To rapidly analyze the biomass quality(soluble matter, cellulose, hemicellulose, lignin, and ash content, as well as cellulose crystallinity and polymerization) of Miscanthus lutarioriparius in Dongting Lake, 126 germplasm resources were collected using two spectral pre-processing methods and three characteristic spectral screening methods. Using different combination methods with optimized spectra and original spectra, a near-infrared spectral analysis model was constructed based the partial least squares method. A dual optimization model was selected for seven quality indicators, and the 126 germplasm resources were evaluated for their industrial potential based on the grey correlation degree method. The results showed that the contents of soluble matter, cellulose, hemicellulose, lignin, and ash, as well as the crystallinity and polymerisation of cellulose in M. lutarioriparius, Dongting Lake, existed in rich diversity and was normally distributed, which was in line with the requirements of near-infrared modeling. The PLS model optimized based on DD derivation and CARS algorithm performed well in the prediction of soluble matter content in M. lutarioriparius, with the root mean square error(RMSEC) of 0.27 and the coefficient of determination(R2C) of 0.99 in the calibration set. The RMSECV of 0.77 and the R2CV of 0.97 in the cross-validation set, and the relative analytical error of 5.07 and the R2V of 0.88 in the prediction set. The PLS model optimized by VCPA-IRIV based on DD combined with variable combination cluster analysis exhibits excellent performance in predicting the cellulose, hemicellulose, lignin and ash content of M. lutarioriparius and its crystallinity and degree of polymerization. The model's RMSEC ranges from 0.14 to 10.20, R2C from 0.98 to 0.99, RMSECV from 0.28 to 19.46, R2CV from 0.94 to 0.98, R2V from 0.87 to 0.98, and the relative percent deviation (RPD) ranges from 4.84 to 15.65. The results also showed that the NIR spectral model based on the dual optimal spectral subset could effectively predict the biomass quality of M. lutarioriparius and had high stability. Through the evaluation of 126 germplasm resources of M. lutarioripariu by the gray correlation degree method, it was found that the industrial utilization potential scores of the 126 samples were normally distributed, with an average value of 54.4. There were 4 first-grade germplasm resources, 40 second-grade germplasm resources, 63 third-grade germplasm resources, 14 fourth-grade germplasm resources, and 5 fifth-grade germplasm resources.