摘要
为减小发射场气温预报误差,将BP神经网络和遗传算法结合起来,以2018—2022年的欧洲细网格气温预报数据和发射场实况数据为基础,利用相关系数筛选预报因子,建立了发射场气温预报模型。结果表明:模型气温预报平均绝对误差为1.132℃,较业务使用的欧洲细网格预报平均绝对误差优化了7.8%;模型气温预报的误差标准差为0.907℃,模型能较好修正欧洲细网格的温度预报误差离散值,较欧洲细网格预报的误差更稳定,能大幅减小预报员的人工订正工作;在神舟15号任务保障中,该模型预报的窗口温度为-18.02℃,而实况为-17.9℃,在临界条件下温度精细化预报保障提供了一种可靠性较高的预报手段。
In order to reduce the temperature forecast error in the launch site,the GA and BP were combined.Based on the EC fine grid forecast data and the launch site observational data from 2018 to 2022,the prediction factors were screened using the correlation coefficient.In the end,the temperature forecast model of the launch site was established.The results showed that the mean absolute error of the model temperature forecast was 1.132℃which was 78%better than that of the EC fine grid;The error standard deviation of the model temperature forecast was 0.907℃which was more stable than that of the EC fine grid and the manual correction work was greatly reduced;During the support process of Shenzhou-15 manned space mission,the window temperature predicted by the model was-18.08℃,and the actual temperature was-17.9℃.It could provide higher reliable forecasting method for temperature precision forecasting,especially in critical conditions.
作者
张芳
王刚
张朝飞
潘泉
陈锋
谭文秋
ZHANG Fang;WANG Gang;ZHANG Chaofei;PAN Quan;CHEN Feng;TAN Wenqiu(Jiuquan Satellite Launch Center,Jiuquan 732750,China)
出处
《载人航天》
CSCD
北大核心
2024年第2期219-223,共5页
Manned Spaceflight
关键词
BP神经网络
遗传算法
航天发射保障
气温预报
BP neural network
genetic algorithm
space launch support
temperature forecast
