Model Predictive Control of Natural Gas Pipeline Systems - a case for Constrained System Identification

摘要

Natural gas pipeline systems are not commonly discussed as target systems for Model Predictive Control (MPC). Over decades, complex, non-linear dynamic system models have been developed and trials to directly use these complex models as a part of a real-time (optimal) control or real-time optimization solution have been presented. Quite few are those approaches known from power systems and process industries: just pick the relevant dynamics and try to linarize, if possible. This paper describes how MPC can be applied on a natural gas pipeline system. The dynamic models required for MPC are assumed linear, which turns out to be a good approximation of reality. Natural Gas Pipeline Systems offer, because of their physical properties, an opportunity to use a'priori information in the model identification phase, which we make full use of. A demonstration example, though simplified, shows that significant improvement in pipeline operations can be achieved with MPC.