Large Signal Modeling for Microwave Power Devices Based on LSTM

The characteristics of microwave power devices working in large signal state can be accurately characterized by X-parameters. However, the extraction of X-parameter is inefficient. To improve this problem, an X-parameter modeling method based on long short-term memory (LSTM) neural network is proposed. It is proved by experiments that the mean square error (MSE) of the LSTM model is 0.1291. Therefore, X-parameters of the transistor can be accurately represented by the established model.