摘要
利用高斯过程回归模型描述不同时间尺度下负荷序列的基本演化趋势、强波动和随机噪声水平。对历史负荷数据进行可视化分析,挖掘整理出不同时间尺度下的历史负荷数据,与气候数据一起作为特征数据。采用模糊聚类算法对整理出的特征数据进行筛选,消除冗余信息,构造出紧凑有效的最优特征集。将历史负荷数据代入高斯过程回归模型进行训练,并利用实际电力数据对模型进行了测试,验证了所提出模型具有较高的预测精度与鲁棒性。
The Gaussian process regression model is used to describe the basic evolution trend,strong fluctuation and random noise level of load sequence on different time scales. Visualization analysis of historical load data is carried out. The historical load data on different time scales are mined and sorted out,which are used as characteristic data together with climate data. The fuzzy clustering algorithm is used to filter the characteristic data sorted out to eliminate the redundant information and construct the compact and effective optimal feature set. The historical load data are brought into the Gaussian process regression model for training,and the model is tested by using the actual power data,which proves that the proposed model has high prediction accuracy and robustness.
作者
武志宇
周筱淋
王雨佳
WU Zhiyu;ZHOU Xiaolin;WANG Yujia(Liaoning Technical University,Huludao 125000,China)
出处
《山东电力高等专科学校学报》
2021年第4期5-9,共5页
Journal of Shandong Electric Power College
关键词
负荷预测
高斯过程回归模型
模糊聚类算法
load forecasting
Gaussian process regression model
fuzzy clustering algorithm