A simulation testbed for SDR transmitters and a workstation designed for training a Graph Neural Network (GNN) in laboratory conditions has been developed, intended for use prior to the modernization of single-frequency networks (SFN) in broadcasting areas. The testbed is designed for the experimental evaluation of technical requirements for the equipment of regional multiplexing centers, the capital costs for project implementation [1], and the refinement of the software for its main components. Automated optimization of SDR transmitters, combining the reliability of a deterministic algorithm based on the DVB-T2 standard with the speed of adapting transmitter operating modes to changing broadcasting conditions, is ensured by the capabilities of the Graph Neural Network (GNN).