基于XGBoost的温室环境预测与卷膜决策方法研究  

Research on Greenhouse Environment Prediction and Film Rolling Decision Method Based on XGBoost Model

作　　者：李柱 朱德兰[1,2] 陆丽琼 韩煜琪 涂泓滨 刘禹晗 许彤 LI Zhu;ZHU De-lan;LU Li-qiong;HAN Yu-qi;TU Hong-bin;LIU Yu-han;XU Tong(College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling 712100,Shaanxi Province,China;Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas,Ministry of Education,Northeast A&F University,Yangling 712100,Shaanxi Province,China)

机构地区：[1]西北农林科技大学水利与建筑工程学院,陕西杨凌712100 [2]西北农林科技大学旱区农业水土工程教育部重点实验室,陕西杨凌712100

出　　处：《节水灌溉》2023年第3期67-74,共8页Water Saving Irrigation

基　　金：国家重点研发计划项目(2021YFE0103000);陕西省重点研发计划项目(2020ZDLNY01-01);宁夏回族自治区重点研发计划项目(2022BBF02026)。

摘　　要：为了实现温室大棚环境由人工管理到自动控制,将农民经验管理模式模型化、参数化,对温室大棚环境进行控制,在陕西杨凌选择管理优质的温室作为研究对象,利用机器学习的极端梯度提升算法(extreme gradient boosting,XGBoost)建立室外环境(温度、湿度和不同高度温度)、室内环境(温度、湿度)、控制(卷膜) 3者之间的关系,分别对4个不同生育期(新梢生长期、开花坐果期、果实膨大期、着色成熟期)葡萄温室的温、湿度进行模拟,并建立温室环境控制卷膜决策模型,将该模型应用于杨凌地区“锦田农庄”3号葡萄温室,实现了直接通过温室外界环境对温室卷膜进行远程控制。结果表明:与实际情况相比,模型决策准确率为95%,根据结果进行卷膜远程控制,昼间卷膜开启后,温度、湿度变化趋势缓慢,均处于目标区间,说明所建立的卷膜决策模型可以有效调控温室内的温度和湿度,能够减少温室内传感器的使用,具有较强的应用价值和推广意义。In order to realize the change of greenhouse environment management mode from manual management to automatic control, this study modeled and parameterized the farmers’ experience management mode to control the greenhouse environment. In Yangling, Shaanxi Province, a greenhouse with excellent management was selected as the research object, and the machine learning extreme gradient boosting(XGBoost) was used to establish the relationship between outdoor environment(temperature, humidity and temperature at different heights),indoor environment(temperature, humidity) and control(film rolling). The temperature and humidity of four grape greenhouses in different growth periods(new shoot growth period, flowering and fruiting period, fruit expansion period and coloring maturity period) were simulated,and the decision-making model of greenhouse environmental control film rolling was established. Then the model was applied to No.3 grape greenhouse in Jintian Farm, Yangling area, and the remote control of greenhouse film rolling was realized directly through the external environment of the greenhouse. The results show that, compared with the actual situation, the decision-making accuracy of the model is 95%.According to the results, the remote control of film rolling is conducted. After the rolling film is opened in the daytime, the change trend of temperature and humidity is slow, and they are all in the target range. It shows that the established film rolling decision model can effectively control the temperature and humidity in the greenhouse, and reduce the use of sensors, which has strong application value and popularization significance.

关 键 词：温室 环境模拟 XGBoost模型 卷膜决策 远程控制 

分 类 号：S223[农业科学—农业机械化工程] S625.51[农业科学—农业工程]