【作者】 张军；

【导师】 张侃谕；

【作者基本信息】 上海大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 温室节能控制的首要问题是如何建立温室的综合能耗模型。温室的培育对象——农作物，是有生命的，其生理响应和生态过程难于检测，温室环境影响植物的生长过程，反过来，植物的生理作用，如蒸腾作用、光合作用、呼吸作用等又影响温室环境，因此，对温室进行环境控制难度较大。温室生产过程运行机理复杂，非线性、时变性、强耦合性、多干扰性、不确定性、不精确性、时滞性严重，而温室中的能量关系更加复杂，为此，要找到一种更为有效的建模方法，用于温室节能目标与设备运行状态之间关系的描述，从而为节能优化控制创造条件，为实现面向节能减排降耗的温室环境系统的优化控制奠定精确的模型基础。本研究用智能集成建模的方式建立温室的能耗模型，并以此智能集成模型作为控制对象，实现温室的节能优化控制。对温室环境优化控制策略进行探讨和研究，通过控制温室环境因子（温度、湿度和CO₂浓度）以使植物良好生长，同时尽可能降低能耗，以达到增产、节能、减排、降耗、增加种植者收益的目的。为实现上述目标，本文进行了如下几个方面的研究和探索：1.建模方法方面：提出了基于条件熵的GM（1,1）和GM（1,N）智能集成灰色预测建模，可有效应对农作物难以检测的生理响应和生态过程等植物特有特性。建立了温室环境系统的过程神经元网络模型，实现了对温室环境系统所有历史数据和实时数据的综合应用。由于受到输入的同步瞬时限制,传统人工神经网络难以表达时间序列中实际存在的时间累积效应,并且传统人工神经网络难以解决较大样本的学习和泛化问题,因此,传统人工神经网络在解决复杂非线性时间序列预测问题时还存在一定的局限性。本文采用过程神经元网络，网络的输入输出是过程或时变函数，放宽了传统神经元网络模型对输入的同步瞬时限制,解决了传统神经网络存在的上述问题。提出了基于信息熵的智能集成模型参数优化辨识方法，充分利用温室数据中更有价值的信息。在辨识机理分析模型的参数时，不是采用某种单一的辨识算法，而是将几种模型参数辨识算法有机集成起来，充分利用尽可能多的数据信息，发挥各类优化算法的长处，避免短处，使机理分析模型的参数尽可能最优。建立了温室系统基于灰色预测补偿的机理模型,对温室环境系统的不确定性、不精确性、多干扰性、农作物生理响应和生态过程难以检测的特性等进行了有效补偿。提出了基于信息熵的智能集成建模思想，利用信息熵对以上各类模型进行智能集成，充分利用每种模型中更有价值的信息。该集成算法能充分利用所采用的各种单独智能优化算法的所有有价值信息，能充分利用温室系统的所有历史数据所包涵的有价值信息。实现了将温室环境系统所有历史数据包含和隐藏的一切历史信息以及当前信息全部应用，让其在建模和控制中全部发挥出价值。提出了基于免疫优化算法的自校正广义方差最小二乘辨识、神经网络、最小二乘支持向量机的智能集成建模方法，在全封闭温室中央空调系统的建模中实现了效果良好的应用。将上述建模方法其应用到玻璃温室及全封闭温室中央空调系统的建模和能耗预测中，为实现节能优化控制奠定有效的准确的模型基础。上述建模方法能有效应对温室对象的特殊性：农作物是有生命的，其生理响应和生态过程难于检测；温室环境影响植物的生长，而植物的生理作用如蒸腾作用、光合作用、呼吸作用等又反过来影响温室环境。这种特殊性导致温室环境控制难度大，现有建模方法控制效果不理想，本文提出的建模思想和建模方法效果满意。智能集成建模的精度指标为：模型绝对误差均值0.166330255、相对误差均值-0.12%、RMSE平均为0.453116572、RE平均为3.72%、模型预测数据与温室对象实测数据相关系数为0.9331535⁰.97287241、决定系数为0.870858467⁰.952664962，这些指标表明提出的温室智能集成建模方法是有效的和准确的。2.优化控制策略方面：提出了基于TOPSIS策略的智能集成节能优化控制策略，该策略不是仅仅把能耗最低作为唯一评价指标，不是以牺牲温室环境因子为代价而实现节能，而是对温室系统各种优化控制策略下的能耗情况、最适合作物生长的环境情况、算法的迭代次数、算法的收敛速度、算法的执行时间进行综合考虑，在充分满足多个评价指标的情况下实现节能，避免出现顾此失彼的情况，以便得到综合效益最好的节能优化控制策略。将遗传算法、粒子群算法、标准模拟退火算法、改进模拟退火算法一、改进模拟退火算法二分别应用于玻璃温室的节能优化控制，并用TOPSIS策略实现了这五种优化控制算法的智能集成节能优化控制。将遗传算法、模拟退火算法、改进模拟退火算法一分别应用于全封闭温室的智能节能优化控制，并用TOPSIS策略实现了这三种优化控制算法的智能集成节能优化控制。分别采用捕食搜索算法、禁忌搜索算法、改进模拟退火算法一、标准粒子群算法、改进粒子群算法对多台冷水机组的负荷优化分配问题进行了研究。将上述控制策略应用到了玻璃温室和全封闭温室的节能减排优化控制中，节能效果满意：单一智能优化控制策略节能率为最小10.80%、最大35.75%、平均20.93%；智能集成优化控制策略的节能率为最小37.32%%、最大44.19%、平均40.67%。3.优化算法方面：用上述多种标准智能优化算法实现了温室环境系统的建模和优化控制；对标准模拟退换算法进行了改进，提出了改进模拟退火算法一：变异操作变搜索空间单循环SA算法；提出了改进模拟退火算法二：混沌遍历搜索特殊算法确定初始温度增设方差判定准则作为停止条件的SA算法；对标准粒子群优化算法进行了改进。将上述多种智能优化算法及其改进算法在温室环境系统智能集成建模和智能集成节能减排优化控制中应用。4.其他方面：将统计学理论与方法应用在建模和控制领域，并对预测值与实测值进行统计学分析。对所建立的各种智能集成模型的有效性和准确性进行了统计学分析。分析数据均表明：模型准确，精度较高。对每种智能集成优化控制策略的控制效果进行了节能分析，也进行了减排分析。对温室常规的燃煤加热方式与地源热泵加热方式进行了经济性分析和减排分析。更多还原

