Sivaji Chakravorti
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sivaji Chakravorti has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Arijit Baral
Indian Institute of Technology (Indian School of Mines) Dhanbad
23 shared publications - Deepak Mishra
Department of Marine Engineering, Indian Maritime University, Kolkata, India
10 shared publications - Biswendu Chatterjee
Electrical Engineering Department, Jadavpur University, Kolkata, India
7 shared publications - Chandra Madhab Banerjee
Philips
7 shared publications - Soumya Chatterjee
National Institute of Technology
7 shared publications - Sovan Dalai
Electrical Engineering Department, Jadavpur University, Kolkata, India
7 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (19) , 28806-28814
Ultrahigh Frequency Sensor for Multisource Partial Discharge Detection
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (19) , 28806-28814
Ultrahigh Frequency Sensor for Multisource Partial Discharge Detection
- 13 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. PP (10709878) , 1
A Novel Method for Estimation of Partial Discharge Locations using Acoustic Sensors and Variation Bayesian Divided Difference Filter
This paper investigates the estimation performance of variational Bayesian divided difference filter (VB-DDF)-based algorithm for localization of partial discharge (PD) events measured using wall mounted acoustic sensors. Localization of PD using nonlinear filters necessitates proper tuning and processing of sensor noise covariance, which requires rigorous offline tuning. VB-DDF proposed in this contribution can adapt the unknown noise covariance matrices using variational Bayesian approach and demonstrably present reliable estimation results. Unknown elements of the covariance matrices are modelled with inverse Wishart distribution and are further estimated by minimizing the Kullback-Leibler (KL) divergence of approximated joint probability density function (PDF) of system states and unknown elements of covariance. The proposed algorithm has been experimentally validated for estimation of PD locations inside a high voltage (HV) apparatus. It has been observed that the proposed VB-DDF method can accurately estimate the location of PD events, which can be implemented in real-life for localization of PD events inside a HV apparatus.
- 13 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. PP (10709878) , 1
A Novel Method for Estimation of Partial Discharge Locations using Acoustic Sensors and Variation Bayesian Divided Difference Filter
This paper investigates the estimation performance of variational Bayesian divided difference filter (VB-DDF)-based algorithm for localization of partial discharge (PD) events measured using wall mounted acoustic sensors. Localization of PD using nonlinear filters necessitates proper tuning and processing of sensor noise covariance, which requires rigorous offline tuning. VB-DDF proposed in this contribution can adapt the unknown noise covariance matrices using variational Bayesian approach and demonstrably present reliable estimation results. Unknown elements of the covariance matrices are modelled with inverse Wishart distribution and are further estimated by minimizing the Kullback-Leibler (KL) divergence of approximated joint probability density function (PDF) of system states and unknown elements of covariance. The proposed algorithm has been experimentally validated for estimation of PD locations inside a high voltage (HV) apparatus. It has been observed that the proposed VB-DDF method can accurately estimate the location of PD events, which can be implemented in real-life for localization of PD events inside a HV apparatus.
- 28 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Delivery
- Vol. 41 (1) , 227-236
Short-Duration Current Analysis for Predictive Aging Estimation in Power Transformer Insulation
The lifespan of power transformer insulation is significantly influenced by its Degree of Polymerization (DP), which reflects the average number of polymerized units in the cellulose chains of paper insulation. As cellulose forms the structural core of the insulation, its DP critically affects both mechanical integrity and electrical performance. Conventional methods for determining DP are invasive and require physical sampling, rendering them unsuitable for in-service transformers. This study presents a non-invasive, circuit-model-based approach to estimate DP using time-domain dielectric response data acquired over a short measurement interval (typically 25 seconds). The method utilizes the initial polarization current decay and its corresponding system response, which are modeled through a transfer function representation of the insulation system. A distinct correlation is observed between the early-time decay characteristics and the underlying DP of the insulation. The proposed technique has been validated on datasets obtained from multiple in-service power transformers. Results confirm that the approach can reliably extract DP-related parameters from short-duration dielectric signals, enabling practical, on-site condition assessment without interrupting transformer operation.
