摘要
以实际风力机功率数据为基础,通过经验模态分解(EMD)将风力机功率时间序列分解为多个特征模态函数。利用分形理论对风力机功率时间序列各分量的局部时频特性进行研究判断,并对新的风力机功率时间序列进行重构。然后,利用最大Lyapunov指数等特性指标分析风力机功率时间序列的混沌特性,并分别分析3个尺度子序列的行为动力学特性。最后,建立基于蚁群优化极值学习机制,构建风力机功率时间序列预测模型。仿真结果表明,该模型比其他单一预测模型具有更高的预测精度,可用于工程实际。
In this paper,the time series of wind power(TSWP)is decomposed into multiple eigenmode functions by empirical mode decomposition(EMD)based on the actual wind power data.The local time-frequency characteristics of each component of the TSWP are judged,and the new TSWP is reconstructed by fractal theory.Then,the chaos characteristics of the TSWP are analyzed by using the maximum Lyapunov index and other characteristics,and the behavior dynamics characteristics of the 3 scale subsequences are analyzed respectively.Finally,the prediction model of the TSWP is built based on the ant colony optimized extreme learning mechanism.The simulation results shown that the proposed model is more accurate than other single forecasting models,which can be used in engineering practice.
作者
文孝强
许洋
Wen Xiaoqiang;Xu Yang(College of Automation Engineering,Northeast Electric Power University,Jilin 132012,China;Jilin Province International Research Center of Precision Drive and Intelligent Control,Jilin 132012,China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2021年第11期293-298,共6页
Acta Energiae Solaris Sinica
基金
吉林省科技厅重大科技专项基金(20180201001SF)
吉林省教育厅“十三五”科学研究规划项目(JJKH20190709KJ)
2019吉林市杰出青年人才培养专项(20190104156)。
关键词
风力机功率
预测
建模
神经网络
EMD
蚁群优化
wind power
prediction
models
neural networks
empirical mode decomposition
ant colony optimization
