Due to a number of technological and economic changes in the pipeline industry, interest in optimization software is growing rapidly. Recent governmental regulations and widespread use and acceptance of modeling software in the industry have given rise to applications of optimization in new, existing, and changing pipelines. The goal of a pipeline optimizer is to be able to model a variety of pipeline configurations robustly and efficiently and to ensure feasible and optimal operating strategies according to a specified objective and the demands and resources of the pipeline system. The method should be flexible enough to optimize a number of different types of objectives using the same algorithm. In this paper, we shall discuss the application of the Generalized Reduced Gradient (GRG) method to the problem of steady-state optimization of the operation of a gas pipeline network for several different objectives. We shall present the problem to be solved, the method used to solve it, and the strengths and weaknesses of the method when applied to the problem. Several different criteria for optimality are considered including minimizing fuel consumption by the compressors and maximizing throughput. The optimal solution must satisfy all the constraints imposed by the limitations of the pipeline and the compressors' operating ranges. In mathematical terms, this is a nonlinear constrained optimization problem with some of its decision variables discrete. The GRG method is a state-of-the-art method for problems in which the objective value depends smoothly on continuous decision variables. The challenge in applying it to the pipeline optimization problem lies in finding a way to handle the discrete aspects of the problem. An evaluation of the method is performed by testing on three different pipeline configurations. Results are presented for a range of flow rates for each configuration to demonstrate the amount of savings to be expected from the optimizer and CPU execution times required to run the program.
The interest and need for pipeline optimization software has been getting stronger over the past five years. This is the result of a number of factors. First, many pipeline companies have seen the success of using transient pipeline models only over the past 10 years. Despite previous reservations concerning the validity of transient modeling among pipeline companies, many of them have been convinced that the technology is sound and today many have successful operating systems in the field. The belief in the validity of optimization programs is probably at the same place that confidence in transient models was five years ago. Today, however, there is probably less of a hurdle to leap in convincing pipeline companies that optimization software is a practical tool. Second, there is high activity in the field in terms of papers being written and systems currently being worked on.