A key consideration of any engineering discipline is addressing uncertainty to develop technology, products, or designs that meet business needs. The modeling and design of natural gas systems is no different; system modeling contains many uncertainties that challenge the goal of accurate modeling for system operations and investments. In fact, system modeling likely has more uncertainties than many other engineering disciplines. These uncertainties range from physical properties such as pipeline roughness, the inaccuracies of data used by system models, to large uncertainties in estimating customer demands. Since the hydraulic design of gas systems is driven by maximum flow conditions, modeling peak customer demand during extreme cold design days is likely the most challenging and largest contributor to modeling error. In particular, local distribution and transmission systems with numerous small residential and commercial customers create uncertainties that are significantly more complex than the large custody transfer deliveries on interstate backbone lines (see Appendix for definitions of backbone, local transmission and distribution systems). Due to the rare occurrences of extreme cold design days there is no solution to completely eliminate the uncertainty in modeling system demand. This paper will briefly discuss various sources of modeling uncertainty on local distribution and transmission systems and then focus on customer demand uncertainty during extreme cold conditions, the causes of this uncertainty, and options available to address uncertainty when operating and designing these systems. The paper will then recommend sensitivity analyses as the best way to address demand uncertainty. Hydraulic results from sensitivity analyses on various distribution and transmission systems will be explored, significant differences in system response discussed, and resulting operating and design decisions will be reviewed. Finally, future plans to improve demand modeling and planning policies will be summarized.
Physical - Modeling assumptions for pipeline roughness, equipment pressure drop (valves, filters, stations), regulator droop, and other equipment performance. Model input data - Uncertainty in source data such as billing data, and mapping descrepencies. Errors in data used for model calibration; flow, pressure, temperature, and gas heat content measurement errors. Customer demand modeling - The rare occurrence of a cold design day results in a lack of actual design day demand data, thereby introducing significant modeling uncertainty. Demand uncertainty includes both the lack of daily demand vs. temperature data and hourly demand vs temperature data. Of the above four categories, customer demand modeling creates the highest uncertainty and largest source of error affecting system operations and design. Weather uncertainty is a given and operating decisions and system design must be made with weather forecast error in mind. Physical uncertainties, while significant, can be reduced to reasonable levels by the use of operational flow and pressure data and calibration of models so model data matches reasonably well with actual data at the moderate temperatures where data is available. Likewise, error in operational data must be accounted for but with proper measurement design and maintenance, error can be minimized and accurate representation of physical pressure drops vs. flows obtained.
