Regresar

A Chance-Constrained Nonlinear Programming Approach for Equipment Design Under Uncertainty

Abstract:

In this work there are shown different strategies to cope uncertainty in large-scale chance-constrained nonlinear programs. We present the design of a flare system as a case study. The design of this system is influenced by several uncertain factors, such as the volume and composition of the waste flow stream to be combusted and the ambient conditions. These systems are currently designed based on typical historical values for waste fuel gases and ambient conditions. Consequently, an improperly designed flare can be susceptible to extreme events previously not experienced. Particularly, we use moment matching (MM) when the algebraic form of the moments and the quantile function of the chance constrained (CC) distribution is known, and for more general settings when the distribution cannot be pbkp_redicted we use the scenario approach (AS), the popular conditional value at risk (CVaR) and the recently proposed sigmoid value at risk (SigVaR). We demonstrate that the SigVaR approximation offers the best results and this approach overcome the conservative results of the AS and CVaR.