基于CEEMDAN与量子粒子支持向量机的电力负荷组合预测

为更精确地对电力系统负荷进行预测,提出一种基于添加自适应白噪声的完全集合经验模态分解与量子粒子支持向量机的组合预测方法。首先针对原始经验模态分解办法中存在的模态混叠及集合经验模态分解方法引入白噪声造成信号失真等问题,提出添加自适应白噪声的完全集合经验模态分解方法,并用其将原始信号分解到不同时间尺度。利用支持向量机方法分解结果分别进行预测,并采用量子粒子方法对支持向量机中的不敏感损失系数、惩罚系数及核宽度系数进行寻优,从而得到最好的预测结果。最后,通过对青海某区域的电力系统负荷预测,并引入不同方法进行对比,证实了该方法的有效性与实用性。

The Power Load Combined Forecasting Based on CEEMDAN and QPSO-SVM

To predict the power system load more accurately , this paper proposes a combined forecasting method based on the complete ensemble empirical mode decomposition with adaptive noise and quantum particle swarm opti -mization.Firstly, aiming at the modes overlap problem and signal distortion existing in ensemble empirical mode de -composition , this paper proposes the complete ensemble empirical mode decomposition with adaptive noise , and de-composes the original signals into the different time scales .Then, it uses the support vector machine to predict the de-composition result , and employs the quantum particle swarm optimization method to optimize the insensitive loss coeffi -cient, penalty coefficient and kernel function .Finally, by forecasting the power system load in a certain domain of Qinghai province and comparing it with another different methods , it proves the validity and practicability of the meth-od mentioned in this paper .