Monitoring and improving LP optimization with uncertain parameters

Linear Programming (LP) remains the workhorse of applied optimization, having open-loop process applications such as production planning, inventory scheduling and closed-loop applications such as Real-time optimization and the steady-state determination at each execution of a Model Predictive Controller (MPC). This paper presents new methods for monitoring performance (estimating the degradation due to uncertainty) and for improvement (reducing the uncertainty, when required, through economically optimal experiments) of closed-loop RTO systems.