基于LPF-VMD和KELM的风速多步预测模型  

Multi-step Wind Speed Prediction Model Based on LPF-VMD and KELM

作　　者：向玲[1] 邓泽奇 赵玥 XIANG Ling;DENG Zeqi;ZHAO Yue(School of Energy Power and Mechanical Engineering,North China Electric Power University,Baoding 071003,Hebei Province,China)

机构地区：[1]华北电力大学能源动力与机械工程学院

出　　处：《电网技术》2019年第12期4461-4467,共7页Power System Technology

基　　金：国家自然科学基金项目(51675178)~~

摘　　要：提出了一种基于低通滤波-变分模态分解的风速信号预处理方法。该方法首先从能量的角度直接通过低通滤波筛选出信号的趋势成分,再利用VMD将剩余信号分解成一系列相对平稳的限带内禀模态函数。将该信号预处理方法与核极限学习机结合,建立了风速多步预测模型。为了提高模型的预测性能,采用鸟群算法优化KELM预测模型的4个参数,以最优参数组合建立预测模型。最后以浙江某风电场采集的实际风速数据为例进行预测验证,结果表明所提出的多步预测方法具有较高的预测精度和运行效率。A novel wind speed signal pre-processing technology based on low pass filter-variational mode decomposition(LPF-VMD) is proposed in this paper. Firstly, the trend component of the signal is filtered directly with low-pass filter on the viewpoint of energy, and then the residual signal is decomposed into a series of relatively stable band-limited intrinsic mode functions(BIMF) with VMD. On this basis, a multi-step wind speed prediction model based on LPF-VMD and kernel extreme learning machine(KELM) is established. In order to improve the prediction performance of the model, bird swarm algorithm(BSA) is utilized to optimize the four parameters of KELM prediction model, and a forecasting model is established based on combination of the optimal parameters. Finally, based on actual wind speed data acquired at a wind farm, experimental results show that the proposed method has high prediction accuracy and operation efficiency.

关 键 词：风速预测 变分模态分解 相空间重构 核极限学习机 鸟群算法 

分 类 号：TM614[电气工程—电力系统及自动化]