| 摘要:基于有机胺吸收剂的化学吸收法能够有效捕集船舶烟气中的CO2，但吸收剂降解失效、挥发泄漏会造成严重的环境问题。虽然傅里叶红外光谱仪能够实时监测挥发的有机胺体积分数，但受限于高饱和蒸气压有机胺类蒸气难标定问题，对胺类体积分数的实际监测精度较低。对此，设计两种傅里叶红外光谱仪中高饱和蒸气压有机胺蒸气体积分数标定方法，分别为固定比例校准算法和基于循环神经网络的序列映射法。结果表明，相较于固定比例法，基于循环神经网络的非线性序列映射法能够实现更好的校准性能，平均绝对误差下降了81.57%，均方误差下降了70.05%，可有效反映船舶碳捕集与封存（carbon capture and storage，CCS）系统对大气环境的影响。 |
| 关键词: 船舶CCS系统 有机胺吸收剂 算法校准 傅里叶红外光谱仪 循环神经网络 |
|
| Volume Fraction Calibration Method of High Saturated Vapor Pressure Organic Amine in Fourier Transform Infrared Spectrometer |
| HUANG Hao,LI Ke,CHEN Qiuyan,HE Zhijun |
| Shanghai Marine Diesel Engine Research Institute， Shanghai 201108， China |
| Abstract:The chemical absorption method based on organic amine absorbent can effectively capture CO2 in ship gas，but degradation，failure，volatilization and leakage of absorbent can cause serious environmental problems.Fourier transform infrared spectrometer（FTIR）can monitor volatile organic amine volume fraction in real time，but monitoring precision is low because of the problem that high saturated vapor pressure organic amine volume fraction is hard to calibrate.Two volume fraction calibration methods of high saturation vapor pressure organic amine in FTIR are put forward，which are fixed proportion calibration algorithm and sequence mapping method based on recurrent neural network.The results show that compared with the fixed proportion calibration algorithm，the nonlinear sequence mapping method based on recurrent neural network can achieve better calibration performance，mean absolute error（MAE）decreases by 81.57%，and mean square error（MSE）decreases by 70.05%，which can effectively reflect the influence of ship carbon capture and storage（CCS）system on atmospheric environment. |
| Key words: ship CCS system organic amine absorbent algorithm calibration Fourier transform infrared spectrometer（FTIR） recurrent neural network |
