A new innovation-weighted adaptive Kalman filtering algorithm is proposed to improve system performance by addressing the issues of poor stability of the internal calibration punctual system of micro-nano satellites and low punctuality accuracy when large changes occur due to the influence of ambient temperature. A model of frequency change of temperature-compensated crystal oscillator with temperature is established, the Kalman filter algorithm is used to filter out the input noise and realize the calibration of the crystal oscillator model parameters, the input field value is filtered out by new innovation-weighted technology, and the influence of system noise on the filtering results is reduced by adaptive technology. The experimental results show that the algorithm can achieve the convergence of model parameters in about 600s and can be adjusted in real time during the calibration period. The influence of input field value and system noise can be successfully controlled throughout the calibration process; when the ambient temperature changes, the punctual accuracy can reach 178 μs/day. The proposed algorithm effectively improves the stability and punctual accuracy of the system.