基于改进神经网络的用电客户信用评价  

Power consumer credit assessment based on improved neural networks

作　　者：李伟[1] 杨照芬[1] 牛东晓[1] 

机构地区：[1]华北电力大学经济管理系,河北保定071003

出　　处：《华东电力》2007年第4期9-12,共4页East China Electric Power

基　　金：国家自然科学基金资助项目(50077007)资助;高等学校博士点专项基金(20040079008)资助

摘　　要：在分析影响用电客户信用因素的基础上,建立了一套适用于用电客户信用评价的指标体系,采用BP神经网络建立用电客户信用评价模型,用遗传算法优化BP神经网络的连接权重和阈值,解决了BP神经网络存在落入局部最小点和收敛速度慢的问题,两者结合,实现了优势互补。实例研究表明,评价值与实际值相差较小,遗传神经网络的评价结果是令人满意的。The power consumer credit assessment has raised close attention among power supply companies and the society as whole. Based on the analysis of factors affecting power consumer credit, an index system suitable for power consumer credit assessment was constructed. The BP neural network was used to establish the model for power consumer credit assessment, and the genetic algorithm was adopted to optimize the connection weights and threshold values of the BP neural network. The problems of trapping into local minimum and low convergence speed of the BP neural network are solved. Practice shows that the assessment conducted by the genetic neural network is satisfactory with difference between the assessment value and the actual one being little.

关 键 词：遗传算法 BP神经网络 用电客户 指标体系 信用评价 

分 类 号：TM73[电气工程—电力系统及自动化]