Comparative performance analysis of fractional-order nonlinear PID controller for complex surge tank system: tuning through machine learning control approach

In this research article, a fractional-order nonlinear Proportional plus Integral plus Derivative (FONPID) controller is incorporated into a complex surge tank system where the gains of the controller are tuned through a machine learning control approach to control the nonlinear variations in level setpoints. Obtaining a proper output response from a non-linear system is challenging and demanding for researchers. In this case, nonlinear Proportional plus Integral plus Derivative (NPID) and conventional Proportional plus Integral plus Derivative (PID) controllers are not sufficient for obtaining desired output robustness in the system performances. Hence, to fulfill the need for an adaptive controller for a spherical surge tank system, FONPID can be a better choice. The machine learning control is applied to the gains of the FONPID controller with the Cuckoo Search Optimization Algorithm (CSA), a swarm-intelligence algorithm mostly known for its levy flights and searching pattern for best quality eggs. The whole idea of using machine learning control is to tune the gains to make the controller adaptive towards parametric variation and uncertainties. The machine learning control uses the Integral of Absolute Error (IAE) performance index criteria as the minimum objective function of CSA for tuning of gain constraints of the controllers. The proposed FONPID controller is then compared with NPID and conventional PID controllers to stabilize level setpoint variations. The results demonstrate that the FONPID controller gives better, robust, and optimum results over NPID and PID controllers. In comparison to NPID and PID controllers, the FONPID controller performs significantly better, with gains ranging from 11.68% to 215.31% across various operational modes and system parametric variations.