An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines

Permanent Magnet Synchronous Machines (PMSMs) are extensively utilized for their ability to deliver accurate position control, and the equivalent circuit characteristics of these machines are essential in several applications, particularly in formulating the control strategy. The study introduces an...

Full description

Saved in:
Bibliographic Details
Main Authors: Sema Nur Ipek, Nur Bekiroglu, Murat Taskiran
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11021566/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849470989555990528
author Sema Nur Ipek
Nur Bekiroglu
Murat Taskiran
author_facet Sema Nur Ipek
Nur Bekiroglu
Murat Taskiran
author_sort Sema Nur Ipek
collection DOAJ
description Permanent Magnet Synchronous Machines (PMSMs) are extensively utilized for their ability to deliver accurate position control, and the equivalent circuit characteristics of these machines are essential in several applications, particularly in formulating the control strategy. The study introduces an ensemble-based methodology for estimating the equivalent circuit parameters of PMSMs consisting of phase resistance (R), magnetizing reactance (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {m}}$ </tex-math></inline-formula>), and leakage reactance (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {l}}$ </tex-math></inline-formula>) via manufacturer catalog data, which eliminates the necessity for experimental setups, high-quality real-time data, and operational disruptions. Six machine learning models-Multilayer Perceptron (MLP), Cascade Forward Neural Network (CFNN), Layer Recurrent Neural Network (LRNN), Transformer-like Network (TRF), Decision Tree (DT), and Support Vector Regression (SVR)&#x2013;were evaluated in the first stage of the study. Among these, LRNN and TRF showed the best performance, with LRNN achieving the highest <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">$0.9212~\pm ~0.0973$ </tex-math></inline-formula>) for the (R) parameter, followed by TRF (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula>: <inline-formula> <tex-math notation="LaTeX">$0.9163~\pm ~0.0561$ </tex-math></inline-formula>). An averaging voting ensemble model is developed by integrating the two highest-performing algorithms, LRNN and TRF, leveraging the strengths of both algorithms. The ensemble model combining TRF and LRNN further improved predictions, achieving an <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of <inline-formula> <tex-math notation="LaTeX">$0.9804~\pm ~0.0151$ </tex-math></inline-formula> and TGF of <inline-formula> <tex-math notation="LaTeX">$0.9827~\pm ~0.0173$ </tex-math></inline-formula> for R, <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of <inline-formula> <tex-math notation="LaTeX">$0.9615~\pm ~0.0306$ </tex-math></inline-formula> for (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {m}}$ </tex-math></inline-formula>), and TGF of <inline-formula> <tex-math notation="LaTeX">$0.9236~\pm ~0.1177$ </tex-math></inline-formula> for (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {l}}$ </tex-math></inline-formula>). It also outperformed individual models in error metrics, with a MAPE of 7.66% for (R) compared to 23.06% (TRF) and 29.42% (LRNN). The visualization analysis confirmed the model&#x2019;s strong predictive capability, as the error distribution is tightly clustered around zero, the estimated values align closely with the ideal line, and the real trends in efficiency and torque across various load conditions are accurately represented. Thus, the model&#x2019;s capacity to accurately predict parameters and represent machine behavior has been revealed, and this method offers a feasible option for the effective use of resources, such as time and labor, in the estimation of PMSM parameters.
format Article
id doaj-art-adc0a29db9bd4eeb8977409f1f7cda3a
institution Kabale University
