基于神经网络和过程机理的锅炉过热系统动态仿真

神经网络模型由于缺乏物理基础，其外推能力不够理想，甚至有时其预报结果与实际系统的基本规律相矛盾；机理模型通常过于复杂，且无法描述未知扰动的影响，一般难以在线应用。该文基于网络和过程机理特性，构建了电厂锅炉过热蒸汽系统仿真模型。在仿真模型设计上，充分地考虑了过热蒸汽系统输入与输出间的物理基础，网络的训练除了具有通常的数值映射关系学习功能之外，还具有过程机理学习功能，从而保证了文中的网络模型常规的神经网络模型具有十分良好的联想能力、对系统过渡过程的预报能力和外推效果。初步研究表明，文中提出的基于神经网络和过程机理的仿真建模方法，可以有效地克服复杂系统仿真精度和仿真速度间的矛盾，拓宽仿系统的工程应用领域。

DYNAMIC SIMULATION OF THE SUPERHEAT SYSTEM OF BOILER BASED ON NEURAL NETWORK AND PROCESS MECHANISM

Neural network model is difficult to be applied on line, because the neural network model lacks physical basis and its extrapolation capability is not ideal, even its forecast result is in contradiction with the basic law of a real system, mechanism model is too complex and unable to describe influence of an unknown disturbance. In this paper, a simulation model of superheat system of power plant boiler is presented, which based on neural network and process mechanism characteristics. On design of the simulation model, the physical basis between the input and the output of the superheated steam system of boiler was well considered. The network model has the learning ability of process mechanism, besides routine learning function of numeric mapping relations. Thus, good association ability, the prediction ability, and the extrapolation effect on system transition process of the network model were well assured, compared with routine neural network models. The preliminary research shows,the simulation method based on neural network and process mechanism could get over the contradiction between the accuracy and the speed of the simulation of the complex system effectively, and widen the engineering applied field of the simulation system.