Research on Establishing Forecast Model of Seasonal PV Power Generating System by Neural Network

• Main Authors: Yu Liu, 劉昱
• Other Authors: Kuo-Hsiung Tseng
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/652gdd

碩士 === 國立臺北科技大學 === 電機工程系所 === 105 === The power generating efficiency of PV power Generating is easily influenced by weather、sunshine hours and module temperature. This thesis predicts PV power Generating from solar power generating system in four seasons of NTUT based on sunshine hours, ultraviolet index, average temperature and relative humidity from Data Bank Atmospheric and Hydrologic of 2016. The solar power generating system is 70.38kwp while the PV power Generating efficiency has great deal with solar power module temperature. Module temperature is considered in input parameter hoping to achieve PV power generating of four seasons. Dynamic Neural Networks is used to build the predicting model in this thesis to determine the capability with the parameters aforementioned to carry out Neural Networks training. To compare the result of prediction and the actual PV power generating of four seasons in 2016, adjust the number of the cell and do the normalization, input data analyzing with trial and error method are taken account. Finally, with the method RMSE to judge and predict the model accuracy and compare the influence to trained model with different input parameter.