Gu Qi,Zhong Wenqi,Shi Yan,et al.Combustion optimization of gas turbine based on CFD numerical simulation and AI algorithm[J].Journal of Southeast University (Natural Science Edition),2020,50(3):545-554.[doi:10.3969/j.issn.1001-0505.2020.03.018]





Combustion optimization of gas turbine based on CFD numerical simulation and AI algorithm
1 东南大学能源热转换及其过程测控教育部重点实验室, 南京 210096; 2 中国华电集团有限公司上海华电闵行能源有限公司, 上海 201108; 3 江苏省产业技术研究院工业过程模拟与优化研究所, 苏州 215123
Gu Qi 1 Zhong Wenqi 1 Shi Yan 1 Shi Yongfeng2 Feng Wei3
1 Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education, Southeast University, Nanjing 210096, China
2 Shanghai Huadian Minhang Energy Co., Ltd., China Huadian Co., Ltd., Shanghai 201108, China
3 Institute for Process Modeling and Optimization, Jiangsu Industrial Technology Research Institute, Suzhou 215123, China
多目标优化 LS-SVM-GA 燃气轮机燃烧 CFD数值模拟 数据预处理
multi-objective optimization least squares support vector machine(LS-SVM-GA) gas turbine combustion computational fluid dynamics(CFD)numerical simulation data preprocessing
为提高燃气轮机的燃烧性能和稳定性,减少污染物排放,提出了一种基于CFD数值模拟和AI算法的燃气轮机燃烧优化方法.利用大涡模拟(LES)和部分预混火焰面生成流形(FGM)燃烧模型来保证CFD数值模拟的准确性,将其计算结果与电厂运行数据相结合,建立更全面的训练数据库,然后利用拉依达法则和核主成分分析法(KPCA)进行数据预处理.在此基础上,建立了3个最小二乘支持向量机(LS-SVM)模型,分别用来预测NOx排放量、燃烧效率和压力脉动最大幅值,预测的平均相对误差分别为0.562%、0.336%和0.469%.结果表明,CFD模拟数据的加入使该预测模型适用范围更广,稳定性和准确性更高.基于最小二乘支持向量机模型的预测结果,采用遗传算法(GA)对燃料比例分配、空燃体积比等参数进行优化,最终得到NOx排放量、燃烧效率和压力脉动最大幅值的平均优化量分别为3.692×10-6、0.568%和0.926 kPa,基本满足优化要求.
A combustion optimization method for the gas turbine based on computational fluid dynamics(CFD)numerical simulation and AI algorithm was developed to improve the combustion performance and the stability, and to reduce the pollutant. The large eddy simulation(LES)and flamelet generated manifold(FGM)model were used to guarantee the accuracy of CFD simulation, whose results were combined with the power plant operation data to build a more detailed training database. Then, the PauTa criterion and the kernel principal component analysis(KPCA)were used for data preprocessing. Based on it, three least square support vector machine(LS-SVM)models were established to predict NOx emission, combustion efficiency and maximum pressure pulsation amplitude, of which the average relative errors were obtained as 0.562%, 0.336% and 0.469%, respectively. The results indicate that the application of CFD simulation data can generalize the prediction models,and improve the stability and the accuracy of the models. Based on the results of the LS-SVM models, a genetic algorithm(GA)is used to optimize the operating conditions, including the fuel proportional distribution and the air-fuel ratio. Finally, the average optimization improvement of NOx emission, combustion efficiency and maximum pressure pulsation amplitude are 3.692×10-6, 0.568% and 0.926 kPa, respectively basically meeting the optimization requirements.


