Interval Prediction Model for Electrical Load Using Fuzzy Time- Series and Grey Theory

[[abstract]]時間序列區間預測法較傳統預測法能更週嚴地考慮到預測事件之變動性，並能提供更充足的資訊供決策者作衝擊評估、風險管理、規劃調度和投資計劃等之參考，因此在企業經營實務上日益受到重視。另一方面，真實世界的預測模型十分複雜，模型中的變數資料有時很難取得精確的數據，僅能運用人類的經驗知識來進行，然而人類經驗的知識通常是模糊的或不明確的，是以一個有效的時間序列區間預測法應能夠處理具模糊語意的能力，以提供更有價值的資訊，供決策者作更週嚴之計劃與決策。為此，本文提出一個結合模糊時間序列法(fuzzy time series)與灰色理論(grey theory)之混合模型(hybrid model)，建構一個區間型趨勢之預測模型，其既能提供單點預測值又能提供區間預測值供決策者作決策時之參考。本研究並以台電民國75年至92年農林魚牧業合計用電量之歷史資料為實證分析對象，以驗證本文所提之模式的效度。[[abstract]]Interval forecasts are often preferable because they take explicitly into account the predictor variability and will provide rich information to policymaker for impact assessment, risks control, planning, dispatching and investment plan. Establishing a forecasting model is essentially important but difficult since the variables are very complicated in real world problems and the associated accurate data are not easily collected. Instead, the data can be gathered from, human experience knowledge, which is usually vague and uncertain. Thus, there is a great necessity for developing an interval forecasting model which can tackle with the uncertainly issue. In this thesis, we propose a hybrid method, which integrate fuzzy time-series and grey theory to construct an interval type of forecasting model that provides forecasting abilities of single-values and interval type-values. We validate the effectiveness of the proposed model by conduction a case study on the data collected from the power load for agriculture, forest, fish and herd industry in Taiwan