Adaptive convex relaxations for Gas Pipeline Network Optimization

The growing use of natural gas for electricity generation has created significant changes in the loads experienced by pipeline systems used for gas transport. This trend compels a fundamental re-examination of natural gas markets to develop pricing that reflects the physical and engineering capacity of pipeline networks. Recent advances in optimization techniques are promising for enabling pipeline optimization to address these needs. In this paper, we present an adaptive partitioning method in combination with optimality-based bound tightening approaches to strengthen standard convex relaxations for a class of steady-state pipeline system optimization problems, mathematically formulated as mixed-integer nonlinear programs (MINLP). Computation time and solution accuracy for objective functions that maximize compressor efficiency and economic welfare, respectively, on meshed gas networks are evaluated and compared to outcomes obtained using an interior point method. The method rapidly solves such problems to near global-optimum solutions. Efficiency and scalability of the technique makes it promising for extension to large-scale problems of dynamic gas-electric co-optimization.