基于串联形式智能融合算法的燃气轮机建模与仿真  

Modeling and Simulation of a Gas Turbine Based on Series Form Intelligent Fusion Algorithm

作　　者：李薇 何兴 许野 荀金柱 王旭 包哲 LI Wei;HE Xing;XU Ye;XUN Jinzhu;WANG Xu;BAO Zhe(College of Environmental Science and Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学环境科学与工程学院,北京102206

出　　处：《动力工程学报》2022年第6期507-513,558,共8页Journal of Chinese Society of Power Engineering

基　　金：国家自然科学基金资助项目(62073134)。

摘　　要：为提高燃气轮机整体动态仿真模型的准确性与实用性,基于预先构建的机理、BP神经网络和逐步聚类分析法3种单一仿真模型,分别将BP神经网络和逐步聚类分析法模型与机理模型进行组合,创新性地构建了2套串联形式的智能融合模型,有效地克服了机理模型部分过程过于简化和参数获取困难、智能预测模型高度依赖数据质量和缺乏外延性等缺陷,并将上述模型应用于辽宁某钢厂燃气轮机的仿真研究。结果表明:单一仿真模型中智能预测模型的预测效果较好,且逐步聚类分析法模型优于BP神经网络模型;智能融合模型的精度比单一仿真模型更高,其中逐步聚类分析法+机理智能融合模型的仿真效果更优。In order to improve the accuracy and practicability of entire dynamic simulation model of a gas turbine,based on the preestablished three types of single simulation models(i.e.mechanism,BP neural network and stepwise cluster analysis models),two types of intelligent fusion simulation models were innovatively formulated through combining mechanism model with BP neural network model and stepwise cluster analysis model,respectively.Fusion simulation models would effectively overcome the oversimplification and the difficulties in parameters identification related to mechanism model,as well as the high dependence on data quality and lack of extensive property associated with intelligent prediction models.The above models were applied to the simulation study of a gas turbine in a steel plant in Liaoning Province.Results show that the prediction effects of intelligent simulation model are better among all single models,in which stepwise cluster analysis model is superior to BP neural network model.Intelligent fusion model has a higher precision than the single models,and the model combining the mechanism with stepwise cluster analysis owns the better performance.

关 键 词：燃气轮机 BP神经网络 逐步聚类分析法 智能融合 仿真模型 

分 类 号：TK14[动力工程及工程热物理—热能工程]