HVAC系统中蒸发器的建模与仿真

【作者】 于海磊；

【导师】 贾磊； 王雷；

【作者基本信息】 山东大学 ， 检测技术与自动化装置， 2006， 硕士

【摘要】 中央空调系统(HVAC)是一个典型的复杂非线性热力系统，随着能源的日益紧张，对其数学模型和优化节能控制的研究，已经逐渐成为一个热点研究问题。特别是对于中国能源相对缺乏的现状，对这种大耗能系统的建模和控制进行深入研究，以达到节能的目的，就更具有实际意义。 传统的HVAC系统建模是利用数学机理建模法，但由于所建模型一般较复杂，难以用于实际HVAC系统的控制。本文利用神经网络辨识原理和多种改进的BP算法，对HVAC系统的一个重要部件——蒸发器进行研究，面向实际应用进行了智能仿真建模。 1)介绍了HVAC系统的基本工作原理和蒸发器的经典的数学建模仿真方法：有限差分模型、集中参数模型和分布参数模型的数学机理，以及国内外学者的研究概况。 2)论述了整个HVAC系统的热力学工作原理以及工作过程中空气和制冷剂的温度、压力、焓值等状态参数的变化情况。采用经典的数学机理建模法，确立了蒸发器的输入输出参数，建立了蒸发器的分布参数模型；并编写了可以方便输入各种参数的仿真界面，对实验采集数据和仿真数据机进行了对比分析。 3)介绍了神经网络的发展史以及其在辨识建模、智能仿真、预测等领域的广泛应用，指出了由于神经网络特有的非线性适应信息处理能力而在非线性复杂系统辨识建模方面的独特性和优越性。 4)采用BP神经网络及其算法在神经网络辨识上的应用，以采集的实验室蒸发器的输入输出数据为基础，使用数学工具MATLAB6.5，编写了蒸发器的BP神经网络辨识仿真程序。给出了相应的仿真曲线和预测结果，并与原始数据进行了比较和分析，得到较为满意的效果。然后采用多种改进的BP算法，进一步改善网络训练误差的精度和网络训练的速度，建立了训练速度快、仿真效果好的蒸发器神经网络辨识模型。 5)通过对比经典模型和神经网络仿真模型，对它们各自的预测结果和仿真曲线进行分析，综合考虑各方面实际因素，可以看出更多还原

【Abstract】 Heating, Ventilation, and Air-Conditioning (HVAC) systems are complex and non-linear. With the increasingly lacking of energy sources, the study about their math model and optimized control was gradually concerned by researchers. Especially for china, which was highly short of energy source, it was truly necessary to lucubrate the modeling and control of the systems that consumed large energy sources.The conventional modeling method in HVAC systems was based on mathematics mechanism, which was complex and difficultly applied in project. In this paper, based on the neural network identify and ameliorated BP arithmetic, an evaporator model was simulated which was aimed at the practicality apply.1) The paper introduced the working process of HVAC systems and the classical math model and simulation about evaporator, such as finite difference model, lumped parameters model and distributing parameter model. The work of researchers was listed, including abroad and at home.2) Discussed the working principle of thermal and environmental variables such as temperature, pressure, enthalpy, etc. Firstly, the input and output parameters of the evaporator were radicated, and the distributing parameter evaporator model was established by classical math mechanism method. Then the emluator interface was compiled, which can be conveniently inputted the parameters. Further more, the data between experiment and simulation was compared and analyzed.3) Neural Network was introduced, which was successfully applied in the field of identification, intelligent simulation predicts and so on. Because of Neural Network’s unique features in dealing and adapting to non-linear information, it was particular and更多还原