| 本文已被：浏览 883次 下载 241次 |
码上扫一扫！ |
|
|
| 基于BP神经网络与遗传算法反演大气温湿廓线 |
|
张天虎1,2, 鲍艳松1,2, 钱芝颖1,2, 林利斌3, 刘旭林4, 李林4, 侯岳5, 雷红玉5, 李广文5, 马军5, 管琴6, 唐维尧1,2
|
|
1. 南京信息工程大学气象灾害预报预警与评估协同创新中心/中国气象局气溶胶与云降水重点开放实验室，江苏 南京 210044;2. 南京信息工程大学大气物理学院，江苏 南京 210044;3.深圳航空有限责任公司，广东 深圳 518101;4.北京市气象探测中心，北京 100089;5.格尔木市气象局，青海 格尔木 816099;6.青海省气象台，青海 西宁 811000
|
|
| 摘要: |
| 为提高地基微波辐射计大气探测精度，融合BP神经网络与遗传算法，研究0～10 km大气温湿度廓线。首先，结合数据特征，基于数值模拟技术，建立一套TP/WVP-3000型号地基微波辐射计的一级数据质量控制和订正模型。然后，为减小训练样本代表性误差对模型反演精度的影响，利用遗传算法优化训练样本数据，建立一套精度更高的神经网络大气温湿度反演模型。最后，利用构建的反演模型，开展大气温湿度反演试验，结合探空资料和微波辐射计二级产品，评价反演模型精度。研究结果表明：(1) 经过质量控制后的实测数据与模拟数据之间的相关性有显著提升；(2) 经过质量控制与订正后建立的神经网络模型对比原微波辐射计二级产品的反演精度有一定提升，温度提升6.77%，湿度提升20.11%；(3) 经过遗传算法优化后的训练样本所建立的神经网络反演模型对比原微波辐射计二级产品反演精度有进一步的提升，温度提升10.21%，湿度提升23.75%，反演结果与该地区同类型研究结果相比有着较大提升。 |
| 关键词: 微波辐射计 遗传算法 BP神经网络 大气廓线 质量控制 |
| DOI：10.16032/j.issn.1004-4965.2020.010 |
| 分类号: |
| 基金项目: |
|
| ATMOSPHERIC TEMPERATURE AND HUMIDITY PROFILE RETRIEVALS BASED ON BP NEURAL NETWORK AND GENETIC ALGORITHM |
|
ZHANG Tian-hu1,2, BAO Yan-song1,2, QIAN Zhi-ying1,2, LIN Li-bin3, LIU Xu-lin4, LI Lin4, HOU Yue5, LEI Hong-yu5, LI Guang-wen5, MA Jun5, GUAN Qin6, TANG Wei-yao1,2
|
|
1. Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Key Laboratory for
Aerosol-Cloud-Precipitation of China Meteorological Administration, Nanjing University of Information
Science & Technology, Nanjing 210044, China;2. School of Atmospheric physics, Nanjing University of Information Science and Technology, Nanjing 210044, China;3.Shenzhen Airlines, Shenzhen 518101, China;4.Beijing Municipal Meteorological Observation Center,
Beijing 100089, China;5.Geermu Meteorological Bureau of Qinghai Province, Geermu 816099, China;6.Qinghai Meteorological office, Xining 810001, China
|
| Abstract: |
| In order to improve the sounding accuracy of ground-based microwave radiometer, the atmospheric temperature and humidity profiles of 0-10km are studied by combining BP neural network and genetic algorithm. Firstly, based on numerical simulation technology, a set of level 1 data's quality control and correction model of TP/WVP-3000 ground-based microwave radiometer is established. Then, in order to reduce the influence of training sample errors on the accuracy of retrieval model, genetic algorithm is used to optimize training sample data, and a set of more accurate neural network to retrieve atmospheric temperature and humidity is established. Finally, the model is used to carry out atmospheric temperature and humidity retrieval experiments, and then experiment results are combined with sounding data and microwave radiometer's level 2 data to evaluate the accuracy of the retrieval model. Results show that: (1) the consistency between the measured data and the simulated data after quality control is significantly improved; (2) Compared with the level2 data of microwave radiometer, the retrieval accuracy of the neural network model after quality control and correction has a certain improvement, including a 6.77% improvement in temperature and a 20.11% improvement in humidity; (3) Compared with the level2 data of microwave radiometer, the retrieval accuracy of the neural network retrieval model based on the genetic algorithm optimized training sample has a further improvement, including a 10.21% improvement in temperature improved and a 23.75% improvement in humidity. Slightly superior results are achieved with our algorithm. |
| Key words: microwave radiometer genetic algorithm BP neural network atmospheric profile quality control |