Post-processing correction is an effective way to improve the model forecasting result. Especially, the machine learning methods have played increasingly important roles in recent years. Taking the meteorological obse...Post-processing correction is an effective way to improve the model forecasting result. Especially, the machine learning methods have played increasingly important roles in recent years. Taking the meteorological observational data in a period of two years as the reference, the maximum and minimum temperature predictions of Shenyang station from the European Center for Medium-Range Weather Forecasts (ECMWF) and national intelligent grid forecasts are objectively corrected by using wavelet analysis, sliding training and other technologies. The evaluation results show that the sliding training time window of the maximum temperature is smaller than that of the minimum temperature, and their difference is the largest in August, with a difference of 2.6 days. The objective correction product of maximum temperature shows a good performance in spring, while that of minimum temperature performs well throughout the whole year, with an accuracy improvement of 97% to 186%. The correction effect in the central plains is better than in the regions with complex terrain. As for the national intelligent grid forecasts, the objective correction products have shown positive skills in predicting the maximum temperatures in spring (the skill-score reaches 0.59) and in predicting the minimum temperature at most times of the year (the skill-score reaches 0.68).展开更多
A novel identification method of aerodynamicmodels using a physics neural network,named the attitude dynamics network,which incorporates the attitude dynamics of an aircraft without any prior aerodynamic knowledge,is ...A novel identification method of aerodynamicmodels using a physics neural network,named the attitude dynamics network,which incorporates the attitude dynamics of an aircraft without any prior aerodynamic knowledge,is proposed.Then a learning controller,which combines feedback linearization with sliding mode control,is developed by introducing the learned aerodynamicmodels.The merit of the identification method is that the aerodynamicmodels can be learned end-to-end by the physics network directly from the flight data.Consequently,the paper uses an offline scheme and an online scheme to combine the identification process and the control process.In the offline scheme,learning the aerodynamic models and controlling the aircraft compose a cascade system,whereas the online scheme,similar to Learn-to-Fly,is a parallel system.Specifically,in the offline scheme,the physics neural network is trained by sufficient offline flight data,and then the trained network is substituted into the controller.The online scheme refers to the controller making the aircraft fly to generate flight data and sending these data to the deep network at the time of training,while the deep network provides the trained aerodynamic models to the controller at other times.Simulation results show that both under nominal and disturbance aerodynamic conditions,the network trained offline with a large amount of nominal data approximate the aerodynamicmodels well.Thus,the performance of the controller reaches a good level;for the online scheme,the predictive capability of the network increases and the performance of the controller improves with more training data.展开更多
基于准对称混合滑动训练期方法,对近两年来中国气象局陆面同化分析系统(CMA Land Data Assimilation System,CLDAS)输出的日最高气温、日最低气温网格实况分析产品进行订正,以期提高该产品在重庆地区的适用性。结果表明:订正前,2021年CL...基于准对称混合滑动训练期方法,对近两年来中国气象局陆面同化分析系统(CMA Land Data Assimilation System,CLDAS)输出的日最高气温、日最低气温网格实况分析产品进行订正,以期提高该产品在重庆地区的适用性。结果表明:订正前,2021年CLDAS日最高气温产品的平均误差为0.63℃,平均绝对误差为1.14℃,订正后平均误差减小至-0.03℃,平均绝对误差减小至0.64℃,误差小于或等于1℃的准确率由约64%提高到约90%,明显改善了该产品在重庆西部和东南部地区的适用性;订正前,2021年CLDAS日最低气温的平均误差为-0.22℃,平均绝对误差为0.75℃,订正后平均误差减小至-0.03℃,平均绝对误差减小至0.55℃,误差小于或等于1℃的准确率由约91%提高到约93%,改善了该产品在重庆中部地区的适用性。展开更多
为提高智能网格的订正能力及预报水平,基于中央台客观指导产品的甘肃省切片数据和中国气象局陆面数据同化系统(Chinese Land Data Assimilation System Version 2.0,CLDAS-V2.0)日网格实况产品,采用卡尔曼滤波和滑动训练订正两种方法,...为提高智能网格的订正能力及预报水平,基于中央台客观指导产品的甘肃省切片数据和中国气象局陆面数据同化系统(Chinese Land Data Assimilation System Version 2.0,CLDAS-V2.0)日网格实况产品,采用卡尔曼滤波和滑动训练订正两种方法,对河西走廊东部地区(101.0°E—104.5°E,36.0°N—40.0°N)0.05°×0.05°格点最高、最低气温进行订正、检验和评估。结果表明:(1)季节对比,卡尔曼滤波和滑动训练订正产品对四季最高、最低气温的平均绝对误差均小于中央台客观指导产品,均小于2.00℃;卡尔曼滤波和滑动训练订正产品对四季最高、最低气温的预报准确率均大于70%,其中最高气温偏高6%~13%,最低气温偏高8%~24%。(2)空间对比,卡尔曼滤波和滑动训练订正产品对最高、最低气温的平均绝对误差绝大部分地区在1.00~2.00℃,个别地区大于2.00℃;卡尔曼滤波和滑动训练订正产品对最高(最低)气温的预报准确率大部分地区大于70%(60%~70%),个别地区大于80%(70%)。(3)总体上,卡尔曼滤波和滑动训练订正产品对最高、最低气温订正技巧基本为正技巧,个别季节和部分地区订正技巧大于0.300。说明两种订正方法具有较好的订正预报能力,可为今后的温度预报业务提供一定的技术支持。展开更多
文摘Post-processing correction is an effective way to improve the model forecasting result. Especially, the machine learning methods have played increasingly important roles in recent years. Taking the meteorological observational data in a period of two years as the reference, the maximum and minimum temperature predictions of Shenyang station from the European Center for Medium-Range Weather Forecasts (ECMWF) and national intelligent grid forecasts are objectively corrected by using wavelet analysis, sliding training and other technologies. The evaluation results show that the sliding training time window of the maximum temperature is smaller than that of the minimum temperature, and their difference is the largest in August, with a difference of 2.6 days. The objective correction product of maximum temperature shows a good performance in spring, while that of minimum temperature performs well throughout the whole year, with an accuracy improvement of 97% to 186%. The correction effect in the central plains is better than in the regions with complex terrain. As for the national intelligent grid forecasts, the objective correction products have shown positive skills in predicting the maximum temperatures in spring (the skill-score reaches 0.59) and in predicting the minimum temperature at most times of the year (the skill-score reaches 0.68).
