Gaussian RBFNN method for solving FPK and BK equations in stochastic dynamical system with FOPID controller

Abstract

Solving the Fokker–Planck–Kolmogorov (FPK) equation and the Backward-Kolmogorov (BK) equation is a crucial task to obtain the transient response of stochastic dynamical systems. Fractional order PID (FOPID) is a new efficient controller to change the system response to be the expected one. Therefore, in this paper, the Gaussian Radial Basis Functions Neural Network (RBFNN) is proposed to solve FPK and BK equations, to obtain the transient probability density function and the reliability function for a generalized Van der Pol system under a FOPID controller. The values of the different fractional orders are analyzed to discuss the performance of the FOPID controller. A data collection strategy is adopted to deal with the associated boundary conditions by way of a one-time Monte-Carlo simulation and uniform distribution in our Gaussian RBFNN method. The advantage of this method is that the solution process of FPK and BK equations is converted into solving algebraic equations. Numerical results with regard to the transient system response prove that the Gaussian RBFNN is efficient and accurate in getting the solutions of FPK and BK equations. The order of the fractional integration and the fractional derivative are critical parameters to control the system response. Moreover, we conclude that the fractional order parameters in a FOPID controller can indeed enhance the system’s response to a certain extent and lead to bifurcation.