This paper presents a novel artificial intelligence (AI) based approach to predict crucial meteorological parameters such as temperature,pressure,and wind speed,typically calculated from computationally intensive weat...This paper presents a novel artificial intelligence (AI) based approach to predict crucial meteorological parameters such as temperature,pressure,and wind speed,typically calculated from computationally intensive weather research and forecasting (WRF) model.Accurate meteorological data is indispensable for simulating the release of radioactive effluents,especially in dispersion modeling for nuclear emergency decision support systems.Simulation of meteorological conditions during nuclear emergencies using the conventional WRF model is very complex and time-consuming.Therefore,a new artificial neural network (ANN) based technique was proposed as a viable alternative for meteorological prediction.A multi-input multi-output neural network was trained using historical site-specific meteorological data to forecast the meteorological parameters.Comprehensive evaluation of this technique was conducted to test its performance in forecasting various parameters including atmospheric pressure,temperature,and wind speed components in both East-West and North-South directions.The performance of developed network was evaluated on an unknown dataset,and acquired results are within the acceptable range for all meteorological parameters.Results show that ANNs possess the capability to forecast meteorological parameters,such as temperature and pressure,at multiple spatial locations within a grid with high accuracy,utilizing input data from a single station.However,accuracy is slightly compromised when predicting wind speed components.Root mean square error (RMSE) was utilized to report the accuracy of predicted results,with values of 1.453℃for temperature,77 Pa for predicted pressure,1.058 m/s for the wind speed of U-component and 0.959 m/s for the wind speed of V-component.In conclusion,this approach offers a precise,efficient,and wellinformed method for administrative decision-making during nuclear emergencies.展开更多
文摘This paper presents a novel artificial intelligence (AI) based approach to predict crucial meteorological parameters such as temperature,pressure,and wind speed,typically calculated from computationally intensive weather research and forecasting (WRF) model.Accurate meteorological data is indispensable for simulating the release of radioactive effluents,especially in dispersion modeling for nuclear emergency decision support systems.Simulation of meteorological conditions during nuclear emergencies using the conventional WRF model is very complex and time-consuming.Therefore,a new artificial neural network (ANN) based technique was proposed as a viable alternative for meteorological prediction.A multi-input multi-output neural network was trained using historical site-specific meteorological data to forecast the meteorological parameters.Comprehensive evaluation of this technique was conducted to test its performance in forecasting various parameters including atmospheric pressure,temperature,and wind speed components in both East-West and North-South directions.The performance of developed network was evaluated on an unknown dataset,and acquired results are within the acceptable range for all meteorological parameters.Results show that ANNs possess the capability to forecast meteorological parameters,such as temperature and pressure,at multiple spatial locations within a grid with high accuracy,utilizing input data from a single station.However,accuracy is slightly compromised when predicting wind speed components.Root mean square error (RMSE) was utilized to report the accuracy of predicted results,with values of 1.453℃for temperature,77 Pa for predicted pressure,1.058 m/s for the wind speed of U-component and 0.959 m/s for the wind speed of V-component.In conclusion,this approach offers a precise,efficient,and wellinformed method for administrative decision-making during nuclear emergencies.
文摘次季节预测在农业规划、防灾减灾和水资源管理等领域具有重要意义。基于人工智能的“风顺”次季节预测模型(CMA-AIM-S2S-Fengshun),结合自主研发的CRA-40再分析数据和FY-3E卫星数据,通过级联Swin Transformer模块和智能扰动生成技术,实现了气候多要素集合预测。对2017—2021年中国区域降水的历史回算检验表明,“风顺”在逐候平均降水预测中的表现显著优于欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts,ECMWF)模式,整体技巧提升18.6%,其中华南地区提升41.2%,东部地区提升26.5%。在MJO(Madden-Julian Oscillation)预测方面,“风顺”将技巧保持时间延长至32 d(CRA-40驱动),超过ECMWF的30 d基准。个例分析显示,模型对2024年7月中旬华北强降水过程的落区和强度预测精度更高,提前3~4候捕捉到关键异常信号。
