1 Abstract:

This document explains the use, benefits, and components of a Gas Load Forecasting System (LFS):

  • Describes how the LFS is used within the Integrated Gas Management System (IGMS)

  • Provides an overview of Artificial Neural Networks

  • Provides an overview of gas load forecasting techniques

  • Provides practical experiences, findings, and recommendations of the authors

2 Introduction

Williams Gas Pipeline (WGP) is an operating group of The Williams Companies, Inc. It is principally involved in the interstate transportation and storage of natural gas in the United States. WGP operates five pipeline systems, consisting of: Northwest Pipeline; Kern River systems; Texas Gas Transmission; Transcontinental Gas Pipe Line (Transco); and Williams Gas Pipeline-Central (Central). The combined WGP pipeline network has more than 27,300 miles of pipeline and is among the nation's largest-volume transporters of natural gas. WGP has recently embarked on a series of twenty-one projects to enhance its information technology (IT) infrastructure. One of the IT systems being implemented is the Integrated Gas Measurement Management System (IGMS). IGMS is an integrated suite of software applications that provides operations analysis and decision support functions to WGP's three Operations Control centers. IGMS is being implemented with the assistance of Energy Solutions International, a Houston-based pipeline simulation company, which provides software development and systems integration services on the project One of the key components of IGMS is the Gas Load Forecasting System (LFS). The LFS is an artificial neural network-based forecasting application that provides short-term (1 to 5 day) gas load forecasts down to the meter level. LFS generates hourly or daily forecasts depending upon the type of historical data available to it. The LFS stores load forecasts in a centralized Oracle database allowing the data to be shared among different applications within IGMS that require the forecasts. The first implementation of the LFS has just been completed at WGP. The system is currently undergoing testing and will be deployed at two WGP locations by the time this paper is presented. As with many IT system implementations, the process has been a challenge. Nevertheless, the authors have learned much about gas load forecasting and have gained practical experience in making the system actually work and provide meaningful results. This paper starts by describing how the LFS is used within IGMS. Next, an overview of artificial neural networks is provided, followed by a discussion of gas load forecasting techniques. Finally, the practical experiences, findings and recommendations of the authors are presented. Within IGMS, there are three principal uses for load forecast data. Two of the uses are as input to other modules, basically serving as data feedstock. These modules are the Real Time Model (RTM) and the Predictive Model (PM). The third use of load forecast data is for advisories to Operations Control personnel and casual users throughout the company.

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