Dynamic modeling and linear model predictive control of gas pipeline networks

A linear model predictive control (LMPC) strategy is developed for large-scale gas pipeline networks. A nonlinear dynamic model of a representative pipeline is derived from mass balances and the Virial equation of state. Because the full-order model is ill- conditioned, reduced-order models are constructed using time-scale decomposition arguments. The ®rst reduced-order model is used to represent the plant in closed-loop simulations. The dimension of this model is reduced further to obtain the linear model used for LMPC design. The LMPC controller is formulated to regulate certain pipeline pressures by manipulating production set- points of cryogenic air separation plants. Both input and output variables are subject to operational constraints. Three methods for handling output constraint infeasibilities are investigated.