BP神经网络在废气测量中的应用  

Application of BP Neural Network in Harmful Gas Parameter Measurement

作　　者：黄建清[1] 朱伟兴[2] 李丽[1] 

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013 [2]江苏大学江苏省现代农业装备与技术重点实验室,江苏镇江212013

出　　处：《农机化研究》2009年第4期191-195,共5页Journal of Agricultural Mechanization Research

基　　金：江苏省国际合作项目(BZ2005035)

摘　　要：针对禽畜养殖场内传感器故障引起有害气体浓度测量数据出现缺失的问题,应用改进的BP算法,构造3层前馈BP神经网络模型,并采用VB编程开发了基于BP神经网络的废气浓度估算系统。该系统具有估算精度高、运行稳定和使用方便等特点,能训练样本数据、估算测试数据和绘制各类相关曲线。用某养殖场3天的实验数据测试系统,训练误差为2.47%,估算缺失数据的最大相对误差为9.9%,最小相对误差为0.07%,平均相对误差为4.5%。系统不仅能仿真应用,而且能直接为有害气体排放总量的计算程序提供有效估算数据,保证了计算的完整性和准确性,具有一定的推广价值。Through application of improved BP algorithm, a three - layer feedforward network is built in order to solve the missed data problem of harmful gas concentration caused on BP algorithm to estimate harmful gas concentration is by sensor fault in livestock breeding farms. And a System based compiled with VB. The system can accurately estimate missing data, run steadly, be used easily, train sample data, and map various related curves. After the system is tested with three days＇ experimental data in a farm, the results show that the training error is 2.47 percent, and the biggest relative error, the smallest relative error and the average relative error of estimation missing data is 9.9 percent, 0.07 percent and 4.5 percent. The system can not only simulate, but also can provide directly effective estimation data for the calculation program of total harmful gas emission amount, have value to popularize.

关 键 词：神经网络 废气测量 禽畜养殖场 有害气体 

分 类 号：S815.9[农业科学—畜牧学] TP183[农业科学—畜牧兽医]