International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 13, No. 4, December 2022, pp. 2172~2180

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v13.i4.pp2172-2180  2172

Comparative analysis of features of online numerical methods

used for parameter estimation of PMSM

Komal M. Naikawadi1, Sudhir Madhav Patil1, Kapil Kalantri2, Maneetkumar R. Dhanvijay1

1
Department of Manufacturing Engineering and Industrial Management, College of Engineering Pune (COEP), Maharashtra, India
2
HELLA eMobionics, Hella India Automotive Pvt. Ltd., Pune, Maharashtra, India

Article Info ABSTRACT

Article history: As permanent magnet synchronous motors (PMSM) have high power
density, efficiency, good dynamic performance, and small size they are
Received Jul 26, 2022 becoming popular in electric vehicle (EV) applications. Control performance
Revised Aug 22, 2022 and the efficiency of the system get affected due to electrical, mechanical
Accepted Sep 23, 2022 parameters. Parameters value gets affected by voltage source inverter (VSI)
non-linearities, temperature and magnetic saturation effects. If exact
parameters for particular torque speed requirement are found, the efficiency
Keywords: of system increases. There are various offline and online methods for finding
parameters. Offline methods are easy to implement but requires extra setup
Electrical parameters and estimate parameters in steady state. Because the effects of transient
Numerical methods conditions are taken into account during identification, online methods for
Online estimation obtaining real-time data under running conditions are becoming more
Permanent magnet synchronous popular. An overview about online numerical methods to estimate electrical
motor parameters of PMSM is given. It discusses difference between various
PMSM methods in terms of computational cost, convergence speed, noise and
identification error. Choosing of method will be easy using this work. For
inductance estimation, the extended Kalman filter (EKF) algorithm has an
identification error of 0.24% under temperature effect and -0.3% under VSI
non-linearities effect. The identification error for Rs and ψf using the
recursive least square (RLS) method is 0.5% and 0.02%, respectively, when
temperature is considered. EKF and RLS algorithms are proposed.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Sudhir Madhav Patil,
Department of Manufacturing Engineering and Industrial Management,
College of Engineering Pune (COEP), (An Autonomous Institute of Government of Maharashtra)
Wellesley Road, Shivaji Nagar, Pune: 411005, Maharashtra State, India
Email: [email protected]

1. INTRODUCTION
Permanent magnet synchronous motors (PMSM) are becoming popular in industry for electric
vehicle (EV) applications [1]−[3] due to their characteristics such as high-power density, high efficiency [4],
good dynamic performance, and small size [5]. There are two types of PMSMs depending upon the rotor
construction-interior permanent magnet synchronous motor (IPMSM) and surface mounted permanent
magnet synchronous motor (SPMSM). In SPMSM the permanent magnets are placed on the surface of rotor
having equal value of d-axis inductance Ld and q-axis inductance Lq. Where as in IPMSM the placement of
magnets is embedded inside the rotor [6]. Value of Lq is larger than Ld for IPMSM. Reluctance torque is
present in IPMSM which gives advantage of high torque compare to SPMSM. For getting high control
performance of the system, estimation of accurate motor parameters is necessary. Motor parameters can be
classified into electrical parameters such as stator resistance Rs, d-q axis inductances , Ld, Lq, flux linkages ψf

