:::

臺灣能源期刊論文全文

臺灣能源期刊第11卷第3期內容

出刊日期:September, 2024

題目
混合神經網路於太陽能電廠之光伏面板瑕疵診斷
Title
Defects Diagnosis for Photovoltaic Panel Used in Solar Power Plant by Hybrid Neural Networks
作者
劉宇森 謝振中 謝宗穎
Authors
Yu-Sen Liu, Jenn-Jong Shieh, Zong-Ying Shieh
摘要
本文提出一結合深度50層残差神經網路(Residual Neural Network 50, ResNet50)與長短期記憶模 型(Long Short-Term Memory, LSTM)的混合神經網路,實現對太陽能電廠之太陽能模組的光伏面板 的物理破壞、電損、鳥屎汙染與灰塵等的複數瑕疵自動化檢測,以提高太陽能模組的光伏面板瑕疵 檢測的準確性。所提混合神經網路模型除透過實驗驗證之有效性與準確性外,並與傳統神經網路檢 測模型進行對比。驗證結果顯示,所提之混合神經網路模型於太陽能板之瑕疵檢測的訓練時間及準 確度與驗證時間及準確度,皆優於傳統神經網路模型。
關鍵字
光伏面板,自動化檢測,瑕疵檢測,混合,神經網路
Abatract
The article proposes a hybrid neural network that combines residual neural network (ResNet50) and long short-term memory (LSTM) to realize automated detection of complex defects such as physical damage, electrical loss, bird droppings contamination and dust on photovoltaic panels of solar modules in solar power plants, in order to improve the accuracy of photovoltaic performance of solar modules. In addition to verifying the effectiveness and accuracy of the proposed hybrid neural network model through experiments, it is also compared with the traditional neural network detection model. The verification results show that the proposed hybrid neural network model is superior to the traditional neural network model in terms of accuracy and verification time for defects diagnosis for photovoltaic panel used in solar power plant.
Keywords
photovoltaic panel, automated inspection, defect detection, hybrid, neural network. Received