Blending as a multi-horizon time series forecasting tool.

摘要

Every day, millions of cubic feet of natural gas is transported through interstate pipelines and consumed by customers all over the United States of America. Gas distributors, responsible for sending natural gas to individual customers, are eager for an estimate of how much natural gas will be used in the near future. GasHour(TM) software, a reliable forecasting tool from the Marquette University GasDay(TM) lab, has been providing highly accurate hourly forecasts over the past few years. Our goal is to improve current GasHour forecasts, and my thesis presents an approach to achieve that using a blending technique.;This thesis includes detailed explanations of the multi-horizon forecasting technique employed by GasHour models. Several graphs are displayed to reveal the structure of hourly forecasts from GasHour. We present SMHF (Smoothing Multi-horizon Forecasts), a step-by-step method showing how a polynomial smoothing technique is applied to current GasHour predications. A slightly different approach of smoothing has also been introduced.;We compare RMSEs of both GasHour forecasts and smoothed ones. Different comparisons resulting from different situations have been demonstrated as well. Several conclusions have been reached. Based on the results, blending techniques can improve current GasHour forecasts. We look forward to applying this blending technique to other fields of forecasting.