| 引用本文: | 胡阳,简睿妮,房方.基于FDD–HSM方法的复杂拓扑供热管道动态等值建模[J].控制理论与应用,2022,39(3):509~518.[点击复制] |
| HU Yang,JIAN Rui-ni,FANG Fang.Dynamic equivalent modelling of complex topological heating pipeline based on finite difference domain-hybrid semi-mechanism method[J].Control Theory and Technology,2022,39(3):509~518.[点击复制] |
|
| 基于FDD–HSM方法的复杂拓扑供热管道动态等值建模 |
| Dynamic equivalent modelling of complex topological heating pipeline based on finite difference domain-hybrid semi-mechanism method |
| 摘要点击 1263 全文点击 419 投稿时间：2021-09-22 修订日期：2022-03-29 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2022.10889 |
| 2022,39(3):509-518 |
| 中文关键词 供热管网 有限差分域 混合半机理 长短期记忆神经网络 信息–物理融合 |
| 英文关键词 heating pipeline finite difference domain hybrid semi-mechanism long short-term memory neural network cyber-physical fusion |
| 基金项目 国家重点研发计划项目(2021YFE0102400), 国家自然科学基金项目(51906064)资助. |
|
| 中文摘要 |
| 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(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. |
| 英文摘要 |
| More and more attention to district electric-heat combined supply system including cogeneration unit is paid.
For its optimal operation, dynamic modelling of heating network is an important basis. According to heat transport principle
of heating pipeline and its operation data, a finite difference domain-hybrid semi-mechanism (FDD–HSM) dynamic
modelling method was proposed in this paper. Firstly, simplify a section of heating pipeline with complex topology into a
section of single-diameter and straight pipeline with simple topology while yielding the low-order equivalent mechanism
model structure. Secondly, considering input-output delay orders of low-order mechanism model, finite difference regression
vector is defined and the concept of finite difference space is presented. Adopting high-dimensional clustering and
hyperplane estimation, compact convex partition of finite difference space is achieved and several finite difference working
domains are obtained. Then, the HSM modelling method is proposed, in each working domain, identifying mechanism
model parameters and adding a dynamic compensation term of modelling deviation via the long short-term neural network
(LSTM) to realize arbitrary accuracy approximation. Finally, utilize measured operation data of the heating pipeline in
a district heating network to validate effectiveness and accuracy of the proposed method. The obtained linear low-order
mechanism models in the multiple working domains could be widely applied to fast simulation, numerical optimization
and control design of district heating network. |
|
|
|
|
|