摘要
为了提高预测模型精度以及降低模型的复杂程度,减小模型的训练时间,文章提出一种基于改进模糊C均值聚类算法的数据预处理方法,以提高风电功率时间序列的预测模型的预测性能。首先,对实测风电功率混沌时间序列进行了相空间重构;其次,对相空间中的各维输入序列与输出功率序列进行相关性分析,使用相点与相对相关系数的加权建立聚类判据;然后,结合减聚类算法对模糊C均值聚类的收敛速度进行改进,改进的模糊C均值算法将输入序列聚为4类,对每类数据建模。结果表明,对原始数据进行聚类预处理后,预测模型的精度得到了提高。
In order to improve the accuracy of the prediction model while reducing the complexity and training time of the model, this paper proposed a data pre-processing technology based on improved fuzzy C-means clustering algorithm to improve the performance of prediction model on predicting wind power time series. Firstly,reconstruct the phase space of wind power chaotic time series. Secondly, analyze the correlation between each dimension series of the input phase space and output wind power time series, and set up clustering criterion vector with phase point vector and relatively correlation coefficient,then, improve the convergence speed of fuzzy C-means clustering algorithm by introducing subtractive clustering,clustering the input space into four clusters with the improved fuzzy C-means clustering algorithm, modeling each cluster of data. The results showed that after clustering the original data, not only the precision of the prediction model is improved; the model of training time is reduced.
作者
张里
王兰
李红军
廖小君
王婷婷
张江林
刘友波
Zhang Li;Wang Lan;Li Hongjun;Liao Xiaojun;Wang Tingting;Zhang Jianglin;Liu Youbo(Skill Training Center, State Grid Sichuan Electric Power Company, Chengdu 610072, China;Control Engineering College, Chengdu University of Information Technology, Chengdu 610225, China;School of Electrical Engineering and Information, Sichuan University, Chengdu 610065, China)
出处
《可再生能源》
CAS
北大核心
2018年第12期1871-1876,共6页
Renewable Energy Resources
基金
国家自然科学基金项目(51207098)
四川省科技厅项目(2015GZ0204)
关键词
数据预处理方法
风电预测
模糊C均值聚类
模糊模式识别
相关性分析
data preprocessing methods
wind power prediction
fuzzy C -means clusteringalgorithm
fuzzy pattern recognition
correlation analysis
