The temperature affects the performance of the accelerometer, and the main source of error is caused by the thermal stress generated by quartz materials and packaging processes. In response to the temperature drift phenomenonr, this article introduces the design and compensation method of the low-temperature drift structure of the integrated quartz resonant accelerometer. Firstly, the first-order temperature coefficient is eliminated by a symmetrical differential structure, and the process optimization is carried out to reduce the thermal stress generated by the packaging process on the resonator. Secondly, the random forest fitting algorithm is used to establish a temperature model to compensate for the temperature drift of the accelerometer. The temperature drift test of the accelerometer prototype is carried out within the temperature range of -20~80 ℃, and the results show that the bias stability of the prototype has been improved by an order of magnitude after process optimization and compensation.