The design of apron floodlighting is influenced by numerous factors, which affect the efficiency and safety of

floodlighting. In this study, 315 sets of lighting schemes were simulated in DIALux evo software to develop an illuminance prediction model. Horizontal illuminance, vertical illuminance, uniformity, glare rating, and zonal illuminance were used as model inputs, while the installation height of high-mast lights, number of LEDs, and

vertical tilt angle were used as model outputs. A prediction model based on particle swarm optimization (PSO)

and back propagation (BP) neural network was constructed and compared with a model built using BP neural

network. The results show that the proposed PSO-BP neural network is more efficient and has higher model fitting accuracy than the BP neural network algorithm, and it avoids the problem of local optima inherent in the BP

neural network algorithm. The predicted illuminance results for the preset scenes using the established optimized

model were highly consistent with the simulation test results, indicating that the model has good predictive accuracy in engineering applications.

PSO-BP neural network, apron floodlighting, high-mast light, LED light, optimization model