Polarization control can effectively recover the polarization state shifted from the polarization phase drift during the fiber transmission, which is a key technology in quantum communication, optical fiber sensing, and coherent optical communication. In this paper, we propose a polarization control algorithm based on adaptive moment estimation (Adam) and establish a polarization control system model. Based on this model, the Adam polarization control algorithm is simulated and compared with classical stochastic gradient descent (SGD) algorithm, at the same time, the influences of control precision and noise amplitude on polarization control effect are analyzed. Simulation results show that the global optimal polarization state can be rapidly obtained by employing the proposed method. When the attenuation rate is 0.03, the noise amplitude is 0.003, and the number of polarization iterations is 53, the control precision is up to 10 ^-4. Compared with the SGD algorithm, the average iterative steps are reduced by 23% and the maximum control accuracy can be improved by 1--2 orders of magnitude. The feasibility of Adam algorithm is verified by field programmable gate array. The proposed algorithm can quickly and stably compensate the polarization change in the channel, and the polarization control time can be effectively shortened by optimizing the value of attenuation rate.