Plant-wide control: reformulation based on mixed-integer quadratic programming

LAUTARO BRACCIA; PABLO MARCHETTI; DAVID ZUMOFFEN

Buenos Aires

Congreso; 25to Congreso Argentino de Control Automático (AADECA 2016); 2016

Asociación Argentina de Control Automático (AADECA)

This work presents a new plant-wide control (PWC) problem formulation based on a mixed-integer quadratic programming (MIQP) model. The proposed representation maintains the linearity of the constraints and the convexity of the objective function, ensuring that the best integer solution is also a global optimum. Model solutions are obtained using a state-of-the-art solver, thus avoiding the implementation of global optimization algorithms and significantly improving the computational time. Results of a stochastic global search method such as genetic algorithms (GA) are compared against those obtained using the solver Cplex in the GAMS environment. The Tennessee Eastman Process is selected as case study.