Short-Term Power Forecasting of Photovoltaic Power Generation Based on Similar Day and Improved Principal Component Analysis

Yun Wu ； Meng-ke Wu ； Li-shan Bao ； Chun-zhe Li

《電腦學刊》 31卷5期 (2020/10) Pp. 187-197

https://doi.org/10.3966/199115992020103105014

摘要

Aiming at the problem of the accuracy of short-term photovoltaic power forecasting, in this paper, a short-term photovoltaic power forecasting method with dual correlation selection of similar day and improved principal component analysis is proposed. First, based on the traditional similar day selection method, a dual correlation method combining weighted Euclidean distance and weighted gray correlation degree is introduced to select similar days with higher similarity. Then, logarithmic processing is performed on the principal component analysis method to increase the principal component factor contribution rate, thereby improving the accuracy of the prediction model. Taking the historical data of a photovoltaic power station in Qinghai as an example, the short-term power forecasting method for photovoltaic power generation using the dual correlation selection similarity day and improved principal component analysis is compared with the method using the traditional similarity day and principal component analysis and the method using dual correlation selection of similarity day and principal component analysis. The results show that the prediction accuracy of the proposed method in this paper is higher.

關鍵字

dual correlation ； photovoltaic power generation ； short-term power prediction