台灣地區住宅部門氣體能源動態需求模式之研究

Abstract

本研究旨在構建台灣地區住宅部門氣體能源需求之動態模式，期能探 討影響家庭使用氣體能源的主要因素與發展合適的模式用來作預測與規劃 。利用一般化 Box-Jenkins多元時間數列模式分析法分別按液化石油氣與 天然氣兩種氣體燃料構建︰(1)自我迴歸移動平均整合模式（簡稱ARIMA模 式）具有年曆變動型態，用來說明氣體能源資料受到每個月不同星期天數 或移動年假之影響效應，(2) 轉換函數模式探討消費行為之動態反應包含 系統內平均每戶家庭收入、燃料平均價格與平均氣溫等變數之時差關係並 考慮干擾項為一種 ARIMA模式，(3) 轉換與介入模式說明介入事件的發生 均與價格調整有關。(4) 聯立轉換函數模式探討氣體能源總需求模式與能 源選擇模式。同時，根據各種構建的模式進行超前一個時期的預測，發現 其預測精確度甚高，顯示近代的時間數列分析法為一種非常有效且實用的 能源預測方法。 The purpose of this research is to construct dynamic demand models for gas energy by residential sector in Taiwan. In this study, we explore the dynamic relationships among several potentially relevant variables and develop appropriate models for forecasting and planning. The generalized Box-Jenkins procedure of multiple time series modeling can be used separately for liquefied petroleum gas and natural gas to build following models :(1) ARIMA models with the calender variation effects, the models are used to account for the trading day and moving holiday patterns in the gas consumption data. (2) Transfer function models, we use these models to study the dynamic response of consumer behavior towards the average income of each household, the average fuel price, and the average temperature. (3) Transfer function with intervention effects, these models are used to re- late the price change with the occurrence of external events. (4) Simultaneous transfer function models, these models are used to study the total gas demand and fuel choice in the residential sector. In addition to the above application, using the above esti- mated models to do prediction on a future period, we find that the prediction is very close to the actual data. Thus modern time series models are also found to be a very practical and useful tool for energy forecasting.