BP神经网络在日光温室湿度预测中的应用

在对冬季环境下典型北方日光温室环境因子实测数据进行分析的基础上,选择影响温室湿度的环境因子和管理情况作为神经网络的输入量,包括室外温度、室外湿度、室外光照、室内3点温度、室内光照、天窗、侧窗开闭等共10项,以温室内部5个点实测平均湿度为输出量。通过900组数据对构建好的BP神经网络进行训练,选取训练数据外的60组数据作为测试。结果表明,60组输出数据平均相对误差为3.234%,预测效果良好。

Application of Nerve Network on Forecasting Temperature in Sunlight Greenhouse

This thesis analysed the measured datas of common northern sunlight greenhouse’s environrmet factors,then took the 10 factors which are effectful to temperature and human intervention as input,such as the temperature,humidity,illuminance outside,and 3 points temperauter inside,took the average temperature value of 5 points inside the greenhouse as output.The formed BP network was trained through 900 datas,then another 60 datas were trained to test it’s performance.The result has shown that average relative error of the 60 datas was 3.234%,and the forecasting effect was great.