應用滾動灰色預測方法建構能源消耗預測模型

Abstract

能源是經濟發展的一個重要來源， 1973年的「石油危機」，使國際間開始重視能源消耗量與經濟發展之間的研究，近年來國際能源局勢動盪不安，能源價格逐年攀升，連帶影響到國際的經濟發展情況，許多能源也面臨即將枯竭的窘境，在自然資源有限的條件下，能源消耗量的相關研究已成為當前國際上一個非常重要的議題。能源消耗量是經濟發展的一個重要指標。然而，準確的預測能源消耗量卻相當困難，此乃因能源消耗量受到許多因素的影響，這些因素包含科技發展、政府的政策及各種社會經濟指標(譬如：國內生產毛額(GDP)、人口數量及進出口數字)等因素。過去已經有很多能源消耗量預測方面的研究，大多是利用迴歸分析(regression analysis)或類神經網路(artificial neural network, ANN)等方法來預測能源消耗量，但是這些方法卻有其不足或缺失之處，灰色預測模型(Grey Model, GM(1, 1))可彌補上述預測方法的缺陷。為了有效提升GM(1, 1)之預測準確性，本研究提出一個改良的灰色預測模型，共分兩階段進行建模，第一階段對原始資料進行滾動(rolling)灰色預測，然後再針對殘差項進行第二階段的滾動灰色預測，以降低GM(1, 1)模型之預測誤差。最後，本研究將以台灣能源消耗量之實例來說明本研究所提出方法確實有效提升GM(1, 1)模型之預測準確性。

Since the Oil crisis in 1973, energy has been an important source of economic development. Therefore, many countries are concerned with energy-related issues and energy consumption becomes an important economic index. However, it is generally difficult to accurately predict energy consumption. Although there are studies relating to energy consumption forecasting based on social-economic indicators such as gross domestic product, population and import–export figures, those studies have certain deficiencies and disadvantages, such as regression models and artificial neural network (ANN)-based models. For example, their predictive accuracy is low when the sample is small. Grey model (GM) only needs at least four data points to construct a reliable and acceptable prediction model, thus GM is appropriate to be utilized to predict future energy consumption. However, GM also has some drawbacks. This study proposed an improve grey model, using a rolling grey model combining a residual modification model to construct an energy consumption forecasting model to enhance the prediction accuracy of GM. A Taiwanese energy consumption dataset is utilized to demonstrate the effectiveness and feasibility of the proposed method.