Ultra-Short-Term Solar PV Power Forecasting Method Based on Frequency-Domain Decomposition and Deep Learning

• Type: Article in International Conference Proceedings Book
• Date: 2020

Authors

Hu, L (Author) Other; Zhen, Z (Author) Other; Wang, F (Author) Other; Qiu, G (Author) Other; Li, Y (Author) Other; Shafie khah, M (Author) Other; Catalno, JPS (Author) FEUP

Conference proceedings International

Title: IEEE Industry Applications Society Annual Meeting, IAS 2020, Detroit, MI, USA, October 10-16, 2020

Pages: 1-8

2020 IEEE Industry Applications Society Annual Meeting, IAS 2020

10 October 2020 through 16 October 2020

Indexing

ISI Web of Knowledge - 7 Citations

Scopus - 5 Citations

Other information

Authenticus ID: P-00T-GVC

DOI: 10.1109/ias44978.2020.9334889

Abstract (EN): Ultra-short-term photovoltaic (PV) power forecasting can support the real-time dispatching of power grid and the optimal operation of PV power station itself. However, due to various meteorological factors, the photovoltaic power has great fluctuations. To improve the refined ultra-short-term forecasting technology of PV power, this paper proposes an ultra-short-term forecasting model of PV power based on optimal frequency-domain decomposition and deep learning. First, the amplitude and phase of each frequency sine wave is obtained by fast Fourier decomposition. As the frequency demarcation point is different, the correlation between the decomposition component and the original data is analyzed. By minimizing the square of the difference that the correlation between low-frequency components and raw data is subtracted from the correlation between high-frequency components and raw data, the optimal frequency demarcation points for decomposition components are obtained. Then convolutional neural network is used to predict low-frequency component and high-frequency component, and final forecasting result is obtained by addition reconstruction. Finally, the paper compares forecasting results of the proposed model and the non-spectrum analysis model in the case of predicting the 1 hour, 2 hours, 3 hours, and 4 hours. The results fully show that the proposed model improves forecasting accuracy.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 8