The paper presents the results of a study on the use of digital twins as an innovative tool in the field of professional training of operational personnel of nuclear power plants. The digital twin is considered as a virtual model of a physical power unit, functioning on the basis of real data and providing an accurate simulation of its operation, including standard and emergency modes. The expediency of introducing digital twins into the educational process as a means of improving the quality of professional training and the level of security at critical infrastructure facilities has been substantiated. A complex structural and functional scheme of interaction of a digital twin with an operator has been developed, which includes processing of simulation data, analysis of user reactions and adaptation of training scenarios. Particular attention is paid to the use of virtual and augmented reality technologies that create a highly realistic and safe learning environment. Separately, the stages of constructing an algorithm for training operators based on simulation data are considered: initial assessment of skills, personalization of the curriculum, use of simulations and automatic adjustment of the content of classes. The use of artificial intelligence allows you to dynamically adapt the educational process to the individual characteristics of each specialist. The results of the study demonstrate the effectiveness of digital twins in the formation of key professional competencies, increasing the stress resistance of operators and reducing the probability of errors in real operating conditions of power units of nuclear power plants.