Most simple, but also the most simplified approach to stellar winds is its spherically-symmetric (1D) and time-independent smooth-wind modelling. Such a modelling enables to explain some main features of wind-modified stellar spectra and to estimate the mean parameters of stellar wind. It shows also which spectral features demand more detailed and more complicated structural treatment. Modelling of such winds can be made based on different variants of input physics. The simpliest way is to specify the dependence of stellar wind velocity on distance and the mass loss rate. In addition, also some assumption about the temperature run in the stellar wind is to be made. Calculating thereafter the model spectra and comparing them with the observed stellar spectra the main shortcomings of the approach become evident.

Correct treatment of stellar wind demands its dynamical modelling taking adequately into account its interaction with radiative flux and describing wind generation. Main drive is due to flux in spectral lines. To elaborate dynamical picture of stellar wind, different approximations have been used. The equation used for such a study is the equation of momentum conservation applied to interacting atomic particles of plasma and photon field in external fields. A general feature of time-independent treatment is that one must solve the boundary value problem with parameter values, which yield fit in the critical point (on the critical sphere) where uncorrect values give singularities.

We propose a new method to solve the problem in dynamically self-consistent manner.We replace the time-independent boundary value problem with a time-dependent initial value problem. Avoiding the singularities on the critical sphere surface we introduce an initial wind with large velocity gradients but relatively low mass-loss rate. In such case the mass loss rate starts to approach to the time-independant regime starting from the inner, i.e. the subsonic region. The mass loss in it can be treated as leakage to a rarefied region. For wide range of initial models the final result for a star with given parameter values approaches asymptotically to self-consistent smooth wind.

The same holds for plasma diagnostics in stellar winds. Also for the study we can start from the time-dependent formulae for population and depopulation of states and follow the relaxational changes in them. The correct temperature run in stellar wind can be obtained starting from the LTE populations both of excited and ionization states and following the subsequent relaxational changes of these rates. The NLTE overionization rates in quiescent stellar winds are determined primarily by the non-Planckian radiation distribution in the corresponding continua. The temperature in it is determined by the law of energy conservation applied to ionization and recombination processes.

Diagnostics of stellar plasma is aimed to solve the reversed problem to modelling, i.e. to find the run of stellar wind parameters with distance from the observed features of wind modified stellar spectra. This ill-posed task can be solved only approximately. Some results obtained for hot star winds will be analysed. Some peculiarities of stellar spectra which cannot be solved in the framework of smooth wind will be described.