[1] 刘志坦, 王文飞. 我国燃气发电发展现状及趋势[J]. 国际石油经济, 2018, 26(12): 43-50. DOI:10.3969/j.issn.1004-7298.2018.12.006.
Liu Z T, Wang W F. Current status and development trend of gas power generation in China[J]. International Petroleum Economics, 2018, 26(12): 43-50. DOI:10.3969/j.issn.1004-7298.2018.12.006. (in Chinese)
[2] 陈俊龙, 王玉璋, 刘闳钊, 等. 微型燃气轮机低热值燃料燃烧室喷嘴流动实验研究[J]. 热能动力工程, 2017, 32(12): 33-40. DOI:10.16146/j.cnki.rndlgc.2017.12.006.
Chen J L, Wang Y Z, Liu H Z, et al.Experimental research of nozzle flow of low calorific value fuel combustion chamber for micro gas turbine[J]. Journal of Engineering for Thermal Energy and Power, 2017, 32(12): 33-40, 137. DOI:10.16146/j.cnki.rndlgc.2017.12.006. (in Chinese)
[3] Mark C P, Selwyn A. Design and analysis of annular combustion chamber of a low bypass turbofan engine in a jet trainer aircraft[J]. Propulsion and Power Research, 2016, 5(2): 97-107. DOI:10.1016/j.jppr.2016.04.001.
[4] 齐心. 锥形燃气轮机燃烧器性能研究[D]. 北京: 华北电力大学(北京), 2017.
  Qi X. Research on taper burner performance of gas turbine[D]. Beijing: North China Electric Power University, 2017.(in Chinese)
[5] 刘洋. 反应机理对某重型燃气轮机燃烧影响的数值研究[D]. 沈阳: 沈阳工程学院, 2018.
  Liu Y. Numerical study on the influence of different chemical reaction mechanisms on combustion in a heavy duty gas turbine[D]. Shenyang: Shenyang Insitute of Engineering, 2018.(in Chinese)
[6] Shui A S, Fang H M, Zong F X, et al. The error prediction of inventory reconciliation during storage and transportation process based on PLS and MFOA-LSSVM[J]. Journal of Petroleum Exploration and Production Technology, 2017, 7(3): 895-904. DOI:10.1007/s13202-016-0292-0.
[7] Sharifi M A, Souri A H. A hybrid LS-HE and LS-SVM model to predict time series of precipitable water vapor derived from GPS measurements[J]. Arabian Journal of Geosciences, 2015, 8(9): 7257-7272. DOI:10.1007/s12517-014-1716-0.
[8] Y�131;lmaz E. An expert system based on fisher score and LS-SVM for cardiac arrhythmia diagnosis[J]. Computational and Mathematical Methods in Medicine, 2013, 2013: 1-6. DOI:10.1155/2013/849674.
[9] Wu X J, Huang Q, Zhu X J. Thermal modeling of a solid oxide fuel cell and micro gas turbine hybrid power system based on modified LS-SVM[J]. International Journal of Hydrogen Energy, 2011, 36(1): 885-892. DOI:10.1016/j.ijhydene.2010.08.022.
[10] 陈永刚. 基于最小二乘支持向量机的PG9171E型燃气轮机气耗特性建模研究[J]. 华电技术, 2010, 32(5): 14-17, 80.
  Chen Y G. The characteristic simulated study of gas consumption for PG9171E gas turbine based on LS-SVM[J]. Huadian Technology, 2010, 32(5): 14-17, 80.(in Chinese)
[11] Fei C W, Bai G C, Tang W Z, et al. Optimum control for nonlinear dynamic radial deformation of turbine casing with time-varying LSSVM[J]. Advances in Materials Science and Engineering, 2015, 2015: 1-9. DOI:10.1155/2015/680406. qwuuuuuu
[12] Valdés M, Durán M D, Rovira A. Thermoeconomic optimization of combined cycle gas turbine power plants using genetic algorithms[J]. Applied Thermal Engineering, 2003, 23(17): 2169-2182. DOI:10.1016/s1359-4311(03)00203-5.
[13] Syed K J, Roden K, Martin P. A novel approach to predicting NOx emissions from dry low emissions gas turbines[J]. Journal of Engineering for Gas Turbines and Power, 2007, 129(3): 672-679. DOI:10.1115/1.2718564.
[14] 蒋丹,李松晶,包钢. 采用遗传算法对压力脉动过程中气泡模型参数的辨识[J]. 物理学报,2008,57(8):5072-5080.
  Jiang D, Li S J,Bao G. Parameter identification of gas bubble model in pressure pulsations using genetic algorithms[J]. Acta Physica Sinica, 2008,57(8):5072-5080.(in Chinese)
[15] 崔玉峰,徐纲,聂超群,等. 数值模拟在合成气燃气轮机燃烧室设计中的应用[J]. 中国电机工程学报,2006,26(16):109-116.
  Cui Y F, Xu G, Nie C Q, et al. Application of numerical simulation in the design of gas turbine combustor for burning syngas[J]. Proceedings of the CSEE, 2006,26(16):109-116.(in Chinese)
[16] Pillier L, Idir M, Molet J, et al. Experimental study and modelling of NOx formation in high pressure counter-flow premixed CH4/air flames[J]. Fuel, 2015, 150: 394-407. DOI:10.1016/j.fuel.2015.01.099.
[17] Pang Y W, Wang L, Yuan Y. Generalized KPCA by adaptive rules in feature space[J]. International Journal of Computer Mathematics, 2010, 87(5): 956-968. DOI:10.1080/00207160802044118.
[18] Wu H, Ihme M. APS-70th annual meeting of the APS division of fluid dynamics-event-application of pareto-efficient combustion modeling framework to large eddy simulations of turbulent reacting flows[J].Bulletin of the American Physical Society, 2017,62(14): 0003-0503.
[19] Merk M, Jaensch S, Silva C, et al. Simultaneous identification of transfer functions and combustion noise of a turbulent flame[J]. Journal of Sound and Vibration, 2018, 422: 432-452. DOI:10.1016/j.jsv.2018.02.040.
[20] Pitsch H, Du champ de Lageneste L. Large-eddy simulation of premixed turbulent combustion using a level-set approach[J]. Proceedings of the Combustion Institute, 2002, 29(2): 2001-2008. DOI:10.1016/S1540-7489(02)80244-9.
[21] van Oijen J A, Donini A, Bastiaans R J M, et al. State-of-the-art in premixed combustion modeling using flamelet generated manifolds[J]. Progress in Energy and Combustion Science, 2016, 57: 30-74. DOI:10.1016/j.pecs.2016.07.001.
[22] Goebel F, Mundt C. Implementation of the P1 radiation model in the CFD solver NSMB and investigation of radiative heat transfer in the SSME main combustion chamber[C]//17th AIAA International Space Planes and Hypersonic Systems and Technologies Conference. San Francisco, CA, USA,2011:750-760 DOI:10.2514/6.2011-2257.


收稿日期: 2019-11-02.
作者简介: 顾颀(1994—),女,硕士生;钟文琪(联系人),男,博士,教授,博士生导师,wqzhong@seu.edu.cn.
基金项目: 国家自然科学基金重大资助项目(51390492)、装备预研教育部联合基金资助项目(6141A02033524).
引用本文: 顾颀,钟文琪,石岩,等.基于CFD数值模拟和AI算法的燃气轮机燃烧优化[J].东南大学学报(自然科学版),2020,50(3):545-554. DOI:10.3969/j.issn.1001-0505.2020.03.018.
更新日期/Last Update: 2020-05-20