Design of Complex Systems under Uncertainty Using Process Simulators

In this work, a two-stage stochastic programming approach is implemented in a commercial simulator. A hybrid algorithm is proposed, where the first-stage decisions (existence of process units and their corresponding design parameters) are handled by a genetic algorithm, while the second-stage decisions (optimization of operational variables such as flows and temperatures) are optimized through the built-in optimization tool of Aspen Plus. In this way, a number of individuals (possible values of the first-stage variables) are defined, selected and combined through genetic operators, while the second stage variables are modified for each individual and different realizations of the uncertain parameters through a mathematical programming code (SQP) to minimize its expected cost. Given the complexity of the optimization problem under uncertainty, several strategies are proposed to minimize the computational requirements of the solution procedure. These strategies resulted in the reduction of up to 75% in CPU time for problems involving the optimization of complex separations systems. .