The importance of hyperspectral remote sensing technology in inland water quality monitoring research has achieved fruitful results. This research used the hyperspectral satellite images of ZY-1 02D and considered Nansi Lake, Shandong Province, China as the main research area. First, the Savitzky– Golay (SG) filtering method was used to denoise ZY-1 02D images. Meanwhile, combined with the XGBoost model, the denoised and original images were applied to retrieve the Chlorophyll-a (Chla) concentration in the water. We found that compared with the original image, the signal-to-noise ratio (SNR) of 7–5D and 9–5D filtered images has been improved in varying degrees. Based on the Chla concentration in the water, the three-band parameters of 7–5D, 9–5D, and the original (OD) image were extracted. The SNR of the characteristic bands obtained from the 7–5D image was significantly higher than other OD images, and it had the highest accuracy for Chla concentration inversion (coefficient of determination R2=0.8737, root-mean-square error RMSE = 4.2259 μg·L-1). This study innovatively utilized the SG filtering method to denoise ZY-1 02D hyperspectral satellite images and the XGBoost model applied to the images was established to invert the Chla concentration of water bodies, which realized large-scale visualization and high-precision monitoring of Chla concentration in the Nansi Lake, and provided a new idea for improving the accuracy of remote sensing methods for monitoring the water quality of inland water.