Deep Neural Network-Based Digital Twin Modeling Method for All-Optical Matching System

In practical applications, the dispersion of high order nonlinear fiber (HNLF) and erbium-doped fiber amplifier (EDFA) will introduce new noise, which may lead to misjudgment of matching results. In this research, based on the actual demand, the multi-layer deep neural network with parallel input is used to model the all-optical matching system based on different matching formats such as OOK, BPSK, QPSK. Experimental results show that this method has a very low bit error rate, can achieve accurate transmission, and greatly simplify our experimental steps. This technique serves as a potential choice for future simulation methods and system modeling.