應用自組性演算法與修正灰預測建構能源消耗預測模型

Abstract

能源是經濟發展中不可或缺的要素，1973年爆發第一次「能源危機」，因此各國政府非常重視能源消耗與經濟發展的問題;能源消耗量也成為一個國家經濟發展的指標。過去已經有許多能源預測之相關研究，提出一些預測能源消耗的方法，例如:Box-Jenkins模型、迴歸預測模型、類神經網路、灰預測模型等，這些模型大致可以分為兩類，一是因果關係之預測模型，二是時間序列預測模型，大多的研究都是利用時間序列預測模型來建構，由於時間序列預測模型在建模時僅考慮能源資料，而能源消耗量是一種系統內部訊息不明確的資料，任何的經濟指標，(例如:人口數、經濟政策、產業結構)，都可能影響到能源的消耗量，故在預測能源消耗量時，應該考慮這些因素才能提升預測模型之準確率。因此，本研究之應用自組性演算法與殘差修正灰預測來建構能源消耗預測模型以有效提升預測之準確率。

Energy is an essential element of economic development for all countries. Energy consumption has become an important economic index in recent years. Time-series models were frequently used to construct energy consumption model, because it only needs the information of energy consumption. Many studies have developed time-series prediction models for energy consumption using Box-Jenkins models, Autoregressive Integrated Moving Average (ARIMA) or gray models. However, energy consumption would be affected by economic indicators such as population, economic policy, and industrialization of a country, etc. Therefore, a cause-result prediction model is more suitable for forecasting the energy consumption. This study employs GMDH and modified gray model to construct a cause-result prediction model for energy consumption model to enhance the accuracy of existing time-series forecasting models.