Journal homepage: http://ijpeds.iaescore.com

Int J Pow Elec & Dri Syst ISSN: 2088-8694  2173

[7], [8] and mechanical parameters like inertia J, viscous friction Bm. Generally, one gets motor parameters
value by requesting manufacturer for data sheet of motor [9] and the nominal parameters like rated voltage,
rated current, number of pole pairs, rated speed, rated torque, and class of insulation are available on motor’s
nameplate. But under running condition the effect of temperature [10], saturation [11], [12] and voltage
source inverter (VSI) non-linearities [13] affects the values such as Rs, Ld, Lq, ψf. This affects the control
performance of the system and indirectly the efficiency [14]. By getting accurate parameters, accurate control
of motor can be developed [15], and efficiency can be increased. Different offline and online methods [16],
[17] are available for characterization of motor. Offline methods which are available for identification of
parameters are finite element method (FEM) [18], AC standstill frequency response method [19], IEC60034-
4, and DC decay test [20]. In offline methods different tests are performed to get parameters [21]. Offline
methods take a lot of time and require additional test settings [22]. In order to improve control performance,
the EV industry is moving toward recognizing the motor parameters in real time utilizing online methods and
updating the control gains of running systems. The goal is to improve system control performance.
There are different methods for online parameter estimation such as recursive least square (RLS)
[23]−[28], model reference adaptive system (MRAS) [29], extended Kalman filter (EKF) [30]−[32], particle
swarm optimization (PSO) [33]−[38], genetic algorithm-based methods [39], modified Jaya algorithm [40],
machine learning (ML) algorithm [41], moving horizon estimator (MHE) [42], Runge-Kutta model based
predictive method [43], recursive error prediction method (RPEM) [44], impedance methods [45], [46], and
Gauss Newton method [47]. This paper includes different numerical methods with their characteristics and
basic working process. The detailed classification of different online methods is provided in section 2.
In almost all literature field oriented control (FOC) for PMSM is considered to implement motor
parameter identification. The basic principle used by FOC is that it converts stationary reference frame to
rotating frame which helps to minimize the complexity of analysis. Clarke and Park transformations are used
for converting the reference. Clarke transformation is used to convert three phase system (abc frame) into an
orthogonal stationary frame (α-β frame). Park transformation is used to convert orthogonal stationary frame
to orthogonal rotating frame (d-q frame). Most online methods from literature are using d-q reference frame
for parameter identification. There is one problem in d-q frame that is ‘rank deficiency’. As per the voltage
(1) and (2) of the d-q frame, the observability of matrix in steady state is 2. There are 4 parameters to be
identified as mentioned above, using voltage equation in d-q frame one can only find 2 parameters. Different
approaches are presented to overcome this problem [21]. These depend on 2 principles: i). To decrease the
number of parameter identification, and ii) To increase the rank of observability matrix. Rank of the matrix
can be increased by considering various running conditions. FOC [22], [48] implementation is done using
Clark and Park transformation in section 2.
This article is further organized as follows. Section 2 gives the basic mathematical model of PMSM
considering the FOC and classification of parameter identification methods. Section 3 provide the different
numerical methods for parameter estimation along with different features. Section 4 presents analysis in
terms of identification error and comparison for various numerical methods. In identification error table
percentage error considering method under some effect or no effect is provided. Finally, section 5 concludes
the paper.

2. METHOD
2.1. Mathematical model of PMSM
In this section mathematical model of PMSM is introduced. By ignoring eddy current loss,
hysteresis loss, magnetic saturation and assuming balanced 3 phase supply [22], PMSM voltage equations in
d-q frame can be written as:
𝑑𝑖𝑑
𝑢𝑑 = 𝑅𝑠 𝑖𝑑 + 𝐿𝑑 − 𝜔𝑒 𝐿𝑞 𝑖𝑞 (1)
𝑑𝑡

𝑑𝑖𝑞
𝑢𝑞 = 𝑅𝑠 𝑖𝑞 + 𝐿𝑞 − 𝜔𝑒 𝐿𝑑 𝑖𝑑 + 𝜔𝑒 𝜓𝑓 (2)
𝑑𝑡

where, ud and uq are stator voltages of d and q axis respectively; id and iq are d and q axis currents
respectively; Rs is the stator resistance; 𝜔𝑒 is angular speed in electrical reference; Ld and Lq are inductances
of d and q axis respectively; ψf is the flux linkages. The (3) provides electromagnetic torque for PMSM.
3
𝑇𝑒 = 𝑃𝑝 𝜓𝑓 𝑖𝑞 (3)
2

Where, Te is electromagnetic torque; PP is number of poles pairs; ψf is flux linkages; iq is q axis current.

Comparative analysis of features of online numerical methods used for parameter … (Komal M. Naikawadi)
2174  ISSN: 2088-8694

Motion equation can be given as (4).

𝑑𝜔𝑒
𝑇𝑒 − 𝑇𝑙 = 𝐽 + 𝐵𝑚 𝜔𝑒 (4)
𝑑𝑡

Where, 𝑇𝑙 is load torque; J is a moment of inertia; Bm is a coefficient of friction; 𝜔𝑒 is angular velocity of

rotor.

2.2. MATLAB model of PMSM with FOC

Figure 1 gives MATLAB model of PMSM with FOC using mathematical model of PMSM.
Conversion of 3 phase to d-q frame is done using a ParkClark function. Mathematical (1), (2), (3) and (4) of
PMSM are implemented for building Simulink model. Figure 2 shows the output speed waveform of the
developed PMSM along with FOC control.

Figure 1. MATLAB Model of PMSM with FOC

Figure 2. Speed waveform

2.3. Steps to reproduce the MATLAB model

− The Clarke and Park transformation function is written using the standard equation of conversion.
− The voltage subsystem shown in Figure 1 is developed using (1) and (2).
− The torque equation subsystem is build using (3).
− The speed is calculated using (4) and built it in speed equation subsystem.
− Constant load torque is given for operation.

