Optimizing the natural gas supply mix of local distribution utilities

A large mixed-integer linear program (MILP) and a much smaller nonlinear programming (NLP) approximation of MILP, involving simulation and response surface estimation via regression analysis, are proposed to solve the problem of the optimal selection of natural gas supply contracts by local gas distribution utilities. Each potential supply source is characterized by several price and nonprice parameters. Weather variability is the basic stochastic factor that drives the demand for gas by various market segments. The model minimizes the total cost of gas supply and maket curtailment, and thus determines the size of the interruptible market. A numerical application of the methodology illustrates the excellent quality of the NLP approximation and the importance of the trade-offs between contract characteristics. A multi-temporal extension of the modeling methodology is outlined.