分布式光伏集群发电功率波动模式识别与超短期概率预测

Pattern Recognition and Ultra-Short-Term Probabilistic Forecasting of Power Fluctuating in Aggregated Distributed Photovoltaics Clusters

分布式光伏功率不确定性的量化评估对于配电网安全稳定运行具有重要意义.考虑到不同出力波动模式下功率特性差异显著,为得出适应功率波动模式的预测模型并对功率不确定性进行精细化评估,提出一种分布式光伏集群发电功率波动模式识别与超短期概率预测方法.首先,综合卫星云图和光伏功率数据,通过出力波动的特征提取构建波动模式识别模型,实现对波动规律的挖掘.在此基础上,通过分类建模考虑不同波动模式的可预测性差异及波动模式与预测误差的关联关系,使预测区间宽度能更好地适应预测误差分布特征.由此实现对不同波动模式下功率不确定性的精细化考虑,从而提高概率预测精度,为电网调度提供更多参考,削弱分布式光伏功率强波动性对电力系统的影响.

The quantitative evaluation of the uncertainty in distributed photovoltaic power is significant for the safe and stable operation of distribution network.Considering the significant differences in power characteristics of different output fluctuation patterns,in order to obtain a prediction model suitable for different fluctuation patterns and to perform a refined assessment of power uncertainty,this paper proposes a method for pattern recognition and ultra-short-term probabilistic forecasting of power fluctuating in aggregated distributed photovoltaic clusters.First,the satellite cloud images and photovoltaic power data are integrated,and the pattern recognition model of fluctuation is constructed via the feature extraction of power fluctuation,realizing the mining of fluctuation rules.On this basis,the difference in predictability of different fluctuation patterns and the correlation between fluctuation patterns and prediction errors are considered via classification modeling,so that the width of prediction interval can better adapt to the characteristics of prediction error distribution.Thus,refined consideration of power uncertainty of different fluctuation patterns is realized to improve the precision of probabilistic prediction,provide more references for power grid dispatching,and weaken the influence of the strong volatility in distributed photovoltaic power on the power system.