具有区间输出约束的温室环境系统多变量优化控制算法的研究

Conventional methods used for solving greenhouse environment control problems lay excessive emphasis on control performance such as fixed set points control. The resulting solution will cause: (1) higher energy cost and (2) no solutions for the corresponding controllers owing to the existing strong coupling environmental factors and constrains. Thus, it may be more reasonable to adopt zone control objectives instead of point control objectives, because it can guaranteed the existence of controllers, and can provide ample range for the optimization of economic index when controllers are designed.This research project is on multivariable optimization and control algorithms based on the above-mentioned idea. First, the quantitative relationship between environmental states and control variables should be clearly analyzed, and an input-output greenhouse climate dynamic model for control and optimization should be created. And some simulation optimization methods should be developed to solve the optimal control objective zone. Then, combining identification, control and decoupling for the controlled process, some algorithms with the requirement of output zone constraints, such as multivariable adaptive control algorithms, multivariable adaptive decouple control algorithms and based multi-objective optimization multivariable control algorithm, will be presented by employing neural networks, fuzzy systems and multi-objective optimization, etc.. Finally, the achieved algorithms will be applied in the practical greenhouse energy-saving production for testing and certification their validities. And the ultimate goal is to achieve some new effective measures for the practical greenhouse production.

在温室生产中，传统的过分强调控制性能的精准控制模式往往存在两方面问题：（1）导致大量的能源消耗和生产成本的急剧上升；（2）因存在诸多约束及环境因子间的强耦合作用而导致无法确保各环境因子控制目标同时存在可行解。为此，本项目设想通过放宽环境因子的目标控制精度，以"区间"控制目标替代传统的"点"控制目标，通过对多变量优化控制算法的研究以解决上述难题。为此，项目首先拟在已有的温室环境模型基础上，进一步明确环境状态与控制变量、作物生长之间相互作用的定量关系，建立面向优化与控制的温室环境（多因子）输入输出动态模型；研究模拟优化方法，以此确定最佳的控制目标区间；并利用神经网络、模糊系统、多目标优化等技术，将被控对象的辨识、控制和解耦结合起来，进而提出具有区间输出要求的多变量自适应控制算法、解耦控制算法以及基于多目标优化的多变量控制方法，最后，在温室生产的实际控制问题中进行工程验证，最终得到工程新方法。

设想放宽环境因子的目标控制精度，以“区间”控制目标替代传统的“点”控制目标，显然会使得各目标因子同时获得可行解的概率大增，从而确保控制器（可行解）存在性问题，同时又能为控制器提供经济优化的自由度，为降低生产成本、满足用户节能增效的目标要求提供了广阔的调节余地。本项目从温室系统建模出发，研究了环境状态与控制变量、作物生长之间相互作用的定量关系，建立了面向优化与控制的温室环境（多因子）输入输出动态模型；研究了各种优化控制方法，利用神经网络、模糊系统、多目标优化等技术，将被控对象的辨识、控制和解耦结合起来，并提出了具有区间输出要求的多变量自适应控制算法、解耦控制算法以及基于多目标优化的多变量控制方法；最后，将得到的算法在温室生产的实际控制问题中进行工程验证，效果良好。