Abstract: The influences of deposition and diagenesis result in the complexity and diversity of the carbonate reservoirs of the fifth section of the Lower Ordovician Majiagou formation in Sudong 41-33 block, Sulige gas field. Therefore, lithology recognition is a key step in the process of reservoir evaluation. Analysis of log and logging data shows that 7 types of reservoirs are mainly developed in the fifth section of Majiagou formation such as limestone, dolomite, dolomitic limestone, calcite dolomite, argillaceous dolomite, argillaceous limestone and mudstone. Based on the analysis of sensitivity parameters, 4 kinds of parameters of GR, density, PEF and CNL which could reflect reservoir lithologies are selected, and Na-ve Bayesian method constrained by Markov chain probability method is used to perform multi-parameter comprehensive interpretation. The results of analysis and comparison of logging and log data show that the accuracy of lithology recognition reaches 85.34%. Compared with the traditional Na-ve Bayesian method, the accuracy of lithology recognition could be improved by 12.39%. High accuracy model for logging lithology interpretation of carbonate reservoirs can be obtained by using this method. The method is an effective way for complex lithology identification