کنفرانس بین المللی ماشین و محرکه های الکتریکی

صفحه اصلی / The 5th International Conference on Electrical Machines and Drives

Fuzzy-Assisted GA and PSO Tuning of PI Controllers for Induction Motor Speed Regulation

نویسندگان :

Raouf Sirjani¹

1- دانشگاه ازاد اسلامی واحد مشهد

کلمات کلیدی :

Induction motor،PI tuning،speed control،(particle swarm optimization (PSO،(genetic algorithm (GA،fuzzy logic،Simulink

چکیده :

Abstract— This study focuses on improving the automatic tuning of a discrete PI speed controller used in induction motor (IM) drives. To explore different optimization strategies, two well-known meta-heuristic approaches—Particle Swarm Optimization (PSO) and Genetic Algorithm (GA)—are applied and compared under identical operating conditions. Alongside these methods, a simple fuzzy supervisor with a single input is introduced to support the conventional PI controller. The performance evaluation relies on a cost function that considers integral absolute error (IAE), steady-state error (SSE), and overshoot (OS), aiming to balance fast response and steady-state accuracy. The system model was implemented in MATLAB/Simulink with a fixed time step of 5×10⁻⁵ s over a 10- second experiment, including a step reference of 1500 rpm and a step load torque. According to the results, PSO tuning yields Kp = 0.1459 and Ki = 41.090, leading to OS ≈ 5.15%, SSE ≈ 16.09 rpm, and IAE ≈ 2.95×10⁴ rpm·s. The GA approach converges to P = 0.5964 and I = 0.7122, achieving the same overshoot and steady-state error but with a significantly lower IAE of about 4.48×10² rpm·s, which represents roughly a 98.5% reduction. Furthermore, the fuzzy supervisor (using triangular 7-MFs and a diagonal rule base) improves the static accuracy, reducing SSE to about 6.56 rpm and maintaining OS around 5.67%. Overall, the findings show that meta-heuristic PI tuning can be both effective and consistent. In this particular case, GA achieved the lowest IAE, while the lightweight fuzzy layer enhanced steady-state performance without adding significant computational complexity. All key parameters and plots are provided to make the study easy to replicate.