- 28 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Delivery
- Vol. 41 (1) , 227-236
Short-Duration Current Analysis for Predictive Aging Estimation in Power Transformer Insulation
The lifespan of power transformer insulation is significantly influenced by its Degree of Polymerization (DP), which reflects the average number of polymerized units in the cellulose chains of paper insulation. As cellulose forms the structural core of the insulation, its DP critically affects both mechanical integrity and electrical performance. Conventional methods for determining DP are invasive and require physical sampling, rendering them unsuitable for in-service transformers. This study presents a non-invasive, circuit-model-based approach to estimate DP using time-domain dielectric response data acquired over a short measurement interval (typically 25 seconds). The method utilizes the initial polarization current decay and its corresponding system response, which are modeled through a transfer function representation of the insulation system. A distinct correlation is observed between the early-time decay characteristics and the underlying DP of the insulation. The proposed technique has been validated on datasets obtained from multiple in-service power transformers. Results confirm that the approach can reliably extract DP-related parameters from short-duration dielectric signals, enabling practical, on-site condition assessment without interrupting transformer operation.
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. 33 (3) , 2249-2256
Estimation of Moisture Content in Metal Oxide Surge Arrester Based on Frequency Distribution Analysis of Dielectric Relaxation Current
In the present work, a method for the estimation of moisture content in a metal oxide surge arrester (MOSA) is proposed using frequency distribution analysis of dielectric relaxation current. To this end, polymeric MOSA samples with different moisture contents are prepared, and polarization depolarization current (PDC) measurement of the same samples is performed using a Keithley 6517B electrometer. From the measured depolarization/relaxation current data, the relaxation frequency distribution (RFD) function is computed for samples with varying moisture contents. Result indicating that RFD spectrum varies for samples with different moisture contents. Based on their nature of variation, three moisture-sensitive parameters are extracted, and their variation pattern with moisture ingress is analyzed. Subsequently, empirical models incorporating these parameters are developed to quantify moisture levels in MOSA. Validation results demonstrate that these empirical models can reliably estimate moisture content in MOSA, regardless of size variations.
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. 33 (3) , 2249-2256
Estimation of Moisture Content in Metal Oxide Surge Arrester Based on Frequency Distribution Analysis of Dielectric Relaxation Current
In the present work, a method for the estimation of moisture content in a metal oxide surge arrester (MOSA) is proposed using frequency distribution analysis of dielectric relaxation current. To this end, polymeric MOSA samples with different moisture contents are prepared, and polarization depolarization current (PDC) measurement of the same samples is performed using a Keithley 6517B electrometer. From the measured depolarization/relaxation current data, the relaxation frequency distribution (RFD) function is computed for samples with varying moisture contents. Result indicating that RFD spectrum varies for samples with different moisture contents. Based on their nature of variation, three moisture-sensitive parameters are extracted, and their variation pattern with moisture ingress is analyzed. Subsequently, empirical models incorporating these parameters are developed to quantify moisture levels in MOSA. Validation results demonstrate that these empirical models can reliably estimate moisture content in MOSA, regardless of size variations.
- 1 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Plasma Science
- Vol. 53 (8) , 1896-1903
An Efficient and Fast Nondestructive Testing Method for Estimating DP Value of Oil–paper Insulation
The degree of polymerization value (DPv) is a widely recognized indicator of oil–paper insulation degradation in power transformers. However, conventional DPv estimation relies on invasive chemical testing, requiring direct sampling from the transformer, which is impractical for in-service condition monitoring. To overcome this limitation, this study presents a noninvasive and computationally efficient method for DPv estimation based on the interfacial charge characteristics of oil–paper insulation. A quantitative approach is employed, wherein the interfacial charge is derived from polarization–depolarization current (PDC) measurements. The proposed method achieves an estimation accuracy within 3% error, ensuring reliable DPv assessment. Additionally, a qualitative evaluation highlights its ability to eliminate the need for chemical analysis, making it a practical alternative for real-time condition monitoring. The approach is cost-effective, computationally efficient, and suitable for large-scale deployment in transformer maintenance strategies. Furthermore, by eliminating direct transformer sampling, it enhances safety and sustainability while maintaining high accuracy. The method has been validated using experimental data from in-service power transformers, demonstrating its feasibility for practical implementation in insulation health assessment.
