- 學位論文
太陽光電系統發電量預測模型之實作
Implementation of Photovoltaic System Generation Prediction Systems
摘要
近年來由於能源危機之議題與環保意識抬頭，且電力電子與材料生產製造技術都大幅提升的時空背景下，激勵更多研究人員及廠商致力於綠色能源的研究與開發；但綠色能源受天候影響變因較為劇烈，為穩定及提高綠色能源在長時間之使用度，數學家與科學家潛心研究，嘗試對綠色能源的發電量輸出進行預測，便可以大幅提升綠色能源替代利用率。 本論文使用倒傳遞類神經演算法，藉由訓練類神經網路達到實現預測太陽光電發電量之目的。透過收集小型太陽能光電發電廠之數據，將資料以天候類型分類並透過不同之數據正規化法則，導入至類神經網路模型訓練。藉由正規化法則可減少類神經網路模型之演算誤差，再透過調整類神經模型參數(隱藏層層數、隱藏層神經元數量)，對類神經模型預測準確度差異進行測試。結果顯示，適度改變類神經隱藏層層數與隱藏層神經元數量可以增進類神經記憶功能。最後在類神經模型預測研究中加入線上學習機制，結果證實具備線上學習機制的類神經模型可以有效提高太陽光電系統發電量之預測準確。
並列摘要
In recent years, due to the issues of environmental consciousness coupled with the energy crisis and the rapid development of power electronics technologies, more researchers and manufacturers are encouraged to further research and develop of green energy. This thesis applied the Back Propagation Artificial Neural Network Algorithm, which is used to train the neural network, in order to predict power output from photovoltaic system. Operational data sorted by the weather type are collected from a small photovoltaic system. Different normalization pre-processes are applied to adjust data for the neural network training. The prediction accuracy of network can be improved by properly adjusting the amount of hidden layer and neurons. The online training function is also adapted for the power prediction. The results show that with the online training ability, the performance of the neural network can be improved.