基于灰箱理论的燃气轮机建模方法及输出预测  

Modeling Method and Output Prediction of Gas Turbine Based on Grey Box Theory

作　　者：张辉 阮圣奇 李亚飞 夏永放 李敬豪 王斌 王严 ZHANG Hui;RUAN Sheng-qi;LI Ya-fei;XIA Yong-fang;LI Jing-hao;WANG Bin;WANG Yan(China Datang Corporation Science and Technology General Research Institute Ltd.,Beijing 100040,China;China Datang Corporation Science and Technology General Research Institute Ltd.,East China Electric Power Test and Research Institute,Hefei 231283,China;School of Environment and Energy Engineering,Anhui Jianzhu University,Hefei 230601,China)

机构地区：[1]中国大唐集团科学技术研究总院有限公司,北京100040 [2]中国大唐集团科学技术研究总院有限公司华东电力试验研究院,合肥231283 [3]安徽建筑大学环境与能源工程学院,合肥230601

出　　处：《汽轮机技术》2025年第2期150-155,共6页Turbine Technology

摘　　要：聚焦于燃气轮机运行工况不稳定及仿真建模难度高的问题,鉴于传统建模方法单一且模拟效果难以满足实时工况需求,基于灰箱理论的模型建立方法被提出,旨在提高模型的预测精度。首先,采用机理建模与数据驱动建模相结合的方法,对压气机模块进行机理建模,对燃烧室和透平模块进行数据驱动建模。随后,将压气机模块运算所得数据结果作为GA Elman神经网络的输入参数。最后,选取均方根误差(Root Mean Square Error,RMSE)和相关系数(Pearson Correlation Coefficient,PCC)作为评判模型精确度的标准。机理模型与GA Elman神经网络模型相融合的灰箱模型在预测燃气轮机实际工况的输出功率时,其测试集均方根误差(RMSE)为2.892,测试集相关系数(PCC)为0.98366。表现为均方根误差较小,相关系数接近1。该模型可精准预测燃气轮机输出功率,有效及时预防燃气轮机运行过程中出现的故障问题。This study focuses on the problem of unstable gas turbine operating conditions and the difficulty of simulation modelling.Since the traditional modelling method is single and the simulation effect is difficult to meet the requirement of real-time operating conditions,a model building method based on grey box theory is proposed to improve the prediction accuracy of the model.Firstly,a combination of mechanistic modelling and data-driven modelling is adopted to carry out mechanistic modelling of the compressor module and data-driven modelling of the combustor and turbine module.Then,the data results obtained from the compressor module operations were used as input parameters for the GA Elman neural network.Finally,Root Mean Square Error(RMSE)and Pearson Correlation Coefficient(PCC)were chosen as the criteria to evaluate the accuracy of the model.The grey box model fused with the GA Elman neural network model can accurately predict the output power of the gas turbine under the actual working conditions with the test set RMSE of 2.892 and the test set correlation coefficient(PCC)of 0.98366,which shows that the RMSE is small and the correlation coefficient is close to 1.The model can accurately predict the output power of the gas turbine and prevent the failure of the gas turbine during the operation process.

关 键 词：燃气轮机 灰箱理论 神经网络 预测 

分 类 号：TK479.1[动力工程及工程热物理—动力机械及工程]