Adaptive PI Controller using Puma Optimization with Kernel Extreme Learning Machine for Dynamic Response Improvements of Brushless DC Motor

This paper presents a novel approach for controlling the speed of Brushless DC (BLDC) motors using an adaptive Proportional–Integral (PI) controller. Conventional PI controllers suffer from its performance when speed and load torque reference change due to their fixed parameter structure. To address these limitations, the proposed method integrates Hybrid Puma Optimization with Kernel Extreme Learning Machine (KELM) to enhance controller adaptability under varying conditions, including speed changes and load torque. In the proposed method, Puma Optimization is employed to optimally tune PI parameters under diverse operating conditions, generating high-quality training data. These optimized parameters are used to train a KELM model, enabling real-time adaptive adjustment of PI parameter based on reference speed and load torque variations. The effectiveness is validated through MATLAB/Simulink simulations and the results are compared with PI controllers tuned using Extreme Learning Machine (ELM) and Artificial Neural Network (ANN). Simulation results demonstrate that PI-KELM effectively adjusts to dynamic operating conditions, thereby improving the overall performance of the BLDC motor. The proposed method achieves superior dynamic performance with smallest overshoot, faster settling time, and improved damping behavior. PI-KELM significantly improves stability compared to the baseline with 25% improvement in settling time, 83.33% in overshoot, and 50% slower in rise time. Furthermore, the PI–KELM controller yields the lowest mean squared error (MSE) of 0.567 during training and significantly reduced ITAE, IAE, and ISE indices during testing. Compared to conventional scenarios, the proposed method exhibits superior dynamic response and robustness.