基于FDD-HSM方法的复杂拓扑供热管道动态等值建模

The regional combined heat and power supply system with cogeneration units has been increasingly emphasized, and the dynamic modeling of the heating pipe network is an important foundation for its optimal operation. Based on the principle of thermal transport in pipelines and their operating data, a finite difference domain - hybrid semi - mechanism (FDD - HSM) dynamic modeling method is proposed. Firstly, the complex topological heating pipeline is simplified into a simple topological shape with a single pipe diameter and straight pipe sections, and a low - order equivalent mechanism model structure is given. Secondly, considering the input - output delay order of the low - order mechanism model, the finite difference regression vector is defined and the concept of the finite difference space is proposed. High - dimensional clustering and hyperplane estimation are used to achieve its compact convex partition and obtain several finite difference working domains. Then, a hybrid semi - mechanism modeling method is proposed. The mechanism model parameters are identified in each working domain, and a long short - term memory neural network deviation dynamic compensation term is added to achieve approximation with arbitrary accuracy. Finally, based on the measured operating data of a regional heating network pipeline, the effectiveness and accuracy of the proposed method are verified. The obtained multi - working domain - linear low - order mechanism model can be widely used in the rapid simulation, numerical optimization and control design of the heating network.

含热电联产机组的区域电热联合供给系统日益受到重视,供热管网动态建模是其优化运行的重要基础.基于管道热力输运原理及其运行数据,提出一种有限差分域-混合半机理(finite difference domain-hybrid semi-mechanism,FDD-HSM)动态建模方法.首先,简化复杂拓扑供热管道为单一管径、直管段的简单拓扑形状并给出低阶等值机理模型结构.其次,考虑低阶机理模型输入-输出延迟阶次,定义有限差分回归向量并提出有限差分空间概念,采用高维聚类和超平面估计实现其紧致凸划分并获得若干有限差分工作域.然后,提出混合半机理建模方法，在各工作域辨识机理模型参数,并增设长短期记忆神经网络偏差动态补偿项,实现任意精度逼近.最后,基于某区域热网管道实测运行数据验证了所提方法的有效性及准确性,所得多工作域-线性低阶机理模型可广泛应用于热网快速仿真、数值优化及控制设计。