- 期刊
自我組織式模糊系統作短期電力負載預測
A Self-Organization Fuzzy System for Short-Term Power System Load Forecasting
摘要
本文提出一種新型的自我組織式模糊系統模式，作爲電力系統短期負載預測之用，以預測未來24小時內系統每小時負載。 模糊理論因具有解決主觀性、不確定性問題及以人類語言表達及思考模式的特徵，故應用該理論於電力系統短期負載預測上，可改善以往建構短期負載預測模式時的複雜性。本文利用進化演算的技術，於求取傳統模糊邏輯模式歸屬函數之參數最佳解與對應之最終推論法則。由於進化演算法採隨機搜尋、自然演化的過程，故應用於調整歸屬函數的參數值極爲有效，可改善以往模糊系統無法自動由過去負載與各類天候資料學習的缺點，因而可快速且便利的建立預測系統，亦可提高負載預測的準確性。 本文針對台電系統小時負載及氣象局的溫度資料作測試，首先以其相關性決定輸入變數，以預測未來24小時之系統負載。結果並與倒傳遞神經網路及傳統模糊預測的方法做一比較，以驗證本文所提方法之預測性能。
並列摘要
The paper proposes a new self-organizing fuzzy model for the short-term forecasting of the one-day ahead hourly load demands. Due to the abilities of the fuzzy theory in resolving the problems of subjectivity, uncertainty, as well as human linguistic expression and thinking manner, application of the fuzzy theory to the short-term load forecasting (STLF) can alleviate the complication of constructing the STLF model. In the paper, the evolutionary computing techniques are used to obtain the optimal membership functions and the corresponding inference rules. The mechanisms of stochastic search and natural evolution in the evolutionary computing techniques can adjust the fuzzy membership functions effectively. Therefore, the techniques can overcome the defects that the traditional fuzzy system can notlearn automatically from the historical load and diverse weather data. In addition, not only is the forecasting system established conveniently, but the forecasting accuracy is improved by the proposed method. The practical hourly load demands of the Tai power system and the temperature data collected from the Central Weather Bureau were employed to test the proposed fuzzy systems. In the process, the load and weather variables with higher relationship were chosen as the input variables to predict the future 24-hour load curve. Serving as performance verification of the proposed forecasting method, the results were compared to those obtained from the back-propagation artificial neural network and the existing fuzzy load forecasting method.