利用全天空影像分析進行日射量及太陽能發電量之估算與預測

All-Sky Image Analysis for Solar Irradiance and Solar Power Estimation and Prediction

朱財平(Chu Tsai-Ping)

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(2021年)

https://doi.org/10.6345/NTNU202101168

摘要

日射量容易受天氣因素影響而產生突坡事件，並造成太陽能發電量不穩定，以致難以將其整合入區域電網當中。本研究以全天空影像儀捕獲白晝之全天空影像，並建立一以影像特徵為基礎之日射量及太陽能發電量估算與預測系統。分析當前空影像以萃取全域及局部雲層之權重；利用光流法推算雲層移動狀態，並製作一分鐘之天空預測圖像。將雲層權重作為長短期記憶(LSTM)之輸入，以進行日射量估算與預測。並且，製作功率曲線，以進一步估算太陽能之發電量。本研究使用多個效能指標評估效果，包括平均絕對誤差(MAE)、均方根誤差(RMSE)、正歸化均方根誤差(nRMSE)以及判定係數(R^2)；並設計數個實驗方法進行比較，實驗結果顯示，本研究所提出方法具有較佳的性能。

關鍵字

日射量 ； 太陽能發電量 ； 全天空影像 ； 光流法 ； 長短期記憶

並列摘要

The ramp event of solar irradiance is prone to occur due to atmospheric conditions, and makes it difficult to integrate solar power into regional power grid. In this study, all-sky images are captured by the all-sky imager. A system is established to perform solar irradiance and solar power estimation and prediction based on image features. The global and the region cloud weights are extracted by analyzing the images. The predicted images for the next minute are made after the cloud movements are derived by using the optical flow. The long short-term memory (LSTM) is used as a training model with cloud weights as inputs for the solar irradiance estimation and prediction. Furthermore, power curves are made to estimate the solar power output. Several performance indices are used to evaluate the performance, including MAE, RMSE, nRMSE and R^2. Several methods are compared, and the experimental results show that the performance of the proposed method is better.

並列關鍵字

solar irradiance ； solar power ； all-sky image ； optical flow ； LSTM

參考文獻

[1] T. H. Hsiao, K. L. Huang and Y. J. Chang, “A model for forecasting of solar power annual electricity and its strategy application,” Journal of Taiwan Energy., vol. 4, no. 4, pp. 401-430, December 2017.

[2] M. Q. Raza, M. Nadarajah and C. Ekanayake, “On recent advances in PV output power forecast,” Solar Energy., vol. 136, pp. 125-144, 2016.

[3] S. Sobri, S. Koohi-Kamali and N. Abd. Rahim, “Solar photovoltaic generation forecasting methods: A review,” Energy Conversion and Management., vol. 156, pp. 459-497, 2018.

[4] A. Al-lahham, O. Theeb, K. Elalem, et al., “Sky imager-based forecast of solar irradiance using machine learning,” Electronics (Switzerland)., vol. 9(10), pp. 1-14, 2020.

[5] M. Khodayar, G. Liu, J. Wang and M. E. Khodayar, “Deep learning in power systems research: A review,” in CSEE Journal of Power and Energy Systems., vol. 7, no. 2, pp. 209-220, 2021.