LAUSR

A novel virutal sensing (VS) method with artificial neural network (ANN) and K-means clustering for insulated-gate bipolar transistor (IGBT) current measuring is first proposed in this paper. It provides an alternative way to measure the IGBT current indirectly without a physical current sensor. It satisfies the requirements of cost and integratability in some smart gate drive controlling. In this study, the IGBT analytical model is abstracted to an equivalent expression for the prediction of IGBT current. The machine-learning technologies, the ANN and K-means clustering, are implemented for solving this equivalent expression based on statistical data to work out the empirical VS model for IGBT current prediction. The simulations show that this method is effective and it conforms to the outputs of pure analytical calculation from IGBT PSpice model. The experimental platform is also built to qualify the feasibility and practicability of this method, and it results 3% error by average in IGBT global current prediction.