On-line optimization of gas pipeline networks

Abstract

Compressor stations in gas transmission systems consume significant amounts of energy. Currently, operating policies for these stations in the face of varying demand are established by trial and error using simulators. The control of these systems is a challenging task due to the systems' large dimensions, their non-linearity, and the operational constraint present. In addition, the number of measurements is small and the system states are not known. In this paper a general on-line optimization scheme for the operation of gas pipeline networks is presented. It is built around the dynamic simulator GANESI. This simulator is used to predict the behavior of the network when it is subject to any operating policy; a successive quadratic programming optimizer is used to compute the optimal of such operating policies. In order to correct for modelling errors, demand forecast errors and errors in the initial conditions, the on-line measurements are used in a state estimator and the optimization is carried out over a moving horizon. The optimization scheme has been tested in simulations on networks of medium complexity involving up to seven compressor stations. Energy savings of up to 20% over conventional operating policies were found even in cases with 10% demand forecast error.