Research On Data Imputation Technology Of Intelligent Meteorological Gateway Based On GAN

The observation of surface meteorological elements is an important part of work in meteorological observation,which provides important basis for weather forecast and real-time meteorological element values that are closely related to people’s life,as well as climate analysis and scientific research.In recent years,with the rapid increase of meteorological element data and the improvement of meteorological observation technology,the demand for surface meteorological observation equipment is no longer only satisfied with a simple data collection equipment,but also puts forward further requirements for the application of artificial intelligence technology,Internet of Things technology and big data technology in the equipment,so as to obtain more intelligent and networked meteorological data products with reliable data quality and prediction ability.The emergence of intelligent meteorological gateway and its application in surface meteorological observation equipment provide development and research direction for this demand.With the development and application of intelligent meteorological gateway,the problems caused by missing meteorological data can not be ignored.During running of meteorological observation equipment,one or more measurement element data could be missing or abnormal due to damage of sensors or acquisition boards,abnormal network transmission or other reasons.The missing of meteorological element data could lose part of the information contained in the historical data,which will influence the data processing,analysis and prediction functions of the intelligent meteorological gateway.Reasonable and effective methods for processing missing values of meteorological element data are of great practical significance for the analysis and prediction of meteorological data.Most of the traditional missing data processing methods fail to take into account the correlation of meteorological element data in the time dimension and the correlation between meteorological elements,therefore,the imputation effect of these methods is not ideal.After fully studying and analyzing the time series characteristics of meteorological element data,this paper proposes a missing data imputation model named GAN-TRTI based on the improved generative adversarial network WGAN-GP.This model uses the idea of tracking-removed LSTM autoencoder to construct the generator,and the model can directly impute the missing data on incomplete data sets efficiently and accurately.In this paper,the GAN-TRTI model is compared with other missing data imputation methods such as mean imputation method,KNN imputation method and missing forest imputation method through experiments,and the experimental results show that the GAN-TRTII model has better imputation performance.Finally,the missing data imputation model was combined with the meteorological element prediction model based on BI-LSTM,and applied to the intelligent meteorological gateway deployed in three sets of intelligent meteorological observation equipment,which finally proved the effectiveness and practicability