基于数据和神经网络的燃气轮机NOx排放预测与优化

NO_x Emission Prediction and Optimization for Gas Turbines Based on Data and Neural Network

基于大量运行数据,采用神经网络对燃气轮机NOx排放及燃烧稳定性参数进行建模,针对影响燃烧性能的参数进行了敏感性分析,确定其大小及正负相关性,并据此对实际工况点参数进行调整,将模型预测结果与物理规律定性分析结果进行对比。结果表明:模型的预测结果与基于机理的定性分析结果基本一致,即NOx质量浓度随燃料质量流量的增加而升高,随压气机进口导叶(IGV)开度增大而降低;燃烧稳定性参数随值班火焰燃料质量流量增加而减小,随IGV开度增大而增大。

Based on a large sum of operation data from power plants, the artificial neural network was used to model the parameters concerning the NO_x emission and combustion stability of a gas turbine, while a sensitivity analysis was conducted on the parameters influencing the combustion performance to find out the correlation between the inputs and outputs. Subsequently, the parameters at actual operation point were adjusted, and the results of model prediction were compared with that of qualitative analysis based on physical mechanism. It has been found that the results of model calculation are basically consistent with the qualitative analysis, that is, the NO_x emission grows with the increase of fuel flow rate and decreases with the increase of IGV opening;the combustion stability index(CSI) reduces with the increase of fuel flow rate and rises with the increase of IGV opening.