Due to the dense network of rivers and lakes in Shanghai, scientific and effective monitoring of river and lake water quality is conducive to consolidating the achievements of river and lake management, and serving for the protection and management of water resources in the new situation. The purpose of this paper is to make satellite remote sensing technology be applied effectively in urban water quality monitoring. The research methods are as follows: Based on Sentinel-2 multi-spectral images, an inversion model of urban river water quality parameters is established by using machine learning technology. Dissolved oxygen, permanganate index, ammonia nitrogen and total phosphorus of 103 rivers in Shanghai from 2019 to 2021 are inverted by remote sensing, and the spatio-temporal variation characteristics of water quality parameters of main rivers in Shanghai are analyzed. The water environment of Shanghai is evaluated. The results show that the inversion accuracy of DO, CODMn and TP is better than 80%, and the inversion accuracy of NH3-N is better than 70%. The four water quality parameters are better than that of class V, and the water quality in the first and fourth quarters is better than that in the second and third quarters.