【Abstract】 The primary problem of energy conservation control in greenhouse is how tobuild a greenhouse energy model. Crops, the protagonist of greenhouse,are alive,and it is difficult to detect its physiological processes and ecological processes.Environment affects the growth of greenhouse plants, and the physiological role ofplants,such as transpiration, photosynthesis, respiration, etc., affects thegreenhouse environment smultaneously. So,it is difficult to control the greenhouseenvironment. Greenhouse production process is complicated,with nonlinear,time-varying, strong coupling, multi-interference, uncertainty, imprecision, serioustime lag, and the energy transfer relationship in each part of the greenhouse ismore complex. To solve this problem, it is nessary to find a more effective methodfor modeling, used for energy conservation of the greenhouse.In this paper, the intelligent integrated modeling approach is explored to buildenergy consumption model of the greenhouse for energy conservation. Optimalcontrol strategies for greenhouse environment is explored by controlling thegreenhouse environment factors so that plants grows well in the greenhouse andenergy consumption reduces.The following aspects in the intelligent integrated modeling and intelligentintegrated optimization control of greenhouse is researched or investigated:1. Modeling methods:Conditional entropy-based intelligent integration of grey prediction modelingis proposed which can effectively deal with the plant-specific features dfficult tobe detected such as the physiological response and ecological processes of thegreenhouse plants.Process neural network model of the greenhouse environment system isestablished,which can use both the real-time data and the historical data of thegreenhouse, overcoming the limitations of traditional neural networks such aslarge sample study and generalization.Information entropy-based intelligent integrated optimized model parametersidentification is proposed, making full use of the more valuable information ingreenhouse data.Grey prediction compensation-based mechanism model of the greenhouse environment is established to compensate effectively the uncertainty, imprecision,and more disturbing of the greenhouse and the crop physiological responses andecological processes difficult to detected.Intelligent integrated modeling ideas based on information entropy isproposed to use more valuable information in each single model approach.Self-correcting generalized variance least squares identification model, neuralnetwork model, least squares support vector machines model are established forthe central air conditioning system of closed greenhouse. On the basis of the abovethree models,Intelligent integration model based on immune optimizationalgorithm is established with Satisfactory.Modeling idea and modeling methods described above works well with goodresults: mean absolute error is0.166330255, mean relative error is-0.12%,meanRMSE is0.453116572,mean RE is3.72%,the correlation coefficient betweenForecast data of the model and measured data of greenhouse is0.9331535~0.97287241, The coefficient of determination between them is0.870858467~0.952664962. These indicators show that the proposed intelligentintegrated modeling of the greenhouse is effective and accurate.2. Optimized control strategy:An energy-optimized intelligent integrated control strategy based on TOPSISis proposed, which do not put energy as the only evaluation, but consider a varietyof factors such as energy consumption,environmental conditions,algorithmiterations, etc, meeting the requirements of a number of evaluation fully, to obtainan energy-optimized control strategy with good comprehensive benefits.By using the TOPSIS strategies to integrate genetic algorithms, particleswarm optimization, the standard simulated annealing algorithm, an improvedsimulated annealing algorithm, improved simulated annealing algorithm two toachieve the purpose of saving energy,which is used in glass greenhouse for energysaving.Integration algorithm by the constitution of genetic algorithms, simulatedannealing algorithm, the improved simulated annealing algorithm based on theTOPSIS strategy is used in closed greenhouse.Predatory search algorithm, tabu search algorithm, an improved simulatedannealing algorithm, the standard particle swarm optimization, improved particle swarm optimization is applied to multiple chillers for the purpose of energysaving.The saving rates of single intelligent optimization control strategy when usedin greenhouse is minimum10.80%,maximum35.75%,average20.93%. The savingrates of all these intelligent integrated control strategy above-mentioned are:minimum of37.32%,maximum of44.19%, with an average of40.67%while usedin greenhouse.3.Optimization algorithms:The above variety of standard intelligent optimization algorithms has used inthe greenhouse for the purpose of energy conservation.Two improved simulation algorithm were proposed, standard particle swarmoptimization algorithm was improved.The above mentioned intelligent optimization algorithms and improvedalgorithm is used in the greenhouse for energy saving and emission reduction.4. Other aspects:The statistical theory and methods were used in the field of modeling andcontrol, the forecast data and measured data were statistically analyzed. Emissionswere analyzed for each intelligent integrated optimization control strategy.Economic analysis and mitigation analysis had studied for the greenhouseheating means of conventional coal-fired and ground source heat pump.更多还原