The invention relates to the technical field of model construction, and particularly discloses a 100GNRZ direct dimming anti-
lightning stroke SOP scrambling
voltage model construction method based on
deep learning, and the method comprises the following steps: S1, collecting the bias
voltage, modulation depth, SOP scrambling data and environmental parameters of a thin film
lithium niobate modulator in a simulated
lightning stroke environment to construct a
data set; s2, performing normalization and
noise filtering on the data, extracting correlation characteristics of bias
voltage and modulation depth and
time sequence characteristics of SOP scrambling by adopting a CNN-LSTM
hybrid network, and introducing physical characteristics of a modulator as constraints; s3, taking SOP tolerance maximization and photoelectric
loss minimization as targets, and training the model through an adaptive
gradient descent algorithm to obtain an anti-
lightning model; and S4, deploying the model to 100GNRZ direct modulation
light transmission equipment, and dynamically generating an optimal bias voltage for compensation by monitoring equipment parameters in real time.