Recent investigations of the microbial communities in an oil reservoir using culture-independent molecular techniques has indicated that this environment contains many uncultured assemblages, and that the cultured bacteria may represent very minor components of the true microbial diversity. In this study, 16S rDNA based PCR amplification and denaturing gradient gel electrophoresis (PCR-DGGE) was used to monitor the microbial diversity in an injection well (12#9-11) and two related production wells (12#9-9S, 13#11-8) in the KELAMAYI oilfield (Xin Jiang Autonomous Region). The DGGE gel profiles showed significant difference between the microbial community structure of injection and production wells. The bacterial species richness in the injection well was higher than those of the production ones. Unweighted Pair Group Method Clustering (UPGMC) showed the similarity between the injection well and two production wells was 30% and 20% respectively, whereas the similarity between the two production wells was 54%. In addition, the dominant bands in the DGGE gel were excised and analyzed by nucleotide sequence analysis. This analysis of 10 excised bands showed that the dominant microorganisms in all samples had high sequence identity with rRNA genes from uncultured bacteria in the GenBank database (NCBI). Nine sequences of excised band were affiliated with α, γ, δ, ε-Proteobacteria, and one was related to Bacteroides. Two bands exhibited 100% identity to sequences of Pseudomonas stutzeri and Aminobacter sp. cox, respectively. Three bands showed relatively low identity (92% to 93%) to cultured bacteria in the GenBank database, indicating possible new taxa. This work analyzed the composition of the microbial community in KELAMAYI oilfield, which provides important data which may facilitate the isolation of new bacteria from this special underground environment. It also provides useful information for understanding the mechanism of MEOR and applying MEOR in practical oil production in KELAMAYI oilfield.