- 學位論文
應用類神經網路及基因演算法預測風速與風力發電量
Wind Speed and Wind Power Forecasting Using Neural Networks and Genetic Algorithms
摘要
風速的連續變化將影響風力發電機組的發電量亦隨之改變。若能準確的預測發力發電機組在任一時刻的風速所獲得的實際發電量，便能評估其經濟效益與發電成本。類神經網路是一個應用於預測的最佳工具，其具有強大的適應性學習能力。本文將應用類神經網路來進行風速與發電量之預測。 本論文利用倒傳遞網路、放射狀基底函數網路、適應性網路架構的模糊推論系統等三種類神經網路及結合基因演算法，進行風速及風力機發電量預測，並比較各網路的輸出結果，以找出最佳的預測網路與方法。 本文使用氣象觀測站與電力品質分析儀，長時間記錄與監控，得到相關的風速與發電量資料，並應用資料探勘(Data Mining)PolyAnaly模擬軟體，針對輸入資料做前置處理，去除影響期望值最低的參數，而減少輸入參數後，再利用類神經網路軟體NeuroSolutions所產生的類神經網路進行訓練與預測，並計算實際值與期望值兩者間之誤差量及相關參數，以驗證使用之類神經網路預測方法之有效性及準確性。
並列摘要
The generated power of wind-turbine generator (WTG) varies with variation of wind speed. If the generated power of WTG can be forecasted accurately at anytime, cost of conventional power generation can be reduced. Artificial Neural Networks (ANN) is one of the best tools applied to forecast. It provides a powful learning ability. This thesis focuses on ANN applied for forecasting wind speed and generate electricity. The thesis uses three ANNs, Back-Propagation Network, Radial Basis Function Network and Adaptive network fuzzy inference system, as well as Genetic Algorithm to process the forecasts of wind speed and generated electricity. Besides, the results of the network output are compared in order to find the appropriate method. The thesis uses the weather station and power quality monitor. The information of wind speed and the information of generated power in every hour are recorded. The simulation software Data Mining is applied to pre-process the input data and removes the least correlated input data. Then the NeuroSolutions software is used to conduct training and forecasting. The error values and correlative parameters between realistic values and expected values are compared to verify the effectiveness and accuracy of the proposed method based on ANN.