摘要
针对超高压输电线路可听噪声BP网络预测模型影响因素多的问题,运用主成分分析算法(PCA)对影响可听噪声的环境因素、地理参数、导线结构参数等14个因素进行简化,建立PCA-BP网络预测模型。选取甘肃省内多条750kV、330kV输电线路的可听噪声的实测资料为样本集,采用Matlab神经网络工具箱进行模型训练与预测,并与BP网络模型预测结果比较。结果表明:主成分分析方法在可听噪声影响因素的简化中不适用,预测结果没有BP网络模型预测结果理想。分析了主成分在可听噪声影响因素简化中不适用的原因。
Aimed at solving the multiple-influence-factor problem in the audible noise BP network predictive model on EHVtransmission line,14 factors,which affect the value of audible noise,such as environment,geographical parameter and conduc-tor structure are simplified using Principal Component Analysis（PCA）and a PCA-BP network predictive model is established.Based on measured data of audible noise from 330 kV and 750 kV EHV transmission lines in Gansu province,the princi-pal-component-based model is trained and predicted with Matlab neural network toolbox,and compared with the predictive re-sult of BP neural network model.Results show that PCA is not suitable for the simplification of audible noise influence fac-tors and its predictive ability is worse than that of the BP network,and possible reasons of the PCA failure are pointed out
出处
《计算机工程与应用》
CSCD
北大核心
2011年第14期233-236,241,共5页
Computer Engineering and Applications
基金
甘肃电力科学研究院项目(NoH14200906)
关键词
主成分分析(PCA)
BP神经网络
超高压输电线路
可听噪声预测
principal component analysis
BP neural network
Extra High Voltage（EHV） transmission line
audible noise prediction
