Ultra-short-term photovoltaic power forecast of TD-BP neural network based on SSA and K-means

被引：0

作者：

Huang Y. ^{[1
]}

Peng D. ^{[1
]}

Yao J. ^{[2
]}

Zhang H. ^{[1
]}

Yu H. ^{[1
]}

机构：

[1] College of Automation Engineering, Shanghai University of Electric Power, Shanghai

[2] Shanghai Minghua Electric Power Technology Engineering Co., Ltd., Shanghai

来源：

Taiyangneng Xuebao/Acta Energiae Solaris Sinica | 2021年 / 42卷 / 04期

关键词：

Forecasting; K-means; Neural network; PV generation; Singular spectrum analysis; Time delay characteristics;

D O I：

10.19912/j.0254-0096.tynxb.2018-1268

中图分类号：

学科分类号：

摘要：

Aiming at the PV power output prediction without enough accuracy, a PV output prediction model that combines Singular Spectrum Analysis (SSA), K-means clustering, Time delay characteristics (TD), and BP neural networks is proposed. Using the similar day theory to select a variety of weather types as training samples, through the decomposition and reconstruction of the singular spectrum analysis, the trend and quasi-period components contained in the time series are extracted, using K-means clustering method to cluster the reconstructed weather samples into K types. Using temperature, wind speed, weather type, and historical output as sample attribute, each weather type is processed by a delayer to form a sample set with delay characteristics. This sample set is used as the input of BP neural network to construct a TD-BP neural network PV power forecast model based on SSA and K-means. The results show that the model has more accurate forecast for PV output power and has certain feasibility and practicality. © 2021, Solar Energy Periodical Office Co., Ltd. All right reserved.

引用

页码：229 / 238

页数：9