融合机理模型与数据模型的温室环境自适应鲁棒控制方法研究

The microclimate environment in greenhouse is the key to the efficient regulation of facility agriculture. This project takes the adaptive adjustment of model parameters and the real-time update of hybrid model as the research focus, solves the key problems of accurately extracting the spatio-temporal characteristic parameters of the greenhouse environment and the robust adaptive optimization of the model, which is the frontier research direction for realizing the dynamic regulation of crops on demand and scientific issues in the field of environmental regulation. Taking tomato as an example, this paper intends to use time series analysis, deep learning algorithm combined with robust control to construct the greenhouse spatial-temporal parameter adaptive update control model as the entry point. Focus on the following research: (1) A multi-factor coupled environment dynamic mixture model combining mechanism model and data-driven model is established to estimate and adjust the parameter state of the model; (2) The model update mechanism of greenhouse spatial-temporal feature parameter extraction and compensation is proposed to overcome the missing model parameters and the measurable uncontrollable disturbance problem in the system; (3) Based on the interaction between greenhouse environmental factors and crop growth status, a robust adaptive control algorithm was designed, and the adaptive parameters were used to estimate the uncertain parameters online to realize the real-time update of the hybrid model. The research results of the project provide theoretical basis and technical support for clarifying the intelligent and precise regulation of greenhouse microclimate associated with multiple environmental factors, which has important theoretical research significance and broad application prospects.

温室环境模型是设施小气候高效调控的基础，本项目以模型参数自适应调整、混合模型实时更新为研究重点，解决准确提取温室环境时空特征参数和模型鲁棒自适应优化等关键问题，是实现作物按需动态调控的前沿研究方向，是温室环境调控领域亟待解决的科学问题。本课题以温室时空参数自适应更新调控模型构建为切入点，以番茄为例，采用时序分析、深度学习和鲁棒控制相结合，重点开展如下研究：①针对模型参数的时变和不确定性，建立机理模型和数据驱动模型相结合的多因子耦合环境动态混合模型，对模型的参数状态进行实时调整；②提出温室时空特征参数提取及补偿的模型更新机制，解决模型参数缺失和系统中存在的可测不可控的扰动问题；③针对温室环境因素与作物生长状态的互作影响，设计鲁棒自适应控制算法，实现混合模型实时更新。项目研究结果为多环境因子关联的温室小气候智能精准调控提供先进的理论基础和关键技术支撑，具有重要的理论研究意义和广阔的应用前景。

温室环境模型是设施小气候高效调控的基础，本项目以模型参数自适应调整、混合模型实时更新为研究重点，解决准确提取温室环境时空特征参数和模型鲁棒自适应优化等关键问题，是实现作物按需动态调控的前沿研究方向，是温室环境调控领域亟待解决的科学问题。本课题以温室时空参数自适应更新调控模型构建为切入点，以番茄为例，采用时序分析、深度学习和鲁棒控制相结合，开展了如下研究：建立机理模型和数据模型相融合的混合模型，引入模型更新理论，在探究温室内部环境时空特征参数分布规律、温室内外环境参数相互作用关系基础上，提出参数自适应更新机制和鲁棒优化控制方法；设计一种鲁棒自适应控制律，当模型存在不确定或受到外界干扰产生较大误差时，采用自适应策略在线估计不确定参数，利用鲁棒控制策略对外界干扰进行补偿，实现模型随不同温室结构、材料及作物生长状态等差异性和动态因子自适应调整；在上述理论研究基础上，研发多元实时感知信息的温室环境远程测试平台，进行模型与方法试验验证。项目研究结果为多环境因子关联的温室小气候智能精准调控提供先进的理论基础和关键技术支撑，具有重要的理论研究意义和广阔的应用前景。.在本项目的资助下，研究小组在《Computers and Electronics in Agriculture》、《International Journal of Agricultural and Biological Engineering》、《Applied Engineering in Agriculture》、《Phyton-International Journal of Experimental Botany》等国际重要学术期刊和会议中发表基金标注SCI/EI检索论文6篇，其中SCI检索期刊论文4篇，授权国家发明专利1项，培养硕士研究生7名。