Int J Pow Elec & Dri Syst, Vol. 13, No. 4, December 2022: 2172-2180
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2175

− The speed is converted to electrical speed by using formula:

𝜔𝑒 = (𝑝/2) *𝜔𝑚

2.4. Classification of parameter identification methods for PMSM

There are different methods for parameter estimation. Figure 3 shows the classification of different
offline and online methods. Offline methods require the extra test setup [22] and gives estimation in standstill
condition. Generally, data is collected earlier and depending on that parameter are estimated. Whereas for
online methods parameters are found in running conditions and hence different effects like temperature,
saturation, VSI non-linearities can be considered during estimation for developing high control performance.
Online methods can be classified as Numerical methods [11], [12], [14]−[32], [42]−[47], [49]−[51], Observer
based methods [52], AI- ML based methods [33]−[41], [53], [54]. The focus of this article is on different
online numerical methods for electrical parameter estimation.

Figure 3. Classification of parameter identification methods

3. LITERATURE SURVEY
3.1. Online numerical methods for parameter estimation
In this section brief about different online numerical methods is given along with the advantages and
disadvantages of it. Recursive least square method is a classical approach for parameter identification. It is
used to estimate weight coefficients to minimize least square cost function [21]. Forgetting factor f directly
affects the promptness of coefficient. Higher the value of f, the slower the estimation. Closer the value of f to
one, older estimations are equally weighted. Smaller values of f consider only the latest measurements but
increasing the value of the covariance Q as inferable [23]. Figure 4 shows the basic working of RLS method.
Depending upon the current state and estimated state error weighted coefficients are being updated. RLS
algorithm is best in steady state conditions having less execution time compared to EKF [29] and has simple
theoretical derivation and implementation [23]. Apart from this it is sensitive to noise and disturbances [29].
It may not converge to accurate values in d-q frame of reference [21]. It has drawback of data saturation and
has fixed gain due to which estimation accuracy gets reduced [29]. It may not perform well with low speed
and light load.
Moving horizon estimation is the performance of MHE is better than Unscented Kalman Filter and
EKF under transient condition. It has more accuracy compared to RLS, MRAS, EKF and Unscent Kalman
Comparative analysis of features of online numerical methods used for parameter … (Komal M. Naikawadi)
2176  ISSN: 2088-8694

Filter (UKF) [6], [28]. Extended Kalman filter is the optimal recursive estimation algorithm based on least
square method used to estimate states of dynamic nonlinear system [29]. Taylor series is used to linearize the
dynamic model. For estimation, it takes measurement noise R and processing noise Q into account.
Execution time for EKF is longer than RLS and MRAS [12], [17]. It is based on discrete system model of
electrical system and has better optimization capability, good convergence in simultaneously estimating
PMSM electrical parameters [4], [14]. Figure 5 shows the working of EKF considering the noise effects.
Kerid [32] has considered temperature variation while implementing EKF algorithm for parameter
estimation. Main advantage of EKF is that it rejects the process and measurement noise having similar
accuracy as UKF [45], MHE. EKF with Gradient correction has small calculation, high accuracy, and fast
convergence speed. It has been observed in literature that EKF has a complex structure with high
computational burden and comparative longer execution time than MRAS and RLS. It is difficult to design
the algorithm for multi parameter measurement.

Figure 4. RLS algorithm

Figure 5. EKF Algorithm

The main drawback of EKF is that it has large sampling time. This is overcome by DKF method.
The results of Jacobian matrix for DKF is identity matrix. This algorithm is least noisy and has accurate
mean value under load step condition. Computational load can be reduced using this method. The idea behind
model reference adaptive system is that it makes error calculation from reference model and a adjustable
model as shown in Figure 6 using adaptive mechanism error is minimized. Adaptive mechanism includes
Popov stability criteria and Lyapunov stability theorem. MRAS has good results and less implementation
complexity [29]. MRAS possesses advantage of less execution time than EKF. This method is sensitive to
noise, and it is difficult to be used in multi parameter identification of missing rank. It increases the difficulty
of identification algorithm [29]. The gain matrix of recursive prediction error method can be identified using
different algorithm like Gauss Newton method (GNA) and the stochastic gradient algorithm (SGA).
According to Perera [44], in low-speed region GNA has rapid adaptation of flux linkages than SGA.
Simultaneous adaptation without zonal scheduling scheme is possible with GNA [47]. This method is more
effective when steady state solution used for flux linkage and dynamic part for Rs. To define system
characteristics fractional mathematical models are more accurate than integer order models because essence
of capacitance, inductance are fractional order. Li [51] used Levenberg-Marquardt algorithm (LM) with
variable damping factor to identify fractional order model parameters with steps. Akpunar [39] proposed
extended state observer predictive speed control algorithm for PMSM drives and implemented using Runge-
Kutta method. Runge-Kutta method has simplicity of modelling and better constraint handling capability. It
is robust but it has computational burden. Rengifo [46] represented instantaneous input impedance model
along with disturbance discrete model considering features such as harmonics, saturation. Values of
parameters according to interior points and genetic algorithm are given. Interior points algorithm requires

