Abstract: The article is devoted to identifying patterns in the dynamics of decarbonization of national economies based on the methods of self-organizing maps (SOM). The study consists in revealing a typology of countries of the world in the context of energy generation and consumption, as well as analyzing 45 changes in the structure of their energy profiles in 2013–2022. An empirical model for assessing the decarbonization potential using self-organizing maps was developed and tested. The research methodology includes six stages. At the first stage, the authors collected and systematized data from open sources, in particular the World Bank and the EnerData yearbook, which include macroeconomic, energy and environmental indicators. The second stage involved pre-processing of data – eliminating omissions, identifying outliers and converting them into a single format "country-year-indicator-value". At the third stage, a coordinated multidimensional data set with 14 variables was formed (ensuring the absence of multicollinearity), and data normalization was performed. In the fourth stage, a SOM with hexagonal topology was built and trained, the optimal structure of which was defined using quantization and topographic error indicators. The fifth stage is devoted ti the clustering of neuron weight vectors using an agglomerative approach, where the optimal number of clusters was determined using the Silhouette Score, Davies-Bouldin, and Calinski-Harabasz indices. In the final stage, the results were interpreted, countries were typified by energy profiles (leaders, industrial, resource-oriented, etc.) and the dynamics of their movement between clusters over time were analyzed. The results obtained allowed us to identify structural differences in decarbonization models and identify key factors that ensured the success of the countryʼs energy transition at the global level.