设施园艺光环境多因子耦合智能调控模型与方法研究

Creating a favorable light environment for crop growth is an important issue in facility horticulture development. The correlation effluences between crop biological condition and multiple external environment factors are the keys in optimal regulation. In this project, based on intelligent control theory, a multi-factor coupling regulation model and method for facility horticulture light environment adjustment are adopted. First, main affecting factors of crop photosynthetic rate are analyzed, and a multistage and multidimensional large-scale sample data is obtained through nested tests. The multi-factor coupling crop photosynthetic rate modeling method based on genetic neural network are implemented, taking the photosynthetic rate model in facility tomato planting process as the research example. In order to get the optimal photosynthetic rate, the photosynthetic rate optimization algorithm under multi-factor constraints are also studied based on genetic ant colony algorithm. According to the BP neural network, the light environment control factors model are established. The quantitative control algorithm and closed-loop control mechanism of light environment parameters are researched under dynamic multi-factor environment. Finally, based on multi-sensor fusion, an intelligent regulation method for facility horticulture light environment is proposed. The verification and perfection of the models and methods are validated through facility tomato planting process. The research results can be of important significance to understand the multi-factor coupling light environment regulation principle and method, and provide theoretical foundation for improving light environment regulation efficiency and the rapid development of modern facility horticulture.

营造有利作物生长的光环境是设施园艺发展亟待解决的重要问题，其优化调控的难点在于需考虑作物生理与外部环境等多因子关联的影响，拟基于智能控制理论进行设施园艺光环境多因子耦合智能调控模型与方法研究。分析影响作物光合速率的主要因素，设计多阶段多因子嵌套试验获取多维大样本数据，探寻基于遗传神经网络的多因子耦合作物光合速率建模方法，以设施番茄为对象构建光合速率模型实例；以光合速率最优为目标，研究多因子约束条件下基于遗传蚁群的光合速率寻优算法，进而建立基于BP神经网络的光环境调控目标值模型；研究多因子动态变化条件下光环境参数动态定量调控算法与闭环调控机制，提出基于多传感器融合的分阶段设施光环境智能调控方法，优化模型与方法验证平台，以设施番茄为对象开展验证与完善。研究成果对阐明多因子关联影响下的光环境调控原理具有重要意义，为提高光环境调控效率，推动现代设施园艺快速发展奠定基础。

现有补光系统缺乏定量生理依据，控制方法简单粗放，易导致补光不足或过量等问题。本文针对作物光合作用受多种关键环境因子交叉影响，导致光环境调控目标参数随关键因子变化而具有很强动态性等难题，在分析环境对光合速率影响基础上，研究多因子关联的作物光合速率模型建模方法、调控目标值动态寻优方法，设计基于多传感器信息融合和无线传感器网络体系架构的光环境调控系统，并进行试验验证。主要研究内容及结论如下。.（1）研究了基于遗传神经网络的光合速率模型建模方法。该方法在构建神经网络预测模型时，采用遗传算法对其初始权值矩阵进行优化，与随机生成权值矩阵方法相比，网络收敛速度明显提高。针对番茄幼苗光合速率模型实例进行异校验，结果表明，遗传神经网络模型预测值和实测值决定系数为0.9856，直线斜率是0.994，截距是-0.01092，误差在0.1~0.4μmol/m2s范围内，均明显优于未优化神经网络模型。以上研究表明，遗传神经网络建模方法可生成高精度光合速率模型，为设施光环境调控提供了描述作物光合需求的定量生理模型。.（2）提出了基于人工智能算法的光环境调控目标值寻优方法，实现了以温度、CO2浓度等为参数的光饱和点动态计算。其中基于遗传神经网络光合速率模型的异校验结果表明，模型预测值与实测值最大相对误差不超过6%；基于改进型群智算法客服遗传算法全局搜索能力强但局部寻优能力较弱的不足，模型预测值与实测值相关性拟合决定系数为0.976，表明二者之间高度线性相关，最大相对误差不超过2%。.（3）研发了基于多传感线融合的光环境调控系统。在分析整理栽培过程光环境调控需求的基础上，提出多传感器信息融合的闭环反馈控制机制，集成无线传感器网络系统架构和光环境调控目标值模型，构建了设施光环境调控系统，实现了设施内部光子通量密度、CO2浓度、温度等环境信息实时监测，调控目标值自动计算与各补光节点动态变量控制，在满足作物需求条件下增产降低30%电能消耗，番茄增产可达30%以上。