Statistical Performance and Stability Evaluation of Metaheuristic-Based PID Tuning for Flyback Converters

The Flyback Converter (FBC) is widely utilized in low-power applications. A Proportional Integral Derivative (PID) controller is typically employed to ensure output voltage regulation and stability. However, determining the optimal PID controller gains is challenging due to the bilinear nature of the FBC model, which classifies it as a non-convex optimization problem. Metaheuristic algorithms are often applied to optimize PID gains as their gradient-independent nature eliminates the need for complex analytical formulations and trial-and-error methods. They can provide an excellent workable approximate global solution. The global optimality of metaheuristics cannot be guaranteed as they may or may not provide a globally optimal solution, and there is no deterministic way to determine whether the final solution is global or not because of the inherent randomness involved in their structure. This inherent randomness in their algorithmic structure necessitates standard Statistical Performance Evaluation (SPE) to establish the superiority of one algorithm over another in finding controller gains. In this paper, the Improved Accelerated Particle Swarm Optimization (IAPSO) and Dragonfly (DF) algorithms are implemented for the first time to optimize PID gains for FBC. An independent samples t-test is conducted to underscore the importance of performing standard statistical tests while optimizing controller gains with metaheuristic algorithms. The results demonstrate excellent FBC response and stability for a given reference voltage and the statistical superiority of IAPSO over the DF algorithm in finding FBC PID gains.