基于模糊聚类与函数小波核回归的短期负荷预测方法  

Short-term load forecasting based on fuzzy clustering and functional wavelet-kernel regression

作　　者：祖向荣[1] 田敏[1] 白焰[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206

出　　处：《电力自动化设备》2016年第10期134-140,165,共8页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(61571189)~~

摘　　要：短期负荷预测日益成为智能电网的重要课题。将历史日负荷序列表示成等负荷段组成的子序列集合,基于模式相似性方法,采用模糊聚类与函数型小波核非参数回归(FWKNR)相结合的组合预测算法。FWKNR将预测日负荷表示成历史日相应负荷段的加权平均值,将与参考段更相似的段赋予更高权重,并基于离散小波变换的形状相似性度量,采用N-WE计算权重;由预测日各分段预测快速完成日预测。模糊聚类针对单一用户历史负荷进行典型负荷模式的分类预处理,并识别与预测日有更相似行为模式约减的有效训练样本集合参与模型预测。基于某地区实际负荷数据,实验比较分析验证了组合算法的优越性。STLF（Short-Term Load Forecasting） is an important issue of smart grid. The historical daily load curve is expressed as a set of hourly load segments and a hybrid forecasting algorithm based on the pattern similarity method is applied,which combines the fuzzy clustering with the FWKNR（Functional Wavelet-Kernel Nonparametric Regression）. FNWKR is applied to express the load curve of the predicted day as a set of weighted average of the corresponding hourly load segments of historical days,which assigns higher weight to the segment with higher similarity and uses N-WE（Nadaraya-Watson Estimator） to calculate the weight based on the shape-similarity measurement of discrete wavelet transform. The daily load is predicted by the quick forecasting of load segments. Fuzzy clustering is used to pre-classify the historical loads to typical load patterns for a particular customer and recognize the effective reduced training sample set with more similar behaviour pattern to the predicted day for the model forecasting. Based on the practical load data of a region,the experimental analysis verifies the superiority of the proposed algorithm.

关 键 词：短期负荷预测 模式相似性方法 模糊聚类 非参数小波核回归 非线性时间序列 智能电网 

分 类 号：TM715[电气工程—电力系统及自动化]