The aim in this article is to develop and investigate the performance of two numerical methods based on Physics-Informed Neural Networks (PINNs), the so-called Asymptotic-Preserving Neural Network (APNN) methods, for solving the multiscale uncertain Vlasov-Poisson-Fokker-Planck system in the high-field regimes. The second section starts with the presentation of the Vlasov-Poisson-Fokker-Planck (VPNN) system from which, taking the limit as \(\epsilon\to 0\), one obtains the high-field limit of the VPFP system: \[ \partial _t \rho - \nabla _x. (\rho\nabla _x \Phi) =0 \] \[ -\Delta_x \Phi _x = \rho - h(x) \] Further the micro-macro model is introduced and one shows the way high-field limit model can be derived from the micro-macro model. In the third section two formulations of the Asymptotic- Preserving Neural Networks (APNNs) methods are described. The one APNN method is based on the micro-macro decomposition for the VPFP system which means to use PINN to solve the micro-macro decomposition system, rather that the original one. The second APNN method is based on the mass conservation low and one shows that the applicability domain for this method is broader and involves cases with long time duration on non-equilibrium data. The performances of both methods are tested, compared and efficiency discussed on a set of very interesting problems and models. In the fourth section there is an extensive presentation of numerical experiments and results for following problems: the Landau Damping, the Bump-on-Tail Case, a one-dimensional Riemann Problem, the case of Mixing Regimes, the Gravitational Case, Uncertainty Quantification (UQ) problems. Some conclusions can be found in the fifth section.