Chaplanov O. Neurodynamic forecasting models in control systems

The problem of development of methods for nonlinear dynamic stochastic plants modeling based on neural networks and chaos-dynamics technologies is considered. It is shown that neuroemulators are suitable for the reconstruction of chaotic and stochastic characteristics in real time. Neural network methods and architecture for Hurst exponent reconstruction in real time is derived. The neural network architecture and adjusting methods based on resonance filter in real time are derived. The fast and low computational complexity radial-basis neural networks models and learning methods are derived. Computer simulation of the developed neural network architecture and learning methods is carried out. Real-world problems of dynamic reconstruction of chaos and Hurst exponent computation with the application of the developed models and learning methods are solved.