摘要
为更全面、准确地评估10kV配电网线损水平,提出了一种基于粒子群算法(PSO)优化BP神经网络(BPNN)的较为准确有效的10kV配电网理论线损预测方法。首先筛选和构建5个电气特征指标描述10kV配电网结构和运行状态;其次采用惯性因子和学习因子动态调整的粒子群算法,全局搜索BP神经网络的权值和阈值来构建PSO-BPNN线损评估模型;通过对训练样本集的学习,拟合电气特征指标与线损之间的非线性关系,进而对测试样本集线损进行预测。最后应用某地区10kV配电网的实际样本数据验证了所提方法的有效性与合理性。
To estimate the level of 10 kV distribution network line losses more integrally and precisely,an assessment method based on BP neural network(BPNN)improved by particle swarm optimization(PSO)is proposed.Firstly,5 electrical characteristic indexes are selected to reflect the structure and operation state of 10 kV distribution network.Secondly,the inertia weight and the acceleration coefficient of PSO are dynamically adjusted so that the weights and biases of BPNN can be searched more effectively.Then,nonlinear relation between electrical characteristic indexes and line losses is fitted through the learning of training sample sets to predict the line losses of test sample sets.Finally,the PSO-BPNN is proved to be effective and proper through an actual 10 kV distribution network sample data.
作者
刘丽平
张东霞
孙云超
张义涛
王新迎
LIU Liping;ZHANG Dongxia;SUN Yunchao;ZHANG Yitao;WANG Xinying(China Electric Power Research Institute,Beijing 100192,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China)
出处
《供用电》
2018年第7期43-48,共6页
Distribution & Utilization
基金
国家电网公司科技项目"面向同期线损管理的多专业数据治理技术与挖掘应用研究"(XT71-17-027)~~
关键词
配电网
理论线损
BP神经网络
粒子群算法
开集测试
distribution network
theoretical line losses
BP neural network
particle swarm optimization
open-set test