Int J Pow Elec & Dri Syst, Vol. 13, No. 4, December 2022: 2172-2180
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2177

less computing time. For multiparameter coupling problems, an improved PSO algorithm based on Gaussian
decline and Gaussian disturbance can be used [34]. In, a metaheuristic algorithm for a continuous time
system using a photovoltaic model is implemented [35] and compared to PSO in [38]. To estimate Li-ion
battery parameters with high convergence speed and low complexity, artificial ecosystem-based optimization
is used [36]. In the future, a metaheuristic algorithm for PMSM parameter estimation could be investigated.

Figure 6. MRAS algorithm

4. DISCUSSION
Table 1 gives comparative analysis between different methods in terms of convergence speed,
computational complexity, initial value requirement and sensitivity to noise. RLS and RPEM have fast
convergence speed compared to others. EKF has property to reject measurement and process noise, but EKF
have more computational complexity compared to RPEM. Depending upon method used the criteria for
initial value requirement changes. Some methods require initial parameter value so that convergence time get
reduced. Noise sensitivity is an important factor which affects the accuracy of algorithm. Kalman filter have
advantage of considering noise while estimating the parameters.

Table 1. Comparative study of methods

Methods
Features
RLS EKF MRAS HF Injection RPEM Fractional order-based LM
Convergence speed Faster Slower Slower - Fastest Fast
Computational Complexity Less More Less - Substantial -
Initial value requirement - Required - - Required Required
Sensitive to noise More Immune Less More - -

Table 2 gives brief about percentage of identification error for parameters considering the method
and different effects. Operating conditions of motor such as temperature, magnetic saturation affects stator
resistance and inductances. Rs, Ld, Lq vary with rotor position and frequency. In EV application the inverter
non-linearities also need to be considered while controlling the performance for efficiency. Table 2 provides
the identification error percentage for parameter estimation using different methods under various
consideration such as VSI non-linearities, temperature, saturation, or no consideration. The identification
error can be calculated as:

|𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛−𝑡𝑟𝑢𝑒|
𝐼𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑒𝑟𝑟𝑜𝑟 = ∗ 100%
𝑡𝑟𝑢𝑒

Here, identification and true means the estimated or identified and the true value of the parameter
respectively. Identification error is the error in estimating the parameter value and the actual value.
One can use Table 1 results to decide the method for parameter identification depending upon the
features one wants for estimation. Along with it Table 2 gives scope for method finalization in terms of
identification error of method either without any consideration or with considering effects such as
temperature and VSI non- linearities.

Comparative analysis of features of online numerical methods used for parameter … (Komal M. Naikawadi)
2178  ISSN: 2088-8694

Table 2. Identification error considering various effects

Method or Without any consideration Temperature effect VSI Non-Linearities
Add-ons or reference
Effect Rs Ld Lq Ψf Rs Ld Lq Ψf Rs Ld Lq Ψf
RLS 4.61 1.87 2.45 -2.5 - - - - - - - - [21]
- - - - -2.3 -0.06 0.19 0.02 - - - - [49]
- - - - 0.5 0.36 0.86 - - - - - MTPA Control [4]
MRAS (10-6), Improved adaptive law
- - - - 0.003 0.0004 0.0004 0.0004 - - - -
[29]
- - - - - - - - 4 - - - [29]
EKF - - - - 1.8 0.24 0.24 - - - - - Gradient correction [31]
- - - - - - - - -2.3 -0.3 -0.3 -1.5 [32]
HF Injection - - - - - - - - 5 5 6 9.5 [41]
RPEM - - - - - - - - - - 10 SGA [40]

5. CONCLUSION
This paper represents an overview on different numerical methods that are available for online
parameter estimation considering that the position sensor is present in the system. The basic idea of methods
with their advantages and disadvantages are provided. It also gives identification error percentage values for
numerical methods with consideration of various effects. The comparison between methods in terms of
convergence speed, computational complexity, sensitivity to noise and initial value requirement is also
provided. Selection of the numerical method for parameter identification of PMSM is possible using this
literature depending upon the user’s constraint for method selection such as convergence speed, identification
error range, and computational time. Further scope is to implement any of the method to get accurate
parameters of motor.
Having fast convergence speed, RLS is the most used method for motor parameter estimation but it
does not consider the noise effect. Having noise consideration by default and if temperature and saturation
effects are to be considered, EKF is the useful method for parameter estimation but the computational
complexity is more. As a future scope one can use EKF for Ld, Lq estimation and RLS for Rs and flux linkage
estimation of PMSM, as EKF has more computational burden.

