ABSTRACT

Planning for the efficient utilization of pipeline resources requires the accurate estimation of demands placed on the system. Regression analysis provides a convenient procedure for creating an expression with which to estimate the loads. Regression functions involving polynomials and multiple variables can handle a vast number of considerations. With the addition of step-wise regression, the model can take on a very complex mixture of considerations. Systematic errors can be reduced by employing a technique such as exponential smoothing. The most exacting function will still have some uncertainty associated with it. Developing an algebra that will allow operations such as addition, subtraction, multiplication and division of stochastic parameters can provide a more complete handling of many facets of the planning mission than just being able to analyze measured data. A method of moments allows for the first three operators. The remaining operator, division, can be handled via a reciprocal transformation of the stochastic divisor. Weather - sens.d w. e Id& Planning daily operations at Colorado Interstate Gas (CIG) begins with a detailed weather forecast for the front range of the Rocky Mountains. Centrally located on its 3000~mile natural gas transmission system (see Exhibit l), deliveries to front range distribution companies are almost entirely weather-dependent. Summer loads on this portion of the system are low and stable but heating season loads are extremely volatile on both an hourly and daily basis. Since GIG's predominant flow direction is from northwest to southeast, virtually every aspect of transmission system operation must be considered with respect to weather-dependent loads. Weather-dependent loads account for approximately 50 percent of GIG's peak throughput of 2 BCF/day. Peak loads are typically characterized by relatively small hourly variation, as residential and light commercial customers' furnaces operate near their fbll-day capacity. Operational concerns focus primarily upon maintaining peak deliverability fkom compression and storage facilities, ensuring receipt points are performing up to nomination and ensuring delivery points do not exceed nominations. Daily loads can vary as much as 600 MMcf from one day to the next as shown in Exhibit 2. Large changes in ambient temperature drive these extreme variations in daily loads. Small industrial base loads contribute little toward dampening daily load volatility. Meeting such demand is primarily a matter of adjusting storage injection and withdrawal rates and to a lesser extent varying line pack. Temperature variations and behavioral patterns of residential and light commercial customers produce large hourly swings in gas demand. Elevations in excess of 5000 feet and a high percentage of clear days combine to produce large diurnal temperature variations. During the heating season, the difference between average maximum and average minimum temperatures in Denver is 27 degrees. In September and October the difference is 30 degrees. Exhibit 3 shows hourly temperature variations experienced in Denver during the 1996-97 heating season. Heating season hourly load profiles are characterized by two peaks as shown in Exhibit 4. The first and largest peak occurs between 7:00 a.m. and 9:00 a.m.

This content is only available via PDF.
You can access this article if you purchase or spend a download.