Forecasting Natural Gas Design Day Demand from Historical Monthly Data
- Ronald H. Brown (Marquette University) | Paul E. Kaefer (Marquette University) | Calvin R. Jay (Marquette University) | Steven R. Vitullo (Johnson Controls Inc.)
- Document ID
- Pipeline Simulation Interest Group
- PSIG Annual Meeting, 6-9 May, Baltimore, Maryland, USA
- Publication Date
- Document Type
- Conference Paper
- 2014. Pipeline Simulation Interest Group
- 0 in the last 30 days
- 103 since 2007
- Show more detail
In this paper we discuss methods and limitations of building daily natural gas demand models from historical monthly consumption data for the purpose of predicting demand on design day conditions. (Here we define the design day as the bitter cold day that produces the highest demand the system is designed to handle.) Effective gas system planning for design day operations and design day design requires accurate design day demand forecasts. For regional segments of the gas system, often only cumulative demands over billing cycles is available and may not include periods with extreme cold weather events. This paper provides an overview of new methods for building design day forecasting models based on this coarse data.
We discuss building daily demand models using monthly consumption data, including what inputs can be used so that model coefficients are identifiable. Significantly more accurate models are built using two temperature inputs instead of a single temperature input. This approach fits the data much better on extreme cold days showing a higher design day demand. We also discuss observations on recent multiday near design day events showing higher than modeled demand on the peak day, known as the heck-with-it-hook, which impacts confidence levels.
Results are presented comparing model forecasts made with models parameterized on monthly data and on daily data. These test results include models built with no near design day conditions in the training data set but evaluated on testing set data with such data included. Results of forecasts from various gas systems are discussed and the use of proxy data from these gas systems for use on other gas systems.
|File Size||272 KB||Number of Pages||15|