| 引用本文: | 黎静华,赖昌伟.考虑气象因素的短期光伏出力预测的奇异谱分析方法[J].电力自动化设备,2018,(5): |
| LI Jinghua,LAI Changwei.Singular spectrum analysis method for short-term photovoltaic output prediction considering meteorological factors[J].Electric Power Automation Equipment,2018,(5): |
|
| 摘要: |
| 在传统奇异谱分析(SSA)方法的基础上，提出一种嵌入气象因素的改进SSA方法，该方法融合了SSA、相关性分析和灵敏度分析等技术，可有效提高传统SSA方法的预测精度。采用SSA技术将光伏出力时间序列分解为低频序列、高频序列和噪声序列，通过Pearson相关系数法确定温度和辐照为影响光伏出力的主要气象因素，再对光伏出力与气象因素之间的灵敏度进行分析。根据灵敏度分析的结果和基准值分别对待预测日的低频序列和高频序列进行修正，将修正结果进行叠加得到光伏出力预测结果。将所提方法运用于某地区的光伏短期预测中，与自回归模型、BP神经网络及传统的奇异谱分析回归方法的对比结果表明，所提方法具有更高的预测精度。 |
| 关键词: 奇异谱分析 气象因素 相关性分析 灵敏度分析 短期光伏出力预测 |
| DOI：10.16081/j.issn.1006-6047.2018.05.007 |
| 分类号:TM615 |
| 基金项目:国家重点研发计划支持项目(2016YFB0900100)；国家自然科学基金资助项目(51377027) |
|
| Singular spectrum analysis method for short-term photovoltaic output prediction considering meteorological factors |
|
LI Jinghua, LAI Changwei
|
|
Guangxi Key Laboratory of Power System Optimization and Energy-Saving Technology, Guangxi University, Nanning 530004, China
|
| Abstract: |
| An improved SSA(Singular Spectrum Analysis) method embedded with the meteorological factors is proposed based on the traditional SSA method, which combines technologies such as SSA, correlation analysis and sensitivity analysis, effectively improving the prediction accuracy of the traditional SSA method. The PV(PhotoVoltaic) output time series is decomposed into low frequency series, high frequency series and noise sequence series by the SSA technology. The temperature and irradiation are determined as the main meteorological factors influencing the PV output by the Pearson correlation coefficient method. The sensitivity between the PV output and meteorological factors is analyzed, according to the results of which and the reference value, the low frequency series and high frequency series of the prediction day are modified respectively and then superimposed to obtain the PV output prediction results. The proposed method is applied in the short-term PV prediction of an area, and the results compared with the AR(AutoRegressive)model, the BP neural network, and the traditional SSA-AR method show that, the proposed method has higher prediction accuracy. |
| Key words: singular spectrum analysis meteorological factor correlation analysis sensitivity analysis short-term photovoltaic output prediction |