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-adc0a29db9bd4eeb8977409f1f7cda3a2025-08-20T03:24:59ZengIEEEIEEE Access2169-35362025-01-0113968579687310.1109/ACCESS.2025.357610111021566An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous MachinesSema Nur Ipek0https://orcid.org/0000-0003-0369-4231Nur Bekiroglu1Murat Taskiran2https://orcid.org/0000-0002-6436-6963Department of Electricity and Energy, Istanbul Aydin University, Istanbul, T&#x00FC;rkiyeDepartment of Electrical Engineering, Y&#x0131;ld&#x0131;z Technical University, Istanbul, T&#x00FC;rkiyeDepartment of Electronics and Communication Engineering, Y&#x0131;ld&#x0131;z Technical University, Istanbul, T&#x00FC;rkiyePermanent Magnet Synchronous Machines (PMSMs) are extensively utilized for their ability to deliver accurate position control, and the equivalent circuit characteristics of these machines are essential in several applications, particularly in formulating the control strategy. The study introduces an ensemble-based methodology for estimating the equivalent circuit parameters of PMSMs consisting of phase resistance (R), magnetizing reactance (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {m}}$ </tex-math></inline-formula>), and leakage reactance (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {l}}$ </tex-math></inline-formula>) via manufacturer catalog data, which eliminates the necessity for experimental setups, high-quality real-time data, and operational disruptions. Six machine learning models-Multilayer Perceptron (MLP), Cascade Forward Neural Network (CFNN), Layer Recurrent Neural Network (LRNN), Transformer-like Network (TRF), Decision Tree (DT), and Support Vector Regression (SVR)&#x2013;were evaluated in the first stage of the study. Among these, LRNN and TRF showed the best performance, with LRNN achieving the highest <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">$0.9212~\pm ~0.0973$ </tex-math></inline-formula>) for the (R) parameter, followed by TRF (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula>: <inline-formula> <tex-math notation="LaTeX">$0.9163~\pm ~0.0561$ </tex-math></inline-formula>). An averaging voting ensemble model is developed by integrating the two highest-performing algorithms, LRNN and TRF, leveraging the strengths of both algorithms. The ensemble model combining TRF and LRNN further improved predictions, achieving an <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of <inline-formula> <tex-math notation="LaTeX">$0.9804~\pm ~0.0151$ </tex-math></inline-formula> and TGF of <inline-formula> <tex-math notation="LaTeX">$0.9827~\pm ~0.0173$ </tex-math></inline-formula> for R, <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of <inline-formula> <tex-math notation="LaTeX">$0.9615~\pm ~0.0306$ </tex-math></inline-formula> for (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {m}}$ </tex-math></inline-formula>), and TGF of <inline-formula> <tex-math notation="LaTeX">$0.9236~\pm ~0.1177$ </tex-math></inline-formula> for (<inline-formula> <tex-math notation="LaTeX">$X_{\mathrm {l}}$ </tex-math></inline-formula>). It also outperformed individual models in error metrics, with a MAPE of 7.66% for (R) compared to 23.06% (TRF) and 29.42% (LRNN). The visualization analysis confirmed the model&#x2019;s strong predictive capability, as the error distribution is tightly clustered around zero, the estimated values align closely with the ideal line, and the real trends in efficiency and torque across various load conditions are accurately represented. Thus, the model&#x2019;s capacity to accurately predict parameters and represent machine behavior has been revealed, and this method offers a feasible option for the effective use of resources, such as time and labor, in the estimation of PMSM parameters.https://ieeexplore.ieee.org/document/11021566/Averaging votingCFNNdecision treeensemble learningLRNNMLP
spellingShingle Sema Nur Ipek
Nur Bekiroglu
Murat Taskiran
An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
IEEE Access
Averaging voting
CFNN
decision tree
ensemble learning
LRNN
MLP
title An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
title_full An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
title_fullStr An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
title_full_unstemmed An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
title_short An Ensemble Learning-Based Predictive Parameterization Approach for Permanent Magnet Synchronous Machines
title_sort ensemble learning based predictive parameterization approach for permanent magnet synchronous machines
topic Averaging voting
CFNN
decision tree
ensemble learning
LRNN
MLP
url https://ieeexplore.ieee.org/document/11021566/
work_keys_str_mv AT semanuripek anensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines
AT nurbekiroglu anensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines
AT murattaskiran anensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines
AT semanuripek ensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines
AT nurbekiroglu ensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines
AT murattaskiran ensemblelearningbasedpredictiveparameterizationapproachforpermanentmagnetsynchronousmachines