An investigation on IGBT junction temperature estimation using online regression method

In order to obtain junction temperature of insulated-gate bipolar transistor (IGBT) on electric vehicle, a number of studies have proposed junction temperature (Tj) estimation methods. However, some of these methods require additional circuits which lead to cost increase, some methods simulate the IGBT thermal process with thermal model, which induces a deviation between real functional results and simulation, and the other algorithm based methods may involve upgrading the main control unit (MCU) with increased costs. This paper proposes a method of IGBT Tj estimation with regression analysis. The presented method firstly extracts the most sensitive factors from IGBT power loss model, and then the factors are applied in a data fitting function to characterize Tj. After that, the mentioned fitting function is validated by regression analysis with statistics Anderson – Darling (AD) and P-value. Finally the Tj estimation fitting function is demarcated by the negative thermal coefficient (NTC) measured temperature (TNTC). With the method using TNTC demarcation, the error bandwidth between estimated Tj and measured Tj is 3.4 °C in total. Compared with the Tj estimation without TNTC demarcation, the error bandwidth of Tj decreased from 6.9°C to 3.4 °C. The Tj estimation method with TNTC demarcation is efficient for an online Tj estimation of IGBT modules without increased costs.