A novel fractional-order neural network for model reduction of large-scale systems with fractional-order nonlinear structure

In this paper, a new method for reducing the order of nonlinear large-scale fractional-order systems is presented. The considered system has a nonlinear large-scale dynamic. The proposed method is developed by introducing a new fractional-based approach for neural network learning. According to the fractional-order modeling of the system, the structure of the neural network is selected as a recurrent neural network and new design and analysis are done on this network. In order to show the proposed method for model reduction has an acceptable error, a novel fractional-order stability analysis is used to derive the neural network weighting function. Moreover, it can be concluded that the proposed reducing method can preserve the main properties of the original system like a system's stability. Simulation examples are provided to show the effectiveness of the proposed method. Finally, the proposed method is compared with the existing methods and advantages of the proposed method are shown.