Date of Award
Spring 2009
Document Type
Thesis - Restricted
Degree Name
Master of Science (MS)
Department
Electrical and Computer Engineering
First Advisor
Brown, Ronald H.
Second Advisor
Corliss, George F.
Third Advisor
Adya, Monica
Abstract
Natural gas utilities forecast the flow of natural gas of their customers to ensure that they have enough gas to meet their customer demand. Since a majority of forecasting methods rely on the assessment of the relationship between consumption and temperature to forecast consumption, a good part of the error in flow forecasting comes from customers whose consumption does not depend highly on ambient temperature. There is financial incentive for utilities to improve the forecasts made for such customers. This involves addressing two problems. The first is to identify the customers whose consumption is not significantly dependent on temperature, and the second is to forecast natural gas flow for such customers. In this work, we present the Quantitative Customer Identification algorithm, which uses the functional relationship between flow and temperature to quantify the degree of temperature sensitivity of gas customers. We then present the Reconstructed Phase Space Prediction algorithm, which forecasts demand for non-temperature-sensitive customers using only historical flow data. We compare the forecasting accuracy of the Reconstructed Phase Space Prediction algorithm to that of a multiple regression model and show that the Reconstructed Phase Space Prediction algorithm provides more accurate forecasts in most of the cases. The Quantitative Customer Identification algorithm is a simple tool to classify time series based on the variables on which the time series is dependent. The Reconstructed Phase Space Prediction algorithm forecasts flow independent of weather data and therefore can be used to forecast flow without using weather information. An important extension of this algorithm is to extended its application to customers whose consumption is slightly dependent on weather This method is a good starting point for research in the direction of reconstructed phase space based predictors to forecast daily gas consumption.
Recommended Citation
Tenneti, Sidhartha, "Identification of Non-Temperature-Sensitive Natural Gas Customers and Forecasting Their Demand" (2009). Master's Theses (1922-2009) Access restricted to Marquette Campus. 4218.
https://epublications.marquette.edu/theses/4218