文摘A novel identification method of aerodynamicmodels using a physics neural network,named the attitude dynamics network,which incorporates the attitude dynamics of an aircraft without any prior aerodynamic knowledge,is proposed.Then a learning controller,which combines feedback linearization with sliding mode control,is developed by introducing the learned aerodynamicmodels.The merit of the identification method is that the aerodynamicmodels can be learned end-to-end by the physics network directly from the flight data.Consequently,the paper uses an offline scheme and an online scheme to combine the identification process and the control process.In the offline scheme,learning the aerodynamic models and controlling the aircraft compose a cascade system,whereas the online scheme,similar to Learn-to-Fly,is a parallel system.Specifically,in the offline scheme,the physics neural network is trained by sufficient offline flight data,and then the trained network is substituted into the controller.The online scheme refers to the controller making the aircraft fly to generate flight data and sending these data to the deep network at the time of training,while the deep network provides the trained aerodynamic models to the controller at other times.Simulation results show that both under nominal and disturbance aerodynamic conditions,the network trained offline with a large amount of nominal data approximate the aerodynamicmodels well.Thus,the performance of the controller reaches a good level;for the online scheme,the predictive capability of the network increases and the performance of the controller improves with more training data.
文摘基于准对称混合滑动训练期方法,对近两年来中国气象局陆面同化分析系统(CMA Land Data Assimilation System,CLDAS)输出的日最高气温、日最低气温网格实况分析产品进行订正,以期提高该产品在重庆地区的适用性。结果表明:订正前,2021年CLDAS日最高气温产品的平均误差为0.63℃,平均绝对误差为1.14℃,订正后平均误差减小至-0.03℃,平均绝对误差减小至0.64℃,误差小于或等于1℃的准确率由约64%提高到约90%,明显改善了该产品在重庆西部和东南部地区的适用性;订正前,2021年CLDAS日最低气温的平均误差为-0.22℃,平均绝对误差为0.75℃,订正后平均误差减小至-0.03℃,平均绝对误差减小至0.55℃,误差小于或等于1℃的准确率由约91%提高到约93%,改善了该产品在重庆中部地区的适用性。
文摘为提高智能网格的订正能力及预报水平,基于中央台客观指导产品的甘肃省切片数据和中国气象局陆面数据同化系统(Chinese Land Data Assimilation System Version 2.0,CLDAS-V2.0)日网格实况产品,采用卡尔曼滤波和滑动训练订正两种方法,对河西走廊东部地区(101.0°E—104.5°E,36.0°N—40.0°N)0.05°×0.05°格点最高、最低气温进行订正、检验和评估。结果表明:(1)季节对比,卡尔曼滤波和滑动训练订正产品对四季最高、最低气温的平均绝对误差均小于中央台客观指导产品,均小于2.00℃;卡尔曼滤波和滑动训练订正产品对四季最高、最低气温的预报准确率均大于70%,其中最高气温偏高6%~13%,最低气温偏高8%~24%。(2)空间对比,卡尔曼滤波和滑动训练订正产品对最高、最低气温的平均绝对误差绝大部分地区在1.00~2.00℃,个别地区大于2.00℃;卡尔曼滤波和滑动训练订正产品对最高(最低)气温的预报准确率大部分地区大于70%(60%~70%),个别地区大于80%(70%)。(3)总体上,卡尔曼滤波和滑动训练订正产品对最高、最低气温订正技巧基本为正技巧,个别季节和部分地区订正技巧大于0.300。说明两种订正方法具有较好的订正预报能力,可为今后的温度预报业务提供一定的技术支持。
