The invention relates to a port illumination
system life evaluation and
health index model training method, which specifically comprises the following steps: firstly, acquiring operation data such as brightness, temperature and
voltage of an illumination
system and salt mist concentration data in real time through multiple sensors, and constructing a
time sequence data set with a real residual life
label; mapping the operation data to a
quantum space to construct a
quantum state and Hamiltonian, and solving an optimal filtering path through annealing optimization to obtain filtering data; then mapping the filtering data into a Riemannian manifold, calculating a Chen-Simons integral value in combination with a Laplacian operator, external differential and the like, and generating a feature deconstruction operator; constructing a fractional order space-time
convolution network to extract features, and constructing and optimizing a residual life prediction model in combination with salt mist concentration and Weber distribution shape parameters; and finally, inputting new preprocessed data to obtain a prediction result, and calculating a
health index. According to the invention, the
data quality and the characteristic representation capability are improved, and the residual life prediction precision of the port
lighting system can be effectively improved.