- 1 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Plasma Science
- Vol. 53 (8) , 1896-1903
An Efficient and Fast Nondestructive Testing Method for Estimating DP Value of Oil–paper Insulation
The degree of polymerization value (DPv) is a widely recognized indicator of oil–paper insulation degradation in power transformers. However, conventional DPv estimation relies on invasive chemical testing, requiring direct sampling from the transformer, which is impractical for in-service condition monitoring. To overcome this limitation, this study presents a noninvasive and computationally efficient method for DPv estimation based on the interfacial charge characteristics of oil–paper insulation. A quantitative approach is employed, wherein the interfacial charge is derived from polarization–depolarization current (PDC) measurements. The proposed method achieves an estimation accuracy within 3% error, ensuring reliable DPv assessment. Additionally, a qualitative evaluation highlights its ability to eliminate the need for chemical analysis, making it a practical alternative for real-time condition monitoring. The approach is cost-effective, computationally efficient, and suitable for large-scale deployment in transformer maintenance strategies. Furthermore, by eliminating direct transformer sampling, it enhances safety and sustainability while maintaining high accuracy. The method has been validated using experimental data from in-service power transformers, demonstrating its feasibility for practical implementation in insulation health assessment.
- 24 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (9) , 15184-15193
Compensating the Impact of Residual Energy on Time Domain Dielectric Response Using Time-Varying Model
Analysis of polarization and depolarization current is a widely accepted method for diagnosing power transformer insulation. The accuracy of such techniques depends significantly on the premise that measurement of insulation response has been done correctly. During field measurement, equipment sometimes fails to record proper current, even after applying DC charging voltage. In such cases, the polarization current profile gets affected by residual energy. Recently, a Conventional Debye Model (CDM) based approach has been reported to solve the issue. The CDM-based approach relies on identifying the correct time-invariant branch parameters, by minimizing the deviation between measured and estimated value of several performance parameters, through an iterative technique. This coupled with presence of multiple branches in CDM makes the overall method time consuming and computationally intensive. The paper proposes a non-iterative methodology, based on a model with time-varying parameters that is capable of achieving the same result. This not only saves time but also reduces overall data post processing and computation burden required for diagnosis. Performance of the proposed method is tested on data obtained from the oil-paper sample and several real-life power transformers. The proposed method is observed to be capable of estimating paper-moisture (using affected data) with more than 95% accuracy for in-service units. The time required for achieving this is found to be approximately 1/3rd of that required by CDM-based technique (which could provide result with maximum 90% accuracy).
- 24 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (9) , 15184-15193
Compensating the Impact of Residual Energy on Time Domain Dielectric Response Using Time-Varying Model
Analysis of polarization and depolarization current is a widely accepted method for diagnosing power transformer insulation. The accuracy of such techniques depends significantly on the premise that measurement of insulation response has been done correctly. During field measurement, equipment sometimes fails to record proper current, even after applying DC charging voltage. In such cases, the polarization current profile gets affected by residual energy. Recently, a Conventional Debye Model (CDM) based approach has been reported to solve the issue. The CDM-based approach relies on identifying the correct time-invariant branch parameters, by minimizing the deviation between measured and estimated value of several performance parameters, through an iterative technique. This coupled with presence of multiple branches in CDM makes the overall method time consuming and computationally intensive. The paper proposes a non-iterative methodology, based on a model with time-varying parameters that is capable of achieving the same result. This not only saves time but also reduces overall data post processing and computation burden required for diagnosis. Performance of the proposed method is tested on data obtained from the oil-paper sample and several real-life power transformers. The proposed method is observed to be capable of estimating paper-moisture (using affected data) with more than 95% accuracy for in-service units. The time required for achieving this is found to be approximately 1/3rd of that required by CDM-based technique (which could provide result with maximum 90% accuracy).
- 2 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. 32 (4) , 1880-1889
Effect of Charges Freed From Shallow and Deep Traps on Dielectric Response and Associated Insulation Diagnosis
Charges freed from shallow and deep traps affect different portions of the polarization and depolarization current (PDC). Thus, leading to crossover of the two current profiles. As polarization and depolarization currents are disproportionately affected by charge carriers, using either of the profiles may lead to ambiguity or inaccuracy in diagnosis. Detecting the presence of crossover in PDC profiles requires complete current profiles and hence, time-consuming measurements. A methodology is proposed in the present work that can detect such crossover. Results presented in the paper show that the crossover point information, coupled with the polarization current measured for 180 s, is sufficient in obtaining insulation-sensitive parameters for in-service units. The proposed methodology is formulated and validated using data collected from various real-life transformers.
