基于奇异谱分析的超短期风电功率多步预测  

Multi-step prediction of super-short-term wind power based on singular spectrum analysis

作　　者：吴坚 项颂 阎诚 吴晓丹 马继涛 刘福锁[3] Wu Jian;Xiang Song;Yan Cheng;Wu Xiaodan;Ma Jitao;Liu Fusuo(State Grid East Inner Mongolia Electric Power Supply Co.,Ltd.,Hohhot 010000,China;School of Electrical Engineering,Southeast University,Nanjing 210096,China;NARI Technology Co.,Ltd.,Nanjing 211106,China)

机构地区：[1]国网内蒙古东部电力有限公司,内蒙古呼和浩特010000 [2]东南大学电气工程学院,江苏南京210096 [3]国电南瑞科技股份有限公司,江苏南京211106

出　　处：《可再生能源》2021年第11期1548-1555,共8页Renewable Energy Resources

基　　金：国家电网公司总部科技项目(SGMD0000DDJS1900534)。

摘　　要：针对非平稳风电功率序列的波动特性,单一预测模型无法挖掘出深层次的时序特征的问题,文章提出了一种基于奇异谱分析时序的分解组合预测方法。首先采用Cao方法确定奇异谱分析最佳嵌入维度,并对功率时间序列进行分解;然后考虑各分量多时间尺度特性,通过粒子群优化参数的最小二乘支持向量机对各子序列建立预测模型;最后通过对各子序列进行迭代多步预测并将结果叠加,得到实际预测功率值。实际算例表明,与其他分解方法相比,采用奇异谱分析的组合模型能够有效保留原序列中的关键趋势和复杂特征,具有更好的预测性能。For the volatility characteristics of non-stationary wind power sequences,a single prediction model cannot excavate the problem of deep time series characteristics.A time series decomposition combined prediction method based on singular spectrum analysis is proposed.First,the Cao method is used to determine the best embedding dimension for singular spectrum analysis and to decompose the power time series.Second,considering the multi-time scale characteristics of each component,a least squares support vector machine with particle swarm optimization parameters is used to establish a prediction model for each time series.Finally,iterative multi-step prediction is performed on each time series and the results are accumulated to obtain the actual predicted power value.Case study demonstrates that compared with other decomposition methods,the combined model using singular spectrum analysis can effectively retain the key trends and complex features in the original time series,and significantly improve the prediction accuracy of the super-short-term wind power.

关 键 词：时间序列 风功率预测 奇异谱分析 Cao方法 样本熵 

分 类 号：TK8[动力工程及工程热物理—流体机械及工程] TM71[电气工程—电力系统及自动化]