摘要
本文提出了联合采用两个人工神经元网络模型,即Kohonen自组织特征映射模型(Kohonen模型)和误差反向传播模型(BP模型)计算配电网线损的新方法。BP模型理论上可映射任意复杂的非线性关系,但计算表明,当样本的因变量(BP模型的输出量)很分散时,由于计算机舍入误差的影响很难训练到高精度。Kohonen模型可按样本特征的相似性分群,这样,如果先用kohonen模型分群后,再用BP模型拟合各个群的样本数据,则可以得到很准确的结果。本文提出了系统的算法,并用实际配电系统作了测试。计算表明,本文提出的方法是很有效的。
In this paper, a new method is proposed to calculate the energy losses in distribution systems via two artificial neural network models, i. e., Kohonen's self- organization features mapping model (Kohonen's model) and error Back- Propagation model (BP model). BP model can map arbitrarily complex nonlinear relationship theoretically, but simulations show that when the output variables of BP model for all samples are distributed in a large domain. it is difficult to achieve accurate results only via BP model due to the computational errors of computers. While Kohonen's model can divide the samples into several clusters according to their similarities. Thus, compared with utilizing BP model only. combined use of Kohonen's model and BP model can obtain very accurate results. A systematic algorithm is presented, and tested on an actual distribution system. The simulation results show that the proposed methods is very efficient.
出处
《电工技术学报》
EI
CSCD
北大核心
1993年第3期16-20,共5页
Transactions of China Electrotechnical Society
基金
高等学校博士学科点专项科研基金
关键词
电网
损耗
计算
BP模型
: Artificial neural network Distribution System Energy losses Calculation