- 2 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dielectrics and Electrical Insulation
- Vol. 32 (4) , 1880-1889
Effect of Charges Freed From Shallow and Deep Traps on Dielectric Response and Associated Insulation Diagnosis
Charges freed from shallow and deep traps affect different portions of the polarization and depolarization current (PDC). Thus, leading to crossover of the two current profiles. As polarization and depolarization currents are disproportionately affected by charge carriers, using either of the profiles may lead to ambiguity or inaccuracy in diagnosis. Detecting the presence of crossover in PDC profiles requires complete current profiles and hence, time-consuming measurements. A methodology is proposed in the present work that can detect such crossover. Results presented in the paper show that the crossover point information, coupled with the polarization current measured for 180 s, is sufficient in obtaining insulation-sensitive parameters for in-service units. The proposed methodology is formulated and validated using data collected from various real-life transformers.
- 22 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
ML-Based Improved Incipient Fault Detection in Power Transformers Using DGA Data
In power transformers, multiple incipient faults are common and can lead to reduced efficiency, increased energy losses, and eventual failure if not addressed properly. Detecting these faults is crucial for effective condition monitoring. This study proposes an innovative classification approach using optimized Machine Learning (ML) techniques, specifically designed for multiple fault diagnosis in some of India's onsite power transformers. The method, based on three-ratios technique and ML classification, achieves high accuracy (98%), enhancing the reliability of fault detection via Dissolved Gas Analysis (DGA). This approach shows promising diagnostic capabilities, crucial for improving transformer reliability and performance prediction in real-world scenarios.
- 22 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
ML-Based Improved Incipient Fault Detection in Power Transformers Using DGA Data
In power transformers, multiple incipient faults are common and can lead to reduced efficiency, increased energy losses, and eventual failure if not addressed properly. Detecting these faults is crucial for effective condition monitoring. This study proposes an innovative classification approach using optimized Machine Learning (ML) techniques, specifically designed for multiple fault diagnosis in some of India's onsite power transformers. The method, based on three-ratios technique and ML classification, achieves high accuracy (98%), enhancing the reliability of fault detection via Dissolved Gas Analysis (DGA). This approach shows promising diagnostic capabilities, crucial for improving transformer reliability and performance prediction in real-world scenarios.
- 30 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-11
Power Transformer Insulation Diagnosis Using De-Trapped Charge Affected Short-Duration Dielectric Response
The polarization-depolarization current (PDC) gets affected by de-trapped charges. Analysis of such currents lead to inaccurate assessment of the transformer insulation condition. Existing insulation model-based methods assume the polarization current to be free of the de-trapped charges. This assumption does not always hold good. Moreover, most of the time series forecasting methods available in the literature are based on this assumption. A novel method has been proposed for obtaining the depolarization current, which is less affected by the de-trapping current, from short-duration de-trapped charge-affected polarization current. The result presented in the paper demonstrates that the use of data that is less affected by de-trapped charges leads to reliable diagnosis. The proposed analysis has been formulated and tested on data collected from various real-life transformers.
- 30 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-11
Power Transformer Insulation Diagnosis Using De-Trapped Charge Affected Short-Duration Dielectric Response
The polarization-depolarization current (PDC) gets affected by de-trapped charges. Analysis of such currents lead to inaccurate assessment of the transformer insulation condition. Existing insulation model-based methods assume the polarization current to be free of the de-trapped charges. This assumption does not always hold good. Moreover, most of the time series forecasting methods available in the literature are based on this assumption. A novel method has been proposed for obtaining the depolarization current, which is less affected by the de-trapping current, from short-duration de-trapped charge-affected polarization current. The result presented in the paper demonstrates that the use of data that is less affected by de-trapped charges leads to reliable diagnosis. The proposed analysis has been formulated and tested on data collected from various real-life transformers.