Research On Lightning Forecasting Method Based On The WRF Model

In recent years,artificial intelligence technology has received widespread attention from the society,and the use of artificial intelligence to solve practical problems has become mainstream.The occurrence of thunderstorm meteorological disasters has caused serious harm to people and caused serious property losses to the country and society.Features such as the generation,intensity and direction of the thunderstorm are difficult to judge accurately by manual analysis.Traditionally,the prediction of thunderstorm natural disasters in meteorology is subject to artificial experience,which is labor-intensive and inefficient.There is no effective method for thunderstorm prediction in China.Therefore,how to use computer technology to improve the current domestic thunderstorm prediction method,making the thunderstorm prediction more accurate and reducing losses is a more urgent need.The lightning forecasting method with a wide range of applications in meteorology is combined with the comprehensive analysis of the dynamics and micro-physical parameters of the weather research and forecasting model to determine whether there will be lightning in the next two hours.The parameter data of the mode output has strong correlation in time and space.Each parameter data of the output contains information such as longitude,latitude,time,unit of measure and numerical value,which brings great challenges to lightning forecasting.In order to solve the difficulties caused by thunderstorm forecasting,we use these output information,this paper first proposes a lightning correction method based on position correction.In a certain latitude and longitude range,it is necessary to predict whether a position has lightning within a short period of time.Based on the actual measured lightning in the hour before the forecast,calculate the local offset vector and the local offset vector.The lightning prediction results outputted by the WRF mode are corrected to verify the feasibility of the lightning position prediction.From the experimental results,the method can effectively correct the lightning position.However,the scheme is based on the lightning data actually measured on the ground.The WRF mode can output the lightning forecast results for the next 12 hours,and the actual measurement.The implementation conditions of the program cannot be met and cannot be applied to the lightning forecast in actual weather.Therefore,we further improve the experimental model,apply the traditional machine learning method,and propose a lightning forecasting method based on the particularity of the data set outputted by the weather research and forecasting model.Design a tree model,first output the weather research and forecasting model.The parameter data is selected to select the parameters that contribute greatly to the lightning forecast.Then,using these parameters,the data of the Beijing-Tianjin-Hebei region for nearly three years is tested and short-term lightning forecasting is carried out.The experimental results show that the proposed method is superior to the traditional lightning forecasting method in meteorology.It only uses the three parameters of ice,snow and radon ion concentration ratio,TS score method,lightning 6-12 hour lightning accumulation forecast can reach 0.089,the traditional The parameterized lightning forecasting method is higher than 0.054 in 6 hours,the accuracy rate is 0.312,the 4KM neighborhood method is 0.220,the accuracy rate is 0.434,the 6KM neighborhood method score TS is 0.272,and the accuracy rate is 0.535.The experimental results show that our method is excellent.The current mainstream lightning forecasting method in meteorology.