使用灰色理論及基因規劃法建構預測模型-以能源消耗之預測為例

Abstract

由於能源對各國的經濟發展影響鋸大，因此能源消耗之預測目前已經成為各國一個重要的議題。一般多使用傳統統計方法來建構能源消耗預測模型，這些預測模型通常要求資料須符合一些假設，例如：常態分配、獨立性或資料量需相當大。然而，能源消耗方面之資料量常常很少或者屬非常態分配，因此必須用其它預測方法才能得到準確的預測結果。由於建構灰理論中的灰預測模型僅需四筆以上之資料即可，且資料不需假設常態分配，因此適合用來預測能源消耗量。然而，針對一些複雜的資料型態，灰預測模型可能仍會產生相當大的預測誤差。因此為降低灰預測模型之預測誤差，本篇論文針對能源消耗資料，結合了灰色理論以及基因規劃法提出三個改良之灰預測模型以預測能源消耗量。最後，本論文應用1990年到2007年中國以及美國能源消耗資料說明了本研究所提出之模型比文獻上現有之一些能源消耗預測模型包含：灰預測模型、基因規劃法以及自我整合移動平均等所建之模型，在預測能源消耗量之準確度與穩定度上均有較佳之結果。

The issue of energy consumption has become increasingly important due to its impact on the economies of nations. Forecasting energy consumption by conventional statistical methods usually requires assumptions such as normality, independence or large data sets. However, data collected on energy consumption are often very few and do not meet the statistical assumptions. The grey forecasting model, based on grey theory, can be constructed for at least four observations and it does not require any statistical assumptions. Thus, it is suitable to use a grey forecasting model to forecast the energy consumption. In some cases, however, the grey forecasting model may obtain large forecasting errors. To overcome this problem, this dissertation proposes three improved grey forecasting models, which combines grey theory and genetic programming. Finally, two energy consumption data sets from 1990 to 2007 of China and US are used to demonstrate the effectiveness of the proposed forecasting models. The proposed models are also compared with the existing forecasting models of energy consumption. The empirical results indicated that the proposed models can enhance both of the accuracy and the precision of forecasting the energy consumption data.