摘要
光伏出力的精确预测有利于确保电力系统的可靠运行,减小投资者的利益风险。考虑到光伏出力的不确定性和非平稳性,首先采用自适应白噪声的完整集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)将原始光伏出力序列分解为一系列相关性较强、较平稳的子序列,再使用核极限学习机(Kernel Extreme Learning Machine,KELM)分别对每一子序列进行预测。由于KELM学习参数选取对其预测性能有较大影响,提出了基于改进蝙蝠算法(Improved Bat Algorithm,IBA)对KELM模型参数进行寻优。最后,将每一子序列预测结果通过求和相加获取最终的预测值。实际算例表明,该IBA算法收敛速度快,全局搜索能力强,所提的CEEMDAN-IBA-KELM组合方法能有效提高光伏出力的预测精度。
Accurate prediction of PV output is helpful to ensure the reliable operation of power system and reduce the risk of investors. Considering the uncertainty and non-stationarity of the PV output,firstly,the complete ensemble empirical mode decomposition with adaptive noise（ CEEMDAN） is used for the decomposition of the original PV output sequence into a series of strong correlation,stable sub sequences; and then kernel extreme learning machine（ KELM）is built to carry out prediction for each sub sequence. Due to the great influence of the selection of KELM learning parameters on the prediction performance,an improved bat algorithm（ IBA） is proposed to optimize the parameters of KELM. Finally,the final prediction value is obtained by adding each subsequence. The practical example shows that the IBA has a fast convergence speed and strong global search ability,and the proposed CEEMDAN-IBA-KELM combination method can effectively improve the prediction accuracy of PV output.
出处
《电力科学与工程》
2017年第12期15-21,共7页
Electric Power Science and Engineering
关键词
光伏出力预测
自适应白噪声
集合经验模态分解
核极限学习机
参数优化
改进蝙蝠算法
photovoltaic output prediction
complete ensemble empirical mode decomposition
adaptive noise
kernel extreme learning machine
parameter optimization
improved bat algorithm
