摘要
为提高短期负荷预测的精度,构建一种基于自组织特征映射神经网络和模糊理论的短期负荷预测方法。预测分两个阶段,先根据自组织特征映射神经网络聚类特性,进行第一阶段的负荷预测,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的学习方式以缩短学习时间,提高学习精度。第一阶段预测出一个基本的负荷值后,在第二阶段利用模糊理论根据前一个时段的预测误差和误差变化对其进行校正。使用该方法不仅能预测工作日负荷还能预测休息日负荷,实例分析证明了该方法的有效性。
In order to improve the precision of short-term load forecasting, an approach to short-term load forecasting based on self-organizing feature mapping neural network and fuzzy theory was proposed. The forecasting included two steps. First, forecasting load according to the characteristics of self-organizing feature mapping neural network. The learning time was reduced and the learning accuracy was improved by adopting learning under supervision and competition instead of the conventional winner-take-all learning. Second, modifying the forecasting results obtained in step lthe modification was made based with on ,based on the fuzzy theory . the forecasting error and error movement of the previous hour. The method can be applied to both working day and weekend day. The simulation result testifies the effectiveness of the proposed methodology.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2010年第3期129-133,共5页
Proceedings of the CSU-EPSA
关键词
自组织特征映射
神经网络
有监督竞争学习
模糊理论
短期负荷预测
self-organizing feature map（SOM）
neural network
learning under supervision and competition
fuzzy theory
short-term load forecasting
