Granular Decoupling Multivariable Systems Using the ePL Algorithm

This work introduces a novel methodology for decoupling multivariable systems using an evolving participatory learning algorithm. The decoupling is achieved by integrating compensation signals into the control signals generated by controllers designed for single-input, single-output (SISO) loops. Unlike traditional analytical approaches in the literature that typically require a known model of the controlled multivariable system, this proposed methodology does not depend on such a model. Instead, it learns the system’s coupling dynamics in real-time during operation, enabling adaptive control. This ability allows the approach to effectively manage parametric variations and structural changes in the system dynamics, which are often difficult for conventional methods that rely on a fixed process model. Furthermore, the methodology exhibits robustness when faced with parametric or structural changes in the controlled system, making it a promising solution for dynamic and unpredictable environments. The proposed method has been applied to a boiler model with four inputs and four outputs. The results indicate that the granular decoupling outperforms other state-of-the-art methods, especially when the system is subject to model inaccuracies or changes over time. The improved adaptability and performance highlight the approach’s potential to enhance the efficiency of control systems in practical applications.