Date of Award

Fall 2004

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Civil, Construction, and Environmental Engineering

First Advisor

Brown, Ronald H.

Second Advisor

Richie, James

Third Advisor

Povinelli, Richard

Abstract

The focus of this work is on the selection of inputs for combining daily natural gas demand forecast models. One main problem of selecting the inputs is that each utility has different characteristics and sensitivity to inputs. Finding a set of inputs that works well on every utility is desired; however, it is not an easy task due to the diversity of the utilities. The contributions of this research are the development to reduce the multicollinearity problem in the current linear regression model of GASDAYtm, the use of the weekend and/or weekday indicators to help simulate the weekly period of the daily gas demand, and the use of more error terms as exogenous inputs to improve the performance of the forecast model. There are six new forecasting models developed in this study. These new models are simulated on six utilities with different characteristics. All the new forecasting models developed in this study performed well; furthermore, one of them is shown to considerably improve the performance of forecasting daily natural gas demand.

Share

COinS

Restricted Access Item

Having trouble?