Parameters optimization based on neural network of practical wavelength division multiplexed decoy-state quantum key distribution

Abstract

The integration of quantum key distribution (QKD) devices with the existing optical fiber networks is of great significance in reducing the deployment costs and saving fiber resources. Wavelength division multiplexing (WDM) is expected to be a desirable approach to fulfill this ultimate task. In this paper, we analyze the dominant noises in WDM-based QKD system and optimize the key parameters based on a modified model with 200 GHz channel spacing. Then, an appropriate decoy-state method is adopted to estimate the system performance considering statistical fluctuations. Finally, a three-layer artificial neural network is used to train and predict the optimal mean photon numbers within different situations. Our work provides a useful method for the parameters optimization of WDM-QKD system and accelerates the practical development of QKD that coexists with the current backbone fiber infrastructure.