基于CFD数值模拟和AI算法的燃气轮机燃烧优化  

Combustion optimization of gas turbine based on CFD numerical simulation and AI algorithm

作　　者：顾颀 钟文琪[1] 石岩[1] 石永锋 封伟 Gu Qi;Zhong Wenqi;Shi Yan;Shi Yongfeng;Feng Wei(Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education,Southeast University,Nanjing 210096,China;Shanghai Huadian Minhang Energy Co.,Ltd.,China Huadian Co.,Ltd.,Shanghai 201108,China;Institute for Process Modeling and Optimization,Jiangsu Industrial Technology Research Institute,Suzhou 215123,China)

机构地区：[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

出　　处：《东南大学学报（自然科学版）》2020年第3期545-554,共10页Journal of Southeast University：Natural Science Edition

基　　金：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).

摘　　要：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.

关 键 词：多目标优化 LS-SVM-GA 燃气轮机燃烧 CFD数值模拟 数据预处理 

分 类 号：TK16[动力工程及工程热物理—热能工程]