具影像特徵之LSTM深度遞迴類神經網路之日射量預測

Solar Irradiance Forecasting Using LSTM Deep Recurrent Neural Networks with Image Feature

周建華(Jhou Jian-Hua)

指導教授 ： 呂藝光

國立臺灣師範大學/科技與工程學院/電機工程學系/碩士(2019年)

https://doi.org/10.6345/NTNU201901090

摘要

由於日照強度會因為雲層厚度、空氣霾害等問題而受到影響，進而造成太陽光電發電量的不穩定，所以能夠準確的預測日射量是件重要的事情。在本論文中使用具長短期記憶(LSTM)的遞迴類神經網路(RNN)進行日射量的預測。首先建置一日射量紀錄系統，及天空影像採集系統，這兩種系統將記錄每天的日射量及天空影像變化，並儲存於MySQL資料庫。在天空影像方面，利用影像處理方法萃取出天空影像的特徵值，之後將影像特徵值與日射量做為LSTM遞迴類神經網路(LSTM-RNN) 輸入 ，以進行預測。最後，本文以領前五分鐘至六十分鐘進行日射量預測，並進行許多方法比較，以驗證本文所提方法的預測效能。

關鍵字

太陽能預測 ； 深度學習 ； 遞迴類神經網路 ； 影像特徵

並列摘要

Since solar irradiance is affected by factors, such as the thickness of the clouds and the air pollution. These factors can cause instability in solar power generation. Therefore, it is important to be able to accurately predict the amount of solar irradiance. In this paper, recurrent neural networks (RNNs) with long short-term memory (LSTM) are used to develop a solar irradiance forecasting. First, a solar irradiance recording system and a sky image capture system were built. The two systems record daily solar irradiance and sky images, and then store them in the MySQL database. For sky images, some processing methods are used to obtain the feature values. The feature values and the solar irradiance amount are inputs of the LSTM recurrent neural networks (LSTM-RNN). Finally, the solar irradiance forecasting with lead times of 5 to 60 minutes is presented, and then some comparison results are conducted in order to verify the performance of the proposed method in this paper.

並列關鍵字

Solar Irradiance Forecasting ； Deep Learning ； Recurrent Neural Network ； Image Feature

參考文獻

參考文獻

[1] Ahsen Ulutaş, Recep Çakmak, and İsmail Hakkı Altaş, “Hourly Solar Irradiation Prediction by Artificial Neural Network Based on Similarity Analysis of Time Series,” 2018 Innovations in Intelligent Systems and Applications Conference (ASYU), pp. 1-6, 2018.

[2] Stephen M. Ruffing, Ganesh K, and Venayagamoorthy, “Short to Medium Range Time Series Prediction of Solar Irradiance Using an Echo State Network,” 2009 15th International Conference on Intelligent System Applications to Power Systems, pp. 1-6, 2009.

[3] H. Nazaripouya, B. Wang, Y. Wang, P. Chu, H. R. Pota, and R. Gadh, “Univariate time series prediction of solar power using a hybrid wavelet-ARMA-NARX prediction method,” 2016 IEEE/PES Transmission and Distribution Conference and Exposition (T&D), pp. 1-5, 2016.

[4] Rui Huang, Tiana Huang, Rajit Gadh, and Na Li, “Solar generation prediction using the ARMA model in a laboratory-level micro-grid,” 2012 IEEE Third International Conference on Smart Grid Communications (SmartGridComm), pp. 528-533, 2012.