A Novel Simulation Method for Analyzing Diode Electrical Characteristics Based on Neural Networks

Abstract

Based on the equivalent circuit model and physical model, a new method for analyzing diode electrical characteristics based on a neural network model is proposed in this paper. Although the equivalent circuit model is widely used, it cannot effectively reflect the working state of diode circuits under the conditions of large injection and high frequency. The analysis method based on physical models developed in recent years can effectively resolve the above shortcomings, but it faces the problem of a low simulation efficiency. Therefore, the physical model method based on neural network acceleration is used to improve the traditional, equivalent circuit model. The results obtained from the equivalent circuit model and the physical model are analyzed using the finite-difference time-domain method. The diode model based on a neural network is fitted with training data obtained from the results of the physical model, then it is summarized into a voltage–current equation and used to improve the traditional, equivalent circuit method. In this way, the improved equivalent circuit method can be used to analyze the working state of a diode circuit under large injection and high frequency conditions. The effectiveness of the proposed model is verified by some examples.

Characterization of Resistance and Inductance of PIN Diode at mmWave Frequency Using 7-Layer Deep Neural Network

“…It is also sensitive to noise in the measured data since small fluctuations can lead to severe errors in the extracted results, particularly at high frequencies when the signal-tonoise ratio is often low. An alternative method involving machine learning (ML) to analyze the diode characteristics also exists, but it only addresses overall system performance analysis without directly extracting the electrical parameters of the diode [18].…”

Method for the Analysis of Three-Phase Networks Containing Nonlinear Circuit Elements in View of an Efficient Power Flow Computation

“…The computational effort is huge. The reduction of the computational burden proposed in [14,[16][17][18][19][20] can be implemented by only retaining the first and most significant harmonics at the terminals of the nonlinear elements, which can be obtained through measurement. Behavioral frequency-domain models can be computed based on these nonlinear voltage-current relationship measurements.…”