Using a modified Sinc neural network to identify the chaotic systems with an application in wind speed forecasting

Continuous-time models for dynamic nonlinear systems offer greater reliability than their discrete-time counterparts. Discrete-time models can suffer from information loss and increased noise susceptibility. Chaotic and hyperchaotic systems pose significant challenges due to their unpredictable nature. These systems are prevalent in various fields, including weather, climate, finance, and biology. Artificial neural networks, inspired by the human nervous system, are effective in approximating complex nonlinear systems. A recent innovation, the Sinc neural network (SNN), leverages the properties of the Sinc function, which is smooth and oscillatory, making it suitable for approximation tasks. Despite limited research, SNNs have shown promising results in applications like speech recognition, human motion recognition, and fractional optimal control problems. This study introduces a modified Sinc neural network (MSNN) to enhance the performance of SNN in identifying continuous-time nonlinear systems. The MSNN employs a stable online training algorithm based on Lyapunov stability theory. It is utilized to identify several chaotic systems, including the Duffing-Van der Pol oscillator, the Lorenz system, and a financial hyperchaotic system. Additionally, the MSNN is used for forecasting wind speed, an important factor in renewable energy generation. Data from Khorramabad, Iran, is utilized for this purpose. The MSNN's simple structure and strong performance in identifying nonlinear systems and forecasting wind speed demonstrate its potential.

Using a modified Sinc neural network to identify the chaotic systems with an application in wind speed forecasting Keywords:

Chaotic system , hyperchaotic system , modified Sinc neural network , System identification , wind speed forecasting