Qiu Quan, Zheng Chenfei, Wang Wenping, Qiao Xiaojun, Bai He, Yu Jingquan, Shi Kai
Beijing Research Center of Intelligent Equipment for Agriculture, Beijing Academy of Agriculture and Forestry SciencesBeijing, China.
Department of Horticulture, Zhejiang UniversityHangzhou, China.
Front Plant Sci. 2017 Aug 8;8:1297. doi: 10.3389/fpls.2017.01297. eCollection 2017.
State observer is an essential component in computerized control loops for greenhouse-crop systems. However, the current accomplishments of observer modeling for greenhouse-crop systems mainly focus on mass/energy balance, ignoring physiological responses of crops. As a result, state observers for crop physiological responses are rarely developed, and control operations are typically made based on experience rather than actual crop requirements. In addition, existing observer models require a large number of parameters, leading to heavy computational load and poor application feasibility. To address these problems, we present a new state observer modeling strategy that takes both environmental information and crop physiological responses into consideration during the observer modeling process. Using greenhouse cucumber seedlings as an instance, we sample 10 physiological parameters of cucumber seedlings at different time point during the exponential growth stage, and employ them to build growth state observers together with 8 environmental parameters. Support vector machine (SVM) acts as the mathematical tool for observer modeling. Canonical correlation analysis (CCA) is used to select the dominant environmental and physiological parameters in the modeling process. With the dominant parameters, simplified observer models are built and tested. We conduct contrast experiments with different input parameter combinations on simplified and un-simplified observers. Experimental results indicate that physiological information can improve the prediction accuracies of the growth state observers. Furthermore, the simplified observer models can give equivalent or even better performance than the un-simplified ones, which verifies the feasibility of CCA. The current study can enable state observers to reflect crop requirements and make them feasible for applications with simplified shapes, which is significant for developing intelligent greenhouse control systems for modern greenhouse production.
状态观测器是温室作物系统计算机控制回路中的一个重要组成部分。然而，目前温室作物系统观测器建模的成果主要集中在质量/能量平衡方面，忽略了作物的生理反应。因此，很少开发用于作物生理反应的状态观测器，控制操作通常基于经验而非作物的实际需求。此外，现有的观测器模型需要大量参数，导致计算负荷沉重且应用可行性差。为了解决这些问题，我们提出了一种新的状态观测器建模策略，即在观测器建模过程中同时考虑环境信息和作物生理反应。以温室黄瓜幼苗为例，我们在指数生长阶段的不同时间点对黄瓜幼苗的10个生理参数进行采样，并将它们与8个环境参数一起用于构建生长状态观测器。支持向量机（SVM）作为观测器建模的数学工具。典型相关分析（CCA）用于在建模过程中选择主要的环境和生理参数。利用这些主要参数，构建并测试了简化的观测器模型。我们在简化和未简化的观测器上对不同输入参数组合进行了对比实验。实验结果表明，生理信息可以提高生长状态观测器的预测精度。此外，简化的观测器模型能够给出与未简化模型相当甚至更好的性能，这验证了CCA的可行性。当前的研究能够使状态观测器反映作物需求，并使其具有简化形式的应用可行性，这对于为现代温室生产开发智能温室控制系统具有重要意义。
