Date of Award

Fall 1999

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Brown, Ronald H.

Second Advisor

Heinen, James A.

Third Advisor

Garside, Jeffrey J.

Abstract

Existing natural gas load estimation models are based on gas operating areas that are small enough so that the necessary weather information comes from one to three weather sites. Larger operating areas need to use weather information from many weather sites to adequately model the gas consumption for the operating area. One way to handle this problem is to consolidate the weather data from the many weather sites of the large operating area so that it appears to the estimation model that there are very few weather sites. Four approaches to this problem are evaluated using a linear regression (LR)-based estimation model and an artificial neural network (ANN)-based model. The first approach involves dividing the large operating area into smaller ones; the second involves devising a weighted average to combine the many weather sites into just one; the third combines the first two approaches; the fourth involves some combination of three weather sites to use as inputs to a single estimation model for the large operating area. The best of the four approaches yields lower estimation error than the other approaches, and is easy to apply to large operating areas other than the operating area used as a test case.

Share

COinS

Restricted Access Item

Having trouble?