REFERENCES
[1] S. Fang, H. Liu, H. Wang, H. Yang, and H. Lin, “High power density PMSM with lightweight structure and high-performance
soft magnetic alloy core,” in IEEE Trans. Appl. Supercond., vol. 29, no. 2, pp. 1–5, March 2019, Art no. 0602805, doi:
10.1109/TASC.2019.2891630.
[2] S. H. Park, E. C. Lee, J. C. Park, S. W. Hwang, and M. S. Lim, “Prediction of mechanical loss for high power density PMSM
Considering Eddy Current Loss of PMA and Conductors,” in IEEE Trans. Magn., vol. 57, no. 2, pp. 1-5, Feb. 2021, Art No.
6300205, doi: 10.1109/TMAG.2020.3007439.
[3] C. L. Jeong, Y. K. Kim, and J. Hur, “Optimized design of pmsm with hybrid-type permanent magnet for improving performance
and reliability,” in IEEE Trans. Ind. Appl., vol. 55, no. 5, pp. 4692–4701, Sept.-Oct. 2019, doi: 10.1109/TIA.2019.2924614.
[4] Y. Xu, K. Yang, A. Liu, X. Wang, and F. Jiang, “Online parameter identification based on MTPA operation for IPMSM,” in 2019
22nd ICEMS, pp. 1-4, doi: 10.1109/ICEMS.2019.8921761.
[5] P. Kumar and L. Alberti, "Sensorless control of a PMSM for dynamic control performance evaluation," in 2017 Twelfth
International Conference on Ecological Vehicles and Renewable Energies (EVER), 2017, pp. 1-5, doi:
10.1109/EVER.2017.7935898.
[6] T. Zwerger and P. Mercorelli, “A dual Kalman filter to identify parameters of a permanent magnet synchronous motor,” in 24th
ICSTCC, pp. 619–623, IEEE, 2020, doi: 10.1109/ICSTCC50638.2020.9259686.
[7] F. Fernandez-Bernal, A. Garcia Cerrada, and R. Faure, “Determination of parameters in interior permanent-magnet synchronous
motors with iron losses without torque measurement,” in IEEE Trans. Ind. Appl., vol. 37, no. 5, pp. 1265–1272, 2001, Sept.-Oct.
2001, doi: 10.1109/28.952501.
[8] P. H. Mellor, M. A. Al-Taee, and K. J. Binns, “Open loop stability characteristics of synchronous drive incorporating high field
permanent magnet motor,” Proc. Inst. Elect. Eng. - Elect. Power Appl., vol. 138, no. 4, pp. 175–184, 1991,
doi: 10.1049/ip-b.1991.0022.
[9] KOMOTEK, “Ac servo motor product catalogue,” 2013. [Online]. Available: http://komotek.com/wp-
content/uploads/2018/06/Servo-Motor-Catalogue_KOMOTEK-2017.11.pdf.
[10] H. Liu, C. Zhang, and Q. Huang, “Model reference adaptive parameter identification of pmsm based on single-loop model
predictive control,” in 40th CCC, pp. 1365–1372, IEEE, 2021, doi: 10.23919/CCC52363.2021.9549990.
[11] Z. Liu, G. Feng, and Y. Han, “Extended-Kalman-filter-based magnet flux linkage and inductance estimation for PMSM
considering magnetic saturation,” in 36th YAC, pp. 430–435, IEEE, 2021, doi: 10.1109/YAC53711.2021.9486499.
[12] T. Zwerger and P. Mercorelli, “Combining SMC and MTPA using an EKF to estimate parameters and states of an interior
PMSM,” in 2019 20th ICCC, pp. 1–6, IEEE, 2019, doi: 10.1109/CarpathianCC.2019.8766063.
[13] “IEEE guide for test procedures for synchronous machines,” IEEE Standard 115-2009 (Revision IEEE Standard 115-1995), May
2010, pp. 1–219, doi: 10.1109/IEEESTD.2020.9050934.
[14] X. Li, P. Zhang, and L. Yan, “Sensorless control of permanent magnet synchronous motor with online parameter identification,”
in Transactions of China Electrotechnical Societ, vol. 31, no. 14, pp. 139–147, 2016. [Online] Availabke at: http://dgjsxb.ces-
transaction.com/EN/Y2016/V31/I14/139.
[15] J. Yang, W.-H. Chen, S. Li, L. Guo, and Y. Yan, “Disturbance/uncertainty estimation and attenuation techniques in PMSM drives

