Relevance. The assessment of anthropogenic pollutant loads during scheduled and emergency emissions into atmosphere is an important problem to plan new or to exploit the present energy objects of the energy infrastructure. The experimental research in the area is too expensive and does not always give the required accuracy especially under adverse weather conditions. Although the modern mathematical models are able to provide qualitative predictions caused by the effects of air emissions but quantitative estimations, obtained by these models, give an adequate accuracy only for neutrally stratified atmosphere boundary layer. It is required to develop a new more advanced and cost-effective approaches for a more realistic conditions. The main aim of the study is to develop the efficient methods of the atmospheric dispersion modeling to provide information about the environmental review and analysis of the ecological conditions near objects of the energy infrastructure. The method used in the study. The numerical stochastic Lagrangian modeling of the passive tracer in the convective planetary boundary layer (PBL) was used based on the random walk process and Langevin model of the turbulent dispersion. The statistical structure of turbulence is described by the probability density function (PDF) of the vertical velocity fluctuations, which is recovered by the calculated statistical moments of the vertical velocity fluctuations. Four models of the PDF reconstruction are studied. The results have been tested against ones from water-tank experiments (Willis and Deardorff) and from the observational data in convective PBL. The results. Numerical calculations show that the implemented models have the asymmetric distribution of the PBL vertical velocity amplitudes: the fast downward and slow upward flows and may be further used to develop the method of the Lagrangian modeling of the air pollutant dispersion in the convective PBL.