This paper presents the development of the Generative Adversarial Network (GAN) model to study the properties of direct photons produced in proton-proton collisions. The study aims to extend the capabilities of our model (developed previously) by introducing the initial collision energy parameter √sNN in the range from 9 to 27 GeV. The model has been trained on the data generated using PYTHIA8 and tested both at training energies and at intermediate energy values. Special attention was paid to the ability of the model to preserve kinematic dependencies and to reproduce the double longitudinal spin asymmetry ALL. The results proved the possibility of using the GAN to interpolate the characteristics of direct photons in terms of collision energy.