【作者】 王美霞；

【导师】 周强泰；

【作者基本信息】 东南大学 ， 热能工程， 2005， 博士

【摘要】 建立了完整的小型空调系统的动态仿真模型。以稳态集中参数法建立了压缩机模型;以分区集中参数法建立了蒸发器和冷凝器模型,并在模型中考虑了蒸发器管内流型及换热系数的变化;以稳态集中参数法建立了电子膨胀阀模型。将Origin自定义函数拟合方法应用到制冷剂热力性质计算中。对R22替代工质R410A和R407C的热力性质进行简化计算。计算以Cleland简化模型为基础,采用显式拟合方法对R410A和R407C热力性质进行拟合计算,将计算结果和数据源进行比较,误差控制在允许范围内。通过对替代工质R410A和R407C的热力性质拟合结果的误差分析,证明了选用的模型和拟合方法的准确性,为以后研究新工质热力性质提供了新的思路。针对以前模型中存在的问题,建立以毛细管流动特性仿真为目的的毛细管流动数学模型,模型以Stocker’s模型为基础,考虑入口渐缩、过冷和旋转直径对流动的影响。模型不仅可以用于R22,还可以应用于替代工质R410A和R407C。根据所建模型,利用C语言编写计算机程序对绝热毛细管内的流体流动进行模拟计算。模型中考虑了不同摩擦系数公式和粘度公式对计算结果的影响,在实际计算中,需要根据最优的计算结果进行摩擦系数公式和两相粘度公式的选择。从数值角度分析,这种选优的做法实际上是通过摩擦系数公式和两相粘度公式对毛细管模型的计算误差进行修正。在毛细管质量流量的仿真计算过程中,改变以往通过增加步长从而增加工质流量的算法,采用迭代算法进行计算,将弦截法编程计算方程的根的方法引入到毛细管流量计算的程序中,由于误差限ε改变及时和收敛计算方向确定,提高了模型仿真计算的精度和速度。仿真结果与实验数据的吻合验证了算法的准确性。对蒸发器过热度控制进行了PID、模糊控制和神经网络控制研究。针对目前PID控制中存在的问题,提出一种在线调整比例、积分和微分参数的智能PID控制。通过在线参数调整,构成一个模拟人的控制行为的、变速的、非线性智能PID调节器。在模糊控制设计中,选用高斯函数作为隶属函数;对模糊因子采用自寻优方法进行优化,优化时的性能检验指标为误差绝对值积分时间性能指标ITAE最小;利用遗传算法对模糊控制器的模糊控制规则表进行优化,减少了无用的控制规则对控制系统性能不利影响。给出了一种模糊神经网络控制器的遗传算法设计方法,将智能控制的三个主要方法(模糊控制、神经网络和遗传算法)有机结合起来。控制系统采用模糊神经网络结构,其中模糊神经网络控制器采用五层简化的FNNC结构,模糊神经网络控制器的优化主要通过两种方法,一是基于遗传算法的离线优化,一是基于BP算法的在线优化。在离线优化阶段FNNC结构和参数的优化是通过对控制过程的模拟,利用遗传算法优化。因此无需事先提供控制规则,减少了因为被控对象辨识模型不准确带来的影响;在线优化阶段,利用实际的运行参数,进一步优化控制规则,提高了系统的自适应能力。利用遗传算法的全局寻优的特点,克服了基于BP算法的模糊神经网络经常陷入局部最优解的困境,发挥各自的长处,仿真结果证明了模糊神经网络控制器的优越性。通过对比分析可以看出基于神经网络的智能控制方法在制冷系统中的应用是有效可行的。神经网络等智能技术在制冷空调机组中的应用前景将是十分广阔的。更多还原

【Abstract】 A dynamics model of small air conditioning system was proposed in the paper. The compressor model was based on the steady lumped parameter method, and the changes of flow type and heat transfer coefficient of the refrigerant in evaporator and condenser have been taken into account for the first time in the heat exchanger model. The electrical expansion valve model was based on the lumped parameter method. Calculation results indicate that the model is effective in simulating the dynamics characteristics of the small system.Self-defining function curving fitting method in Origin program was first used in the thermal properties calculation of the refrigerants. A simple and convenient method based on the Cleland equations was proposed to calculate thermal properties of refrigerants R410A and R407C, which are known as two main substitutes of R22. The calculating results were compared with the data from the literature and the errors were acceptable. Compared with other thermal properties calculation model, the method has greater computation speed. The calculation results with minor error indicated that this method is an effective method for studying the thermal propertied of the refrigerant substitutes.A new mathematical model to calculate the flow characteristic of the adiabatic capillary tube was proposed in the paper, which was aiming at solving the problems existing in the former models. The Stocker’s basic model was modified with the consideration of various effects due to sub-cooling, area concentration, rolling diameter, and different equations for viscosity and friction factors. The new model not only can be used for R22, but also applied to the substitutes of R22, for example R410A and R407C. With those equations, the simulation results of the modified model are compared to those in literature, and the errors are within 10%. The flowing characteristic was also analyzed in the paper. In the calculating process, the iterated algorithm method was adopted to increasing the accuracy and velocity of the calculation.Simulation on the performance of adiabatic capillary tubes of the refrigerant R22, R407C and R410A was studied in the paper. According to the requirements of the simulation, a simple and convenient method based on the Cleland equations was proposed to calculate thermal properties of refrigerants R410A and R407C, which are known as two main substitutes of R22. The calculating results were compared with the data from the literature and the errors were acceptable.The PID, fuzzy and neural network control on evaporator superheat were studied and compared in the paper. Intelligent PID controller was proposed aiming at the problems existing in the PID controller. It can adjust the parameters on line to acquire the better controller results. In the fuzzy controller designing, the Gaussian function was selected as the membership function. The self-optimum method was adapted to optimize the fuzzy factor, and the minimum ITAE performance figure was used in inspecting the optimization performance. The genetic algorithms were introduced to optimize the fuzzy control rule to reduce the adverse effect of the useless control rule.A new fuzzy neural network controller based on the genetic algorithms was proposed in the paper, which combined the fuzzy control, neural network control and genetic algorithms. The control system based on the basic FNNC structure, and can be optimized through two ways. One way was off line optimization on the basis of genetic algorithms, the other way was on line optimization on the basis of BP. The purpose of off line optimization was optimize the structure and parameter of the FNNC, and the purpose of the on line optimization was utilizing the actual working parameters to further refine the controller rule to suit the actual control system, and enhancing the self-adapting ability. Geneticalgorithm is a kind of global optimization techniques that can be avoid the Back Propagation algorithm’s shortcoming of getting into local solution. Simulation and application in the paper show that the scheme is effective and feasible. The intelligent method based on the neural network can be widely used in the refrigeration system through the comparison between the above methods. The foreground of the neural network intellectual technology application in the refrigeration system controlling is perfecting wide.更多还原