Thesis Implementación de control predictivo en Hardware para aplicación en planta de biogás
No Thumbnail Available
Date
2026-08
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Electrónica
Departament
Campus
Campus Casa Central Valparaíso
Abstract
Esta memoria se centra en la implementación en hardware de una arquitectura de Control Predictivo No Lineal (NMPC) con estimación de estados mediante un Filtro de Kalman Extendido (EKF), orientada a seguir una referencia de producción de metano en un proceso de digestión anaeróbica (DA). La planta de digestión anaeróbica se implementa usando el modelo ADM1, que se emplea como planta virtual junto al modelo reducido AM2 como predictor, ambos modelos ampliamente usados en la literatura. El EKF estima una perturbación que compensa el sesgo entre ambos modelos, otorgando acción integral libre de offset. La planta se simula en MATLAB/Simulink y el controlador se ejecuta en Python sobre una Raspberry Pi, comunicados mediante un servidor Modbus TCP.
This thesis focuses on the hardware implementation of a Nonlinear Model Predictive Control (NMPC) architecture with state estimation via an Extended Kalman Filter (EKF), aimed at tracking a methane production reference in an anaerobic digestion (AD) process. The anaerobic digestion plant is implemented using the ADM1 model, which is used as the virtual plant, together with the reduced AM2 model as the predictor, both models widely used in the literature. The EKF also estimates a disturbance that compensates for the bias between both models, giving the scheme an offset-free integral action. The plant is simulated in MATLAB/Simulink, and the controller runs in Python on a Raspberry Pi, communicating through a Modbus TCP server.
This thesis focuses on the hardware implementation of a Nonlinear Model Predictive Control (NMPC) architecture with state estimation via an Extended Kalman Filter (EKF), aimed at tracking a methane production reference in an anaerobic digestion (AD) process. The anaerobic digestion plant is implemented using the ADM1 model, which is used as the virtual plant, together with the reduced AM2 model as the predictor, both models widely used in the literature. The EKF also estimates a disturbance that compensates for the bias between both models, giving the scheme an offset-free integral action. The plant is simulated in MATLAB/Simulink, and the controller runs in Python on a Raspberry Pi, communicating through a Modbus TCP server.
Description
Keywords
Digestión anaeróbica, Observadores, Control predictivo no lineal, Filtro de Kalman extendido, Producción de metano
