Aiming at problems of unrevealing potential error source, hardly separating specific random error and long data collection time of common random error identification methods, Allan variance analysis method was used to identify random error of MEMS gyro in coal mine underground. Principle of Allan variance analysis method was introduced. Allan variance analysis method was used to process measured data of MEMS gyro, Allan standard deviation curves were given, and main random error coefficients of MEMS gyro were obtained by least square fitting. The experiment results verify validity of the Allan variance analysis method for random error identification of MEMS gyro.