In the Ring Laser Gyro Inertial Navigation System (RLG INS), the variation of working temperature can directly influence its measurement accuracy. In order to enhance the performance of the RLG INS, the operating temperature should be precisely controlled. According to thermal analysis on the accelerometer in the RLG INS, temperature control precision should be better than 0.01 to achieve ℃ 1×10^-5 m/s² output accuracy of the accelerometer. This paper adopt Back Propagation Neural Network algorithm to simulate the accelerometer and set up a precise temperature control system. Based on the simulated model and the control system, experiments are designed to validate the feasibility of temperature control theory. Meanwhile the performance of temperature control system is also verified under different environment conditions.