Int J Pow Elec & Dri Syst, Vol. 13, No. 4, December 2022: 2172-2180
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2179

a survey,” in IEEE Trans. Ind. Electron., vol. 64, no. 4, pp. 3273–3285, April 2017, doi: 10.1109/TIE.2016.2583412.
[16] X. Ma and C. Bi, “A technology for online parameter identification of permanent magnet synchronous motor,” CES Trans. Electr.
Mach. Sys., vol. 4, no. 3, pp. 237–242, 2020, doi: 10.30941/CESTEMS.2020.00029.
[17] W.-H. Kim et al., “Inductance calculation in IPMSM considering magnetic saturation,” in IEEE Trans. Magn., vol. 50, no. 1,
pp. 1–4, 2013, doi: 10.1109/TMAG.2013.2277586.
[18] P. Zhou, D. Lin, G.Wimmer, N. Lambert, and Z. J. Cendes, “Determination of d-q axis parameters of interior permanent magnet
machine,” IEEE Trans. Magn., vol. 46, no. 8, pp. 3125–3128, Aug. 2010, doi: 10.1109/TMAG.2010.2043507.
[19] Y. Gao, R. Qu, and Y. Liu, “An improved AC standstill method for inductance measurement of interior permanent magnet
synchronous motors,” Int. Conf. Electr. Mach. Sys. IEEE, 2013, pp. 927–931, doi: 10.1109/ICEMS.2013.6754404.
[20] M. Hasni, O. Touhami, R. Ibtiouen, M. Fadel, and S. Caux, “Synchronous machine parameter identification by various excitation
signals,” in Electr. Eng., vol. 90, no. 3, pp. 219–228, 2008, doi: 10.1007/s00202-007-0069-z.
[21] Y. Yu et al.,“Full parameter estimation for permanent magnet synchronous motors,” in IEEE Trans. Ind. Electron., vol. 69, no. 5,
pp. 4376– 4386, 2021, doi: 10.1109/TIE.2021.3078391.
[22] X. Li and R. Kennel, “Comparison of state-of-the-art estimators for electrical parameter identification of PMSM,” in IEEE Int.
Symp. Predict. Control Electr. Drives and Power Electron. (PRECEDE), pp. 1-6, 2019, doi: 10.1109/PRECEDE.2019.8753197.
[23] F. Tinazzi, P. G. Carlet, S. Bolognani, and M. Zigliotto, “Motor parameter-free predictive current control of synchronous motors
by recursive least-square self-commissioning model,” in IEEE Trans. Ind. Electron., vol. 67, no. 11, pp. 9093–9100, 2019, doi:
10.1109/TIE.2019.2956407.
[24] D. Shao and Y. Huang, “parameter identification of permanent magnet synchronous motor based on multi-innovation least square
method,” in 2020 IEEE 9th Joint Int. Inf. Technol. Artif. Intell. Conf. (ITAIC), vol. 9, pp. 1810–1814, IEEE, 2020 doi:
10.1109/ITAIC49862.2020.9339029.
[25] M. Bui, D. Xiao, and M. Rahman, “A new online parameters identification method for IPMSMs using current derivative
measurement,” in 10th Int. Conf. Power Electron ECCE Asia (ICPE 2019-ECCE Asia), pp. 1–7, IEEE, 2019, doi:
10.23919/ICPE2019-ECCEAsia42246.2019.8797292.
[26] J. Long, M. Yang, Y. Li, Y. Chen and D. Xu, “Parameter identification of permanent magnet synchronous motors: sequence
strategy comparative study,” in 2017 IEEE Transp. Electrific. Conf. Expo, Asia-Pacific (ITEC Asia-Pacific), 2017, pp. 1-6. Doi:
10.1109/ITEC-AP.2017.8080944.
[27] A. Zwartbol, J. Dong, P. Bauer, and H. Polinder, “Online parameter estimation of PMSM in EV powertrain including thermal
measurements,” in 2019 IEEE Transp. Electrific. Conf. Expo (ITEC), pp. 1–6, IEEE, 2019, doi: 10.1109/ITEC.2019.8790461.
[28] F.-J. Lin, S.-Yi Chen, W.-Ting Lin, and C.-Wei Liu, “An online parameter estimation using current injection with intelligent
current-loop control for IPMSM drives,” in Energies, vol. 14, no. 23, 8138, 2021, doi: 10.3390/en14238138.
[29] Z. Ye, T. Liu, M. Fuller, and G. Griepentrog, “Parameter Identification of PMSM based on MRAS with considering nonlinearity
of inverter,” in IECON 2019-45th Conf. IEEE Ind. Electron. Soc., vol. 1, pp. 1255–1260, IEEE, 2019, doi:
10.1109/IECON.2019.8927595.
[30] X. Li and R. Kennel, “General formulation of Kalman-filter-based online parameter identification methods for VSI-fed PMSM,”
in IEEE Trans. Ind. Electron., vol. 68, no. 4, pp. 2856–2864, 2020, doi: 10.1109/TIE.2020.2977568.
[31] Y. Zhang, Y. Bi, and S. Wang, “parameter identification of permanent magnet synchronous motor based on extended Kalman
