The global energy transition is imposing major challenges linked to the growing demand for electricity, the integration of intermittent renewable energies, and the reduction of CO₂ emissions. In this context, virtual power plants (VPPs) are emerging as an innovative solution, enabling the aggregation and intelligent management of distributed energy resources. Thanks to technological advances in communication, energy management, and algorithmic optimization, they contribute to the stabilization of networks and the transition to a more sustainable energy model. This article presents recent approaches based on machine learning applied to VPPs. It provides a classification of the main techniques, analyses their implementation process, and highlights the strategies adopted internationally to optimize their performance and flexibility.