The research is relevant since there is a necessity to solve identification problems of non-stationary signals of control and communication systems in conditions of uncertainty. Research objective is the development of models and algorithms of non-stationary signal identification with variables, time-dependent parameters and with account of additional a-priory information. Integrated system of mathematical models of researched non-stationary signals and analogue objects models representing additional a-priory information were used as the basis of developed of identification algorithms. Solving identification problems passed in optimizing design, where quality indicators of signal models, comparable models and solutions methods of optimizing problems were used. As the result, models and algorithms of non-stationary signal identification system with consideration of additional a-priory information are developed allowing solving a wide range of filtration, data smoothing, local signal change detection, estimating of their parameters in conditions of uncertainty. The example of solving diagnostics signal problems of turbo generator magnetic stray field showed that the proposed models and identification algorithms with consideration of additional a-priory information allow reliable detect local signal changes from the magnetic stray field sensor that were caused by the rotor winding short circuit.