面向控制的温室系统小气候环境模型要求与现状

以往的温室作物生长和小气候环境模型,主要是从面向研究而不是面向实际生产的温室获得的,这二者的最大不同是:面向研究的模型主要考虑的是得到作物生长高产所需的"最优"的温室内部气候环境参数设定值,而较少考虑温室内控制设备的能力（控制动态过程）、生产过程中温室外气候变化情况和达到"最优"所需付出的能量等代价;而后者在面向实际生产的自动化控制的温室系统模型中是必不可少的。当前温室系统自动化控制面临的一个最大困难,就是缺乏一个这样的可靠的温室系统模型,而只能采用面向研究的温室系统模型去进行实际生产的温室系统控制,这种忽视实际生产条件下的温室系统模型与理想条件下的模型之间差异的"纸上谈兵"的做法,必然导致温室控制技术水平低、达不到预期效果。该文介绍了温室系统的整个控制过程,对一个实际生产的温室系统中各种变量和参数作了简要描述,并概括了面向实际的温室生产控制要求的温室系统模型的基本结构,对温室环境模型、作物生长模型和能耗及CO₂消耗模型的研究现状作了详细的回顾。从满足控制需求出发对现有的温室系统模型所存在的问题进行了分析,并指出了其中的不足和局限性。探讨了未来温室系统的建模方法和需要解决的关键问题,提出了面向控制需求的温室系统建模要满足的要求,为温室系统的建模研究提供了一种新的思路和方向。

Requirement and current situation of control-oriented microclimate environmental model in greenhouse system

Abstract: Economy-based optimal control of greenhouses is an important technique to reduce the operating cost and increase the crop yield. In General, the structure of a typical greenhouse control system consists of two layers: the objective optimization layer and the process control layer. The aim of the former is to obtain the target trajectories of environmental states; while in the latter, environmental states are tuned to track those obtained trajectories. Based on this framework, some relevant models to the greenhouse system, such as greenhouse microclimate model, crop growth and yield model, energy consumption predicting model and CO2 consumption predicting model, need to be built, and some constraints over environmental states and control inputs must be determined. However, most greenhouse climate models and crop growth models proposed in literature are research-oriented rather than based on practical cultivation. The biggest difference between the two is that research-oriented models solely focus on the optimal set-points of internal climatic parameters for maximum crop yield, ignoring the abilities of control actuators, the ambient climate change, and the overall energy consumption. From a practical point of view, the latters are certainly necessary in greenhouse models. It would be difficult to achieve realistic results if the models are incomplete. The lack of reliable greenhouse models has become the greatest difficulty for greenhouse optimization and control. In this paper, the latest trends in greenhouse climate models, crops growth models, energy consumption models, and CO2 consumption models are reviewed in details. The main shortcomings of current models can be summarized as follows: (1) although some models obtained by ample mechanisms, their structures are excessively complicated, such that the corresponding computations are very expensive, which makes it difficult to design an efficient controller based on them. This class of models interprets the real physical laws with large number of parameters and complex structures, and Vanthoor's model and TOMGRO are examples; (2) some models are too simple to accurately reflect the relationships between greenhouse environment and crop growth. Generally, only the dynamics of air temperature and humidity in greenhouses are described in this class of greenhouse climate model, such as Albright's model, while the dynamic of CO2 concentration, which is an important environmental factor to affect photosynthesis, is not included. Furthermore, the influence to crop growth of environmental factors is always partially reflected in simplified crop growth models, e.g. in radiation and thermal effectiveness model, only air temperature and radiation is used to describe the accumulation and allocation of dry matter; (3) only single control input variable is included in many greenhouse climate models, such as the thermal environmental model of solar greenhouse, this is because most greenhouses are not equipped with relevant control actuators. Generally, these greenhouse climate models are used to guide the structure design, material selection, or management of greenhouse production. Additionally, although some greenhouse climate models include various control inputs, including heating, fogging, ventilation and CO2 injection, the dynamic response of control actuators are still not described adequately, and so, they can't be used for greenhouse climate control. In order to obtain a suite of efficient greenhouse models, great efforts need to be made to solve the following key problems: (1) the unknown mechanisms of some processes need to be explored further for a mathematical expression of their input-output relationship. Provided with these results, some complex models can be improved; (2) based on a complex model with detailed mechanistic expressions, some dynamic sub-processes can be reconstructed or simplified to reduce the model complexity; (3) new approaches and theories about system modeling and model validation can be developed.