filter and gradient correction,” in IEEE Int. Conf. Mechatron. Autom., pp. 718–722, IEEE, 2020,
doi: 10.1109/ICMA49215.2020.9233766.
[32] R. Kerid, H. Bourouina, and R. Yahiaoui, “Parameter identification of PMSM using EKF with temperature variation tracking in
automotive application,” PEN, vol. 6, no. 2, pp. 109–119, 2018, doi: 10.21533/pen.v6i2.168.
[33] Y. Cheng, M. Zhao and Q. Liu, “Online parameter identification of PMSM based on LAWPSO,” in 2020 IEEE 4th ITNEC, 2020,
pp. 2188-2192, doi: 10.1109/ITNEC48623.2020.9084925.
[34] Z. Li, D. Chen, Y. Chen, H. Lei, and H. Zhu, “PMSM parameter identification based on improved PSO,”
J. Phys.: Conf. Ser., vol. 1754, no. 1, p. 012235, 2021, doi: 10.1088/1742-6596/1754/1/012235.
[35] J. J. Jui and M. A. Ahmad, “A hybrid metaheuristic algorithm for identification of continuous-time Hammerstein systems,” in
Appl. Math. Modell., 95, 339–360, 2021, doi: 10.1016/j.apm.2021.01.023.
[36] S. Ferahtia, A. Djeroui, H. Rezk, A. Chouder, A. Houari, and M. Machmoum, “Optimal parameter identification strategy applied
to lithium‐ion battery model,” in Int. J. Energy Res. 45, no. 11 ,2021 pp. 16741-16753, doi: 10.1002/er.6921.
[37] B. Arabsalmanabadi, N. Tashakor, Y. Zhang, K. Al-Haddad and S. Goetz, “Parameter estimation of batteries in MMCs with
parallel connectivity using PSO,” in IECON 2021–47th Conf. IEEE Ind. Electron. Soc, pp. 1-6. IEEE, 2021, doi:
10.1109/IECON48115.2021.9589293.
[38] I. B. Ali, M. W. Naouar and E. Monmasson, “Parameters identification for a photovoltaic module: comparison between PSO, GA
and CS metaheuristic optimisation algorithms,” Int. J. Modell., Identif. Control, vol. 38, no. 3-4, pp.211-230, 2022.
[39] A. Avdeev and O. Osipov, “PMSM identification using genetic algorithm,” in 2019 26th Int. Workshop Electric Drives:
Improvement in Efficiency of Electr. Drives (IWED), 2019, pp. 1-4, doi: 10.1109/IWED.2019.8664250.
[40] H. P. Huy Anh, P. Quoc Khanh and C. Van Kien, “Advanced PMSM machine parameter identification using modified jaya
algorithm,” in ICSSE, 2019, pp. 445-450, doi: 10.1109/ICSSE.2019.8823434.
[41] R. Savant, A. A. Kumar and A. Ghatak, “Prediction and analysis of permanent magnet synchronous motor parameters using
machine learning algorithms,” in 3rd ICAECC, 2020, pp. 1-5, doi: 10.1109/ICAECC50550.2020.9339479.
[42] P. Kühl, M. Diehl, T. Kraus, J. P. Schlöder, and H. GeorgBock, “A real-time algorithm for moving horizon state and parameter
estimation,” Comput. Chem. Eng., vol. 35, no. 1, 10 January 2011, pp. 71-83, doi: 10.1016/j.compchemeng.2010.07.012.
[43] A. Akpunar and S. Iplikci, “Control and parameter estimation of PMSM by Runge-Kutta model based predictive control,” in
12th ATEE, pp. 1–6, IEEE, 2021, doi: 10.1109/ATEE52255.2021.9425209.
[44] A. Perera and R. Nilsen, “A framework and an open-loop method to identify PMSM parameters online,” in 23rd ICEMS, pp.
1945–1950, IEEE, 2020, doi: 10.23919/ICEMS50442.2020.9291135.
[45] Q. Wang, G. Wang, N. Zhao, G. Zhang, Q. Cui, and D. Xu, “An impedance model-based multiparameter identification method of
PMSM for both offline and online conditions,” in IEEE Trans. Power Electron., vol. 36, no. 1, pp. 727–738, 2020, doi:
10.1109/TPEL.2020.3000896.
[46] J. Rengifo, F. Vaca, and J. M. Aller, “Instantaneous input impedance method for PMSM parameter estimation,” in IEEE ICIT, pp.
205–210, IEEE, 2020, doi: 10.1109/ICIT45562.2020.9067126.
[47] A. Perera and R. Nilsen, “Gauss-Newton: A prediction-error-gradient based algorithm to track PMSM parameters online,” in
IEEE PEDES, pp. 1–8, IEEE, 2020, doi: 10.1109/PEDES49360.2020.9379424.
[48] “PMSM field-oriented control,” mathworks help manual,
https://in.mathworks.com/help/physmod/sps/ref/pmsmfieldorientedcontrol.html [accessed: 22 June 2022].
Comparative analysis of features of online numerical methods used for parameter … (Komal M. Naikawadi)
2180  ISSN: 2088-8694

