摘要
针对电力负荷的一些特性,提出了使用非线性数据分类学习机的理论来解决短期负荷预测问题。利用有偏最小最大概率机进行数据学习分类,对采集到的信息进行分类、特征提取,形成归一的数据类型;用得到的分类数据作为有偏最小最大概率回归模型的输入进行训练预测。该方法通过核函数将输入向量从低维空间映射到高维空间,在高维空间实现了基于高阶统计信息的负荷影响因数的特征提取,既全面考虑了影响负荷预测的历史时间序列、气象等各种因素,又避免了由于输入变量过多而导致模型结构复杂、训练时间长等不足。计算实例表明,文中提出的方法用于短期负荷预测,其预测精度较高,且训练时间较短,方法可行且有效。
A non-linear data classification machine is presented to deal with the characteristics of power demand in load forecasting. BMPMC is adopted to classify the load data and extract the load characteristics among various factors. BMPMR model is trained using the data classified by BMPMC as the input of the model to accomplish the final forecasting. Using Kernel function to map the input from low-dimension to high-dimension not only avoids the complexity and long training time of the model but considers various factors comprehensively. Case studies on a real power system show that the proposed model is feasible and promising for short-term load forecasting.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2007年第4期46-49,共4页
Proceedings of the CSU-EPSA
关键词
电力系统
负荷预测
数据分类
有偏最小最大概率分类
有偏最小最大概率回归
power system
load forecasting
data classification
biased minimax probability machine classification(BMPMC)
biased minimax probability machine regression (BMPMR)