粒子谱反演算法优化及反演误差分析

发布时间：2019-03-13 10:04

【摘要】：大气中存在着各种来源、大小不同的气溶胶粒子,这些粒子对于生态环境及生活健康具有重要影响。利用多波长激光雷达可以实现大气气溶胶粒子谱及微物理参量信息的高精度、大范围探测,但是与此同时对粒子谱数据反演算法提出了较高的要求。为了获得气溶胶粒径分布,本文提出了粒子谱反演算法,该算法采用正则化理论,结合355nm、532nm、1064nm消光系数和后向散射系数实现粒子谱反演,并针对反演结果不稳定问题,提出了优化方案。该方案是将多组反演结果进行平均化处理,选取满足微物理参量相对误差最小的百分比间隔。利用优化后的反演算法对三种典型气溶胶粒子谱分布进行仿真计算,仿真结果表明:有效半径、体积浓度相对误差控制在10%以内,复折射率实部绝对误差在±0.04以内,虚部误差在±0.005之间,反演效果较好。在上述优化算法基础上,研究分析了各个波长的光学参数误差对反演结果的影响。首先分析了单通道引入误差的情况,当给每个通道光学参数分别引入同比例-20%～20%的误差时,微物理参量反演误差随之增大,355nm、532nm消光系数误差对反演结果影响显著,其后向散射系数次之,1064nm后向散射系数最低。所以反演微物理参量量时需要精确测量355nm、532nm的消光系数。其次在单通道的基础上,分析了多通道误差对于反演结果的影响,结果表明:同时对不同光学通道引入10%的误差时,有效半径相对误差在35%以内,体积浓度在30%以内变化。这些结论在某种程度上可以指导激光雷达系统设计,评估微物理参量相对误差的大致范围。最后利用优化后的算法对实验室已有的多波长激光雷达系统数据进行反演分析,实验结果表明:该算法对于不同天气状况下的粒子谱分布以及微物理参量有较好的反演效果。
[Abstract]:There are various sources and sizes of aerosol particles in the atmosphere, which have an important impact on ecological environment and living health. Multi-wavelength lidar can achieve high precision and wide range detection of atmospheric aerosol particle spectrum and micro-physical parameter information, but at the same time, a higher requirement for particle spectrum data inversion algorithm is put forward. In order to obtain the particle size distribution, a particle spectrum inversion algorithm is proposed in this paper. The algorithm adopts regularization theory, combines 355nm, 532nm, 1064nm extinction coefficient and backscattering coefficient to realize the particle spectrum inversion, and aims at the problem of instability of the inversion results. The optimization scheme is put forward. In this scheme, multiple sets of inversion results are averaged and the percentage interval satisfying the minimum relative error of micro-physical parameters is selected. Three typical aerosol particle spectrum distributions are simulated by using the optimized inversion algorithm. The simulation results show that the effective radius, the relative error of volume concentration is less than 10%, and the absolute error of the real part of complex refractive index is less than 卤0.04. The virtual part error is in the range of 卤0.005, and the inversion effect is better. On the basis of the above-mentioned optimization algorithm, the influence of the optical parameter errors of each wavelength on the inversion results is studied and analyzed. Firstly, the error of single channel is analyzed. When the same ratio of-20% / 20% error is added to the optical parameters of each channel, the inversion error of micro-physical parameters increases. 355nm, 532nm extinction coefficient error has a significant effect on the inversion results, and the error of extinction coefficient of 532nm has a significant effect on the inversion results. The backscattering coefficient is the second, and the 1064nm backscattering coefficient is the lowest. Therefore, the extinction coefficient of 355nm and 532nm should be measured accurately when inverting the parameters of microphysics. Secondly, on the basis of single channel, the influence of multi-channel error on the inversion results is analyzed. The results show that the relative error of effective radius is less than 35% when the error of 10% is introduced to different optical channels at the same time. The volume concentration varies within 30%. To some extent, these conclusions can guide the design of lidar system and evaluate the approximate range of relative error of micro-physical parameters. Finally, the optimized algorithm is used to inverse the data of the multi-wavelength lidar system in the laboratory. The experimental results show that the algorithm has a good inversion effect on the particle spectrum distribution and micro-physical parameters under different weather conditions.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X87;X513

【参考文献】