Comparative Analysis of the Performance of PID and NAMA-L2 Controllers in Fluid Temperature Control

Resumen

The research addresses temperature control in a thermal system where a classical PID controller presents difficulties due to the time required to reach steady state and limited performance during the transient regime, evidenced by overshoot. These limitations affect the efficient control of systems with nonlinear dynamics and variations in operating conditions, motivating the search for more advanced control strategies. In this study, an intelligent control scheme based on artificial neural networks (ANN) with a NARMA-L2 structure is evaluated, capable of approximating the inverse model of the system and generating appropriate control signals. The thermal system was simulated in MATLAB/Simulink using experimental temperature data and a control signal associated with the opening percentage of a fluid inlet valve. Experimental tests were conducted on a laboratory module for system identification and controller training, allowing comparison with a conventional PID Controller. During training, a mean squared error of 1.799×10⁻⁵ and a correlation value of 0.999 were obtained. The results show a settling time of approximately 50 s, a steady-state error close to zero, and no overshoot, confirming that the NARMA-L2 controller is an effective alternative to the classical PID Controller.