Different types of subsidence lakes formed by underground coal mining are severely polluted by mine-production wastewater, domestic sewage, and agricultural irrigation water. Microbial communities perform a crucial role in biogeochemical cycling processes and responses to natural and anthropogenic disturbances in lake waters. Therefore, it is important to investigate the characteristics of microbial community diversity in subsidence lakes for the utilization of water resources in mining areas. In this study, we collected water samples from open and closed subsidence lakes and explored the bacterial communities based on Illumina amplicon high-throughput sequencing. Results showed that bacterial diversity and community and functional composition in open and closed lakes differed greatly, and the bacterial diversity in open lakes was significantly higher than that in closed lakes (p < 0.001). Canonical correspondence analysis found that pH explained 31.73% and 39.98% of the variation in the bacterial community and functional composition in closed lakes, while NH₄⁺ and COD accounted for 29.73% and 26.02% of the changes in the community and functional composition in open lakes, respectively. Modified stochasticity ratios based on null-model analysis demonstrated that stochastic processes were the main factor affecting bacterial community assembly in both closed (MST = 0.58) and open lakes (MST = 0.63). In the closed-lake network, hgcI_clade was inferred to be a keystone species, while Lactococcus, Acinetobacter, Psychrobacter, and Chryseobacterium were detected as keystone species in the open-lake network. By way of discussion, we provide evidence regarding the bacterial community diversity, structure, co-occurrence patterns, and assembly processes in closed and open subsidence lakes. This study contributes to providing a reference for the utilization of subsidence water resources.