In order to investigate the mechanism of cold island effect in plateau lakes, Fuxian Lake, Xingyun Lake and Qilu Lake in Yuxi, Yunnan Province, China, were taken as research objects. Firstly, the land surface temperature inversion of the study area in 2015, 2017, 2019 and 2021 was completed based on remote sensing technology, and Geographic Information System (GIS) technology was used to classify the temperature of the lake region. Then on this basis, the temporal and spatial variation characteristics of the cooling range and cooling intensity of the lakes were analyzed, and Statistical Package for the Social Sciences (SPSS) was used to analyze the correlation between temperature and multiple landscape indices to determine the influencing factors of temperature change. Finally, ridge regression analysis was innovatively adopted to build a fitting model. The results showed that: (1) From 2015 to 2021, the areas of high temperature and sub high temperature zones increased firstly and then decreased, while the areas of low temperature and sub low temperature zones were opposite to the former, and the area of middle temperature zone continued to decrease. (2) The cooling range of Fuxian Lake was about 150–200 m beyond the water boundary, and the cooling ranges of Xingyun Lake and Qilu Lake were about 100–150 m beyond the water boundary. The average cooling intensity of Fuxian Lake, Xingyun Lake and Qilu Lake is about 7.5 ℃, 5.5 ℃ and 5.4 ℃ respectively. (3) The cold island effect of Fuxian Lake was significantly affected by water distribution, vegetation distribution and building distribution, while that of Qilu Lake was significantly affected by water distribution, vegetation distribution and distance factors, and that of Xingyun Lake is significantly affected by all factors. (4) The results of ridge regression relationship model showed that the fitting index R² of the estimated value of the model and the inversion value are all above 0.8, indicating that the fit degree and the accuracy test results are within an acceptable range. In brief, the surface temperature in the studied lake region presented a polarizing trend of high temperature and low temperature, and vegetation, architecture, water body and distance all had influence on the cold island effect of the lakes. Moreover, the temperature fitting method based on ridge regression analysis demonstrated its practical applicability.