基于BP神经网络模型的输电线路造价预测模型研究  

Research on transmission line cost prediction model based on BP neural network

作　　者：张宇晨 张宇霖[2] 封春菲 王晨 ZHANG Yuehen;ZHANG Yulin;FEI Chunfei;WANG Chen(North China Power Engineering Co.,Ltd.of China Power Engineering Consulting Group Beijing 100120,China;HUANENG Power Int’l In c.,Beijing 100031,China;China Hua Neng Group Hong Kong Co.,Ltd.Beijing 100031,China)

机构地区：[1]中国电力工程顾问集团华北电力设计院有限公司技术经济中心,北京100120 [2]华能国际电力股份有限公司,北京100031 [3]中国华能集团香港有限公司,北京100031

出　　处：《电力大数据》2020年第6期35-42,共8页Power Systems and Big Data

摘　　要：针对输电线路工程前期工作中,由于诸多工程技术指标还未确定,工程造价难以预测的问题。本文将采用BP神经网络算法,搭建一个基于过往数据的输电线路工程造价预测模型。通过BP神经网络的强大的非线性函数拟合能力和模式分类能力,来预测新条件下输电线路的工程造价。本文在训练神经网络模型时,还结合PCA主成分分析法来对原始工程数据进行优化,减少“训练集”数据的维度,提升样本数据的特征性并在最后把预测模型与“测试集”的数据进行比较。通过比较,本文证明了基于BP神经网络的输电线路工程造价预测模型在大数据样本的训练下,能够有效拟合造价函数并且能显著降低实际造价与预测值的误差范围因此本文认为该模型的运用,能在项目前期工作中对预测工程造价地准确性产生有益的作用。In the early stage of transmission line project,it is difficult to predict the project cost because many engineering technic;al indexes have not been determined.This paper will use BP neural network algorithm to build a transmission line project cost prediction model based on past data.Through the powerful nonlinear function fitting ability and pattern classification ability of BP neural network,the project cost of transmission line under new conditions can be predicted.When training the neural network model,PCA is combined to optimize the original engineering data,reduce the dimension of the"training set"data,and improve the characteristics of the sample data.Finally,the prediction model is compared with the data of"test set".Through the analysis,this paper proves that the transmission line project cost prediction model based on BP neural network can effectively fit the cost function and significantly reduce the error range between the actual cost and the predicted value under the training of big data samples.Therefore,this paper believes that the application of this model can play a beneficial role in the accuracy of project cost prediction in the early stage of the project.

关 键 词：人工神经网 特征向量 隐藏层 数据集 非线性 激活函数 

分 类 号：TM72[电气工程—电力系统及自动化]