[49] C. Candelo-Zuluaga, J.-R. Riba, and A. Garcia, “Pmsm parameter estimation for sensorless foc based on differential power
factor,” in IEEE Trans. Instrum. Meas., vol. 70, pp. 1–12, 2021, doi: 10.1109/TIM.2021.3096861.
[50] X. Xiao, C. Chen and M. Zhang, “Dynamic permanent magnet flux estimation of permanent magnet synchronous machines,”
in IEEE Trans. Appl. Supercond., vol. 20, no. 3, pp. 1085-1088, June 2010, doi: 10.1109/TASC.2010.2041435.
[51] Y. Li, X. Wang, and S. Lu, “Fractional order parameter identification of PMSM based on improved Levenberg-Marquardt
algorithm,” in CAC, pp. 4219–4223, IEEE, 2019, doi: 10.1109/CAC48633.2019.8996888.
[52] M. A. Hamida, J. De Leon, A. Glumineau, and R. Boisliveau, “An adaptive interconnected observer for sensorless control of PM
synchronous motors with online parameter identification,” IEEE Trans. Ind. Electron., vol. 60, no. 2, pp. 739–748, Feb. 2013, doi:
10.1109/TIE.2012.2206355.
[53] R. Raja, T. Sebastian, and M.Wang, “Online stator inductance estimation for permanent magnet motors using PWM excitation,”
IEEE Trans. Transport. Electrific., vol. 5, no. 1, pp. 107–117, Mar. 2019, doi: 10.1109/TTE.2019.2891047.
[54] Y. Wang, Y. Xu and J. Zou, “Online multiparameter identification method for sensorless control of SPMSM,” in IEEE Trans.
Power Electron., vol. 35, no. 10, pp. 10601-10613, Oct. 2020, doi: 10.1109/TPEL.2020.2974870.

BIOGRAPHIES OF AUTHORS

Komal M. Naikawadi received Bachelor of Technology degree in Electrical

Engineering from Government College of Engineering, Karad, Maharashtra, India. She is
currently pursuing Master of Technology in ‘Mechatronics’ at College of Engineering Pune
(COEP), An Autonomous Institute of Government of Maharashtra. She can be contacted at
email: [email protected].

Sudhir Madhav Patil received Bachelor of Engineering degree in Mechanical

Engineering from the North Maharashtra University, Maharashtra, India and Master’s and
Ph.D. degree in Production Engineering from the Savitribai Phule Pune University (SPPU),
Maharashtra, India. He is currently working as Associate Professor in the Department of
Manufacturing Engineering and Industrial Management of College of Engineering Pune
(COEP), An autonomous Institute of Government of Maharashtra. He is a Member of The
Institution of Engineers (India), American Society of Mechanical Engineers (ASME) and a
Life Member of Tribology Society of India (LMTSI). His main research interest includes
Mechatronics, Manufacturing Automation, Robotics and AI, and Tribology. He has
published several research papers and is a co-inventor for couple of Indian patents. He can
be contacted at email: [email protected].

Kapil Kalantri received Bachelor of Engineering degree in Electronics and

Telecommunication Engineering from the Savitribai Phule Pune University (SPPU),
Maharashtra, India. He is currently working as Head, Design and Development in HELLA
eMobionics at HELLA India Automotive Pvt. Ltd., Pune, India. He can be contacted at
email: [email protected].

Maneetkumar R. Dhanvijay is working as Associate Professor in the

Department of Manufacturing Engg. and Industrial Management, College of Engineering,
Pune (COEP). He completed his Ph.D. in Mechanical Engineering from College of
Engineering, Pune in 2017 and Master’s in Mechanical-Production Engg. from Government
College of Engineering, Karad, Maharashtra in 2004. He is a member of the Institution of
Engineers (India) and Indian Society for Technical Education. His research interest is non-
Conventional machining and mechatronics. He can be contacted at email:
[email protected].

Int J Pow Elec & Dri Syst, Vol. 13, No. 4, December 2022: 2172-2180