rad, which is closely similar to p. The
6
disturbance. With the following control law: finding demonstrates that the position offset estimation was sufficiently
accurate even without relying on true parameters of the PMSM.
vy = Ld, 0 (K pi (i∗y − iy ) − d̂y ), (20) Note that, in Fig. 3b, the speed converged to −w∗m owing to the use of
the absolute value feedback. Without the absolute value feedback, the
where the estimate of the lumped disturbance is obtained using speed would have diverged. Another note is that the model-free
current control worked effectively for the two different initial rotor
ży = −l(zy + liy + vy /Ld, 0 ), (21)
positions. Thanks to these speed and current controllers that do not
rely on the accurate values of the position offset and machine
d̂ y = zy + liy , (22) parameters, the proposed auto-calibration strategy could be realised.

ELECTRONICS LETTERS 30th September 2020 Vol. 56 No. 20 pp. 1048–1051


1350911x, 2020, 20, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/el.2020.1669 by Yuan Ze University, Wiley Online Library on [11/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
speed
zero -current Conclusion: This Letter proposed a novel auto-calibration technique
control offset calibration
control with the estimates for the position offset of PMSM drives, that did not rely on the saliency
& offset estimation
and true parameters of the PMSM. To estimate the position offset, the
1000 proposed technique utilised the condition of zero-current control at
wm, RPM

|wm| = wmmin
high enough speed where parameter dependent terms disappeared.
500 ||wm | – w*m| ≤ εwm The absolute value feedback and model-free control were introduced
as a means to ensure the stability of the speed and current controls,
0 respectively, even with uncertain model parameters. Another feature
20 of the proposed technique was that the sensor delay, which was not
iq ′
current, A

considered in the existing methods, is also included in the position


id ′
10 offset. These features enable an accurate automatic calibration of the
position offset with little information about the target PMSM.
0 Algorithms for both the offline and online calibrations were presented,
× 10–4 and the effectiveness of the proposed approach was experimentally
demonstrated using the 12 kW PMSM drive.
4
m , rad2/s

2.5 × 10–4
0 Acknowledgments: This work was supported by a grant 20TLRP-
C135447-04 from the Transportation & Logistics Research Program
–4
(TLRP) funded by the Ministry of Land, Infrastructure, and Transport
0
of the Korean Government.
θ 0, rad

–0.4 © The Institution of Engineering and Technology 2020


–0.65
Submitted: 10 June 2020 E-first: 10 August 2020
–0.8 doi: 10.1049/el.2020.1669
60 One or more of the Figures in this Letter are available in colour online.
vq ′ K. Choi and Y. Kim (Center for Eco-friendly & Smart Vehicles, Korea
voltage, V

30 vd ′ Advanced Institute of Science and Technology, Daejeon, Republic of


0 Korea)
vd ′ = 0 after calibration S.-K. Kim (Department of Creative Convergence Engineering, Hanbat
–30
0 1 2 3 4 5 6 National University, Daejeon, Republic of Korea)
time, s ✉ E-mail: lotus45kr@gmail.com
a K.-S. Kim (Department of Mechanical Engineering, Korea Advanced
Institute of Science and Technology, Daejeon, Republic of Korea)
Zero-current

speed control
control offset calibration References
& offset with the estimates
estimation 1 Lara, J., Xu, J., and Chandra, A.: ‘Effects of rotor position error in the
0 performance of field-oriented-controlled PMSM drives for electric
vehicle traction applications’, IEEE Trans. Ind. Electron., 2016, 63,
wm, RPM

–500 pp. 4738–4751


||wm| – w*m| ≤ εwm
|wm| = wmmin 2 Moon, H.T., Kim, H.S., and Youn, M.J.: ‘A discrete-time predictive
–1000 current control for PMSM’, IEEE Trans. Power Electron., 2003, 18,
pp. 464–472
10
iq ′ 3 Saadaoui, O., Khlaief, A., Abassi, M., et al.: ‘A sliding-mode observer
current, A

for high-performance sensorless control of PMSM with initial rotor


id ′
5 position detection’, Int. J. Control, 2017, 90, pp. 377–392
4 Antonello, R., Tinazzi, F., and Zigliotto, M.: ‘Benefits of direct phase
0 voltage measurement in the rotor initial position detection for permanent-
magnet motor drives’, IEEE Trans. Ind. Electron., 2015, 62,
× 10–4 pp. 6719–6726
4 5 Xu, P., and Zhu, Z.Q.: ‘Initial rotor position estimation using
m , rad2/s

2.42 × 10–4 zero-sequence carrier voltage for permanent-magnet synchronous


0 machines’, IEEE Trans. Ind. Electron., 2016, 64, pp. 149–158
–4 6 Zhang, X., Li, H., Yang, S., et al.: ‘Improved initial rotor position
estimation for PMSM drives based on HF pulsating voltage signal
4 injection’, IEEE Trans. Ind. Electron., 2017, 65, pp. 4702–4713
7 Wang, G., Li, T., Zhang, G., et al.: ‘Position estimation error reduction
θ 0, rad

2.999 using recursive-least-square adaptive filter for model-based sensorless


2
interior permanent-magnet synchronous motor drives’, IEEE Trans.
Ind. Electron., 2013, 61, pp. 5115–5125
0 8 Kim, S.K., and Ahn, C.K.: ‘Offset-free proportional-type self-tuning
60 speed controller for permanent magnet synchronous motors’, IEEE
vq ′ Trans. Ind. Electron., 2018, 66, pp. 7168–7176
voltage, V

30 vd ′ 9 Bang, J.S., and Kim, T.S.: ‘Automatic calibration of a resolver offset of


vd ′ = 0 after calibration permanent magnet synchronous motors for hybrid electric vehicles’.
0 American Control Conf., Chicago, IL, USA, 2015, pp. 4174–4179

–30
0 2 4 6 8
time, s
b

Fig. 3 Experiment results of offline auto-calibration


a u0 = x 5
b u0 = x − p
6

ELECTRONICS LETTERS 30th September 2020 Vol. 56 No. 20 pp. 1048–1051

You might also like