A Junction Temperature Prediction Method Based on Multivariate Linear Regression Using Current Fall Characteristics of SiC MOSFETs

Haihong Qin, Yang Zhang, Yu Zeng, Yuan Kang, Ziyue Zhu, Fan WuSensors 2025 · DOI

TL;DR

This paper introduces a novel method for accurately predicting the junction temperature of Silicon Carbide (SiC) MOSFETs by using multivariate linear regression (MLR) on two specific parameters measured during the device's turn-off process: the current fall time ($t_{fi}$) and the current fall energy loss ($E_{fi}$). This approach overcomes the limitations of traditional methods that struggle with the low temperature sensitivity of single parameters, offering a more reliable and practical solution for real-time thermal monitoring in power electronics.

What Problem Does This Paper Address?

Power electronic systems, especially those utilizing SiC MOSFETs, are increasingly critical in various high-power applications. A key factor influencing the reliability and performance of these devices is their junction temperature ($T_j$). Overheating or significant temperature fluctuations can lead to premature failure. While monitoring $T_j$ is crucial, accurately estimating it in real-time is challenging. Traditional methods often rely on single "temperature-sensitive electrical parameters" (TSEPs), but for SiC MOSFETs, these parameters (like on-resistance or switching delays) have limited temperature sensitivity, making precise estimation difficult, especially under varying operating conditions like load current. This paper aims to bridge this gap by developing a more robust and accurate $T_j$ estimation method specifically for SiC MOSFETs.

Methods

The core of this research is the development of a Multivariate Linear Regression (MLR) model. The authors first theoretically analyze and then experimentally validate that the current fall time ($t_{fi}$) and current fall energy loss ($E_{fi}$) during the turn-off phase of SiC MOSFETs are indeed temperature-sensitive. These two parameters are chosen as complementary TSEPs.

The methodology involves:

  1. Theoretical Analysis: Deriving mathematical models for $t_{fi}$ and $E_{fi}$ and demonstrating their dependence on junction temperature ($T_j$) and load current ($I_L$).
  2. Simulation: Using LTspice to simulate the behavior of a SiC MOSFET under varying $T_j$ and $I_L$ to confirm the theoretical findings.
  3. Experimental Validation: Building a double-pulse test (DPT) platform to measure $t_{fi}$ and $E_{fi}$ under controlled $T_j$ and $I_L$ conditions.
  4. MLR Model Development:
    • Initially, MLR models were developed to predict $t_{fi}$ and $E_{fi}$ based on $T_j$ and $I_L$.
  5. Crucially, to predict $T_j$, the MLR model was reformulated to use $T_j$ as the dependent variable, with $t_{fi}$ and $E_{fi}$ as independent variables. This approach effectively decouples the influence of load current ($I_L$), which is a significant advantage.
  6. The final model combines $t_{fi}$ and $E_{fi}$ to predict $T_j$.

Datasets

The study utilizes data generated from simulations and experimental measurements.

  • Simulated Data: Obtained using LTspice XVII with a SiC MOSFET model, varying junction temperatures (30°C to 150°C) and load currents (10A to 30A).
  • Experimental Data: Collected from a custom-built double-pulse test (DPT) platform. The experiments involved a NTHL020N090SC1 SiC MOSFET tested at junction temperatures ranging from 30°C to 150°C (in 20°C increments) and load currents from 10A to 30A (in 5A steps).

Key Insights and Ideas

  • Complementary TSEPs are Key: The paper strongly suggests that relying on a single TSEP for SiC MOSFETs is insufficient for accurate $T_j$ estimation due to low temperature sensitivity. Combining parameters like $t_{fi}$ and $E_{fi}$ provides a more comprehensive thermal signature.
  • Current Fall Phase is Rich in Thermal Information: The turn-off current fall phase, often associated with switching losses, contains significant information about the device's thermal state. The authors successfully leverage this phase.
  • Decoupling Load Current is Crucial: A major challenge in $T_j$ estimation is the influence of operating conditions like load current. The MLR approach, by reformulating the model to predict $T_j$ from $t_{fi}$ and $E_{fi}$, effectively mitigates this load current dependency, leading to higher accuracy.
  • MLR for Thermal Modeling: The application of MLR to model the complex interdependencies between TSEPs, operating conditions, and junction temperature proves to be a powerful and effective technique.

Key Results

  • The proposed MLR method using $t_{fi}$ and $E_{fi}$ achieved a high goodness of fit (R² of 94.11%) for predicting junction temperature ($T_j$).
  • This accuracy represents an improvement of 5.23% compared to using $t_{fi}$ alone as a TSEP.
  • The method effectively mitigates the influence of load current, leading to more reliable predictions across different operating points.
  • The $T_j$-$t_{fi}$-$E_{fi}$ model demonstrated superior suitability compared to models based on turn-off time ($t_{off}$) and turn-off loss ($E_{off}$), achieving a higher R² (94.11% vs. 75.41%).

Limitations and Future Work

  • Model Complexity: While MLR is effective, it relies on fitting models to experimental data. The accuracy is dependent on the quality and range of the calibration data.
  • Device Specificity: The models are developed for a specific SiC MOSFET device (NTHL020N090SC1). While the principles are generalizable, device-specific calibration would be required for other SiC MOSFETs.
  • Future Work: The authors highlight the potential for integrating this method into compact, low-cost, embedded temperature-sensing systems for SiC-based power electronics. The non-intrusive nature and high accuracy make it ideal for real-time thermal monitoring, health management, and reliability assessment. Further research could explore extending this to other SiC device types or different operating regimes.

Summary updated

Abstract

The junction temperature (<i>T</i><sub>j</sub>) is a key parameter reflecting the thermal behavior of Silicon carbide (SiC) MOSFETs and is essential for condition monitoring and reliability assessment in power electronic systems. However, the limited temperature sensitivity of switching characteristics makes it difficult for traditional single temperature-sensitive electrical parameters (TSEPs) to achieve accurate estimation. To address this challenge and enable practical thermal sensing applications, this study proposes an accurate, application-oriented <i>T</i><sub>j</sub> estimation method based on multivariate linear regression (MLR) using turn-off current fall time (<i>t</i><sub>fi</sub>) and fall loss (<i>E</i><sub>fi</sub>) as complementary TSEPs. First, the feasibility of using current fall time and current fall energy loss as TSEPs is demonstrated. Then, a coupled junction temperature prediction model is developed based on multivariate linear regression using <i>t</i><sub>fi</sub> and <i>E</i><sub>fi</sub>. The proposed method is experimentally validated through comparative analysis. Experimental results demonstrate that the proposed method achieves high prediction accuracy, highlighting its effectiveness and superiority in MLR approach based on the current fall phase characteristics of SiC MOSFETs. This method offers promising prospects for enhancing the condition monitoring, reliability assessment, and intelligent sensing capabilities of power electronics systems.
BibTeX
@inproceedings{qin2025junction,
  author = {Haihong Qin and Yang Zhang and Yu Zeng and Yuan Kang and Ziyue Zhu and Fan Wu},
  title = {{A Junction Temperature Prediction Method Based on Multivariate Linear Regression Using Current Fall Characteristics of SiC MOSFETs}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25154828},
}
Figure