Training Generative Models for the Task of Object Description in A Strict Format: Analysis of the Efficiency of Their Architectures

In the modern world, digitalization is increasingly penetrating into people's daily lives and covers a wide variety of areas of human activity. Electric power industry is no exception. The introduction of SCADA systems has increased the observability of processes occurring in the network and increased the efficiency and speed of decision-making aimed at maintaining a stable operating mode of the EPS. The next step in the development of digital electric power industry is digital twins, which could be integrated into digital services for their visual display or, for example, for conducting processes of modeling the operating mode of the EPS. The introduction of artificial intelligence (AI) is also gaining popularity. More and more advanced AI models are constantly emerging, capable of solving increasingly complex problems. The introduction of generative artificial intelligence can significantly speed up the process of building digital models. This article provides a comparative review of various generative AI models that are further trained to generate descriptions of electric power substations in a strictly formalized form that can be further computer processed.