发电过程孪生模型在线演化方法研究  

Research on Online Evolution Method of Twin Models in Power Generation Process

作　　者：张悦[1,2] 练有焜 展元 李诺 ZHANG Yue;LIAN Youkun;ZHAN Yuan;LI Nuo(Department of Automation,North China Electric Power University,Baoding 071003,China;Technology Innovation Center of Simulation&Optimized Control for Power Generation of Hebei Province,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003 [2]河北省发电过程仿真与优化控制技术创新中心(华北电力大学),河北保定071003

出　　处：《电力科学与工程》2023年第7期70-78,共9页Electric Power Science and Engineering

基　　金：河北省科技计划项目(22567643H)。

摘　　要：随着机器学习和深度学习的成熟,数据驱动模型得到了快速发展,过程模型的精度也得到了显著提高。目前,模型性能评价已经不再局限于模型精度,模型的泛化、演化等能力也成为重要的衡量指标,尤其体现在数字孪生的建模研究中。数字孪生建模不仅要求模型能够逼真地再现物理实体,而且还要求模型随着物理实体的变化而演化。这就要求数字孪生模型能在保证精度的同时,还具备在线演化的特点。基于串联混合模型建立离线局部模型库,构建了马尔可夫模型,设计了马尔可夫模型与局部模型结合的两层演化模型。利用马尔可夫模型对运行状态进行预测,应用RBF神经网络实现串联混合模型中数据驱动子模型过渡参数的演变,从而达到孪生模型整体演化的目的。最后,建立了空预器换热过程的孪生模型,并进行仿真验证,证明了所提方法的有效性。With the maturity of machine learning and deep learning,data driven models have developed rapidly,and the accuracy of process models has also been significantly improved.Currently,the advantages and disadvantages of models are no longer limited to model accuracy,the generalization and evolution capabilities of models are also important indicators,especially reflected in the modeling research of digital twin.Digital twin model requires not only realistic representation of physical entities,but also needs to evolve as physical entities change.This requires the digital twin model to have the characteristics of online evolution while ensuring accuracy.Based on the series hybrid model,an offline local model library was established,and a Markov model was constructed.A two-level evolution model combining the Markov model and the local model was designed.Markov model was used to predict the operating state,and RBF neural network was applied to achieve the evolution of data driven sub model transition parameters in the series hybrid model,thereby achieving the overall evolution of the twin model.Finally,a twin model of the heat transfer process of the air preheater is established and verified by simulation,indicating the effectiveness of the proposed method.

关 键 词：数字孪生 模型演化 马尔可夫模型 RBF神经网络 混合模型 

分 类 号：TK32[动力工程及工程热物理—热能工程] TK223.3