结合人工神经网络的冷凝器稳态分布参数模型

为了克服由于实际装置的复杂性及生产工艺的差异对冷凝器稳态仿真精度的影响 ,提高冷凝器仿真模型的通用性和准确性 ,提出了冷凝器基本模型结合人工神经网络的仿真思路 .以相区划分和制冷剂出口焓值迭代为基础 ,提出了一种稳定的逆流型冷凝器仿真分布参数模型和算法 ,建立了冷凝器仿真的基本模型 .其计算结果与实验数据的变化趋势一致 ,能够在定性上反映实际物理过程的基本特性 .通过对部分实验数据的学习 ,进一步建立了与基本模型相结合的人工神经网络 .利用其非线性映射能力进行模型修正 ,显著提高了冷凝器的仿真精度 ,从而为同时提高冷凝器仿真的通用性和准确性提供了一种有效的工程应用方法

Steady-State Distributed-Parameter Model Integrated with Artificial Neural Network for Condensers

In order to overcome the effect of complexity of real plants and enhance the generalization and precision of the mathematical model, a simulation method integrating the basic condenser model with artificial neural network was presented. The basic counterflow condenser model is constructed by a steady state distributed parameter model and a corresponding stable algorithm for the counterflow condenser based on the phase change and the iterative calculation of refrigerant enthalpy at exit. The trend of simulation results is in accord with that of the experimental data and can reflect the basic qualitative characteristic of the real physical process. Through learning the part of experimental data by BP algorithm, a multi layer artificial neural network was established and integrated with the counterflow condenser model. The nonlinear mapping ability of the artificial neural network was utilized to modify the counterflow condenser model and improve its predicting results. It shows that this method is an effective way to improve the generalization and precision of the general condenser model.