Combined with optical simulation analysis of Trace pro software, a delivery tray suitable for online detection of pomelo fruit is designed. Through the method of structure optimization and material replacement, the tray model with high irradiance/illumination value in the simulation results was processed in kind pomelo fruit was placed on a tray, and the diffuse transmission spectra of pomelo fruit were collected and compared on an independent spectral platform. The optimal tray with test results and pomelo fruit were used for spectral collection. After the spectral data pretreatment, the spectral characteristic points were extracted by continuous projection algorithm (SPA), and the prediction model of PLSR was established for the soluble solid content (SSC) of pomelo fruit. The optimal prediction determination coefficient R_pre² was 0.957. The root means square error (RMSEP) was 0.271 °Brix. The results show that the prediction ability and accuracy of the model are better, and the design of the transfer tray is reasonable.