基于自组织特征映射神经网络的短期负荷预测  

Short-Term Load Forecasting Based on Self-Organizing Feature Mapping Neural Network

作　　者：赵菁[1] 彭慧敏[1] 张家亮[1] 谢维廉[1] 

机构地区：[1]贵州工业大学电气工程学院,贵州贵阳550003

出　　处：《贵州工业大学学报（自然科学版）》2003年第2期57-62,共6页Journal of Guizhou University of Technology(Natural Science Edition)

摘　　要：提出了一种基于自组织特征映射神经网络(Kohonen网络)的短期负荷预测方法,根据Kohonen网络的聚类特性,样本在输入时就已分好类。输入既有与负荷曲线平滑性有关的数据又有反映负荷周期性变化的数据。在学习训练时,区别于普通的无监督竞争学习采用有监督竞争学习方式,缩短了学习时间,提高了学习精度。实例分析证明了该方法的有效性。A selforganizing feature mapping neural network (Kohonen neural network) based approach to shortterm load forecasting is presented in this paper. Utilizing the generalization of Kohonen neural network, input has been clustered before use. Input data are related to both smoothness and regularity of load curve. Differentiating from conventional winnertakeall learning, learning under supervision and competition reduces the learning time and improves learning accuracy. Simulation testifies the effectiveness of the method.

关 键 词：短期负荷预测 自组织特征映射 神经网络 电力系统 

分 类 号：TM715[电气工程—电力系统及自动化] TP183[自动化与计算机技术—控制理论与控制工程]