A detailed mathematical model of compressor stations and pipes is essential for optimizing the performance of the gas pipeline system. Most of the available literature on compressor station modeling is based on isothermal solutions for pipe flow, which is inadequate for our purposes. In the present work, the pipe flow is treated as nonisothermal unsteady one-dimensional compressible flow. This is accomplished by treating the compressibility factor as a Junction of pressure and temperature, and the friction factor as a Junction of Reynolds number. The solution method is the fully implicit finite difference method that provides solution stability, even for relatively large time steps. The compressors within the compressor station are modeled using centrifugal compressor map-based polynomial equations. This modeling technique permits the designation of different models of compressors in the compressor station. The method can be easily extended to include other types of compressors. Using this mathematical model as a basis, a nonlinear programing problem (NLP) is set up wherein the design variables are the compressor speeds and the objective function to be minimized is the total fuel consumption. The minimum acceptable throughput is imposed as a constraint. This NLP is solved numerically by a sequential unconstrained minimization technique, using the mathematical model of the system for the required function evaluations. The results show that this approach is very effective in reducing fuel consumption. An application of this methodology for selecting the number of compressors to be shut down for the most fuel-efficient operation is also presented. Our results further indicate that station-level optimization produces results comparable to those obtained by network-level optimization. This is very significant because it implies that the optimization can be done locally at the station level, which is computationally much easier.