An Improved Prediction Model of IGBT Junction Temperature Based on Backpropagation Neural Network and Kalman Filter

Table 1

BP neural network training parameters.

Items	Parameters
Nodes in input layer	2
Nodes in hidden layer	10
Nodes in output layer	1
Activation function	Tansig
Output function	Purelin
Training function	Levenberg–Marquardt
Learning rate	0.01
Maximum epochs	1000
Performance goal	10^−5
Validation failures	6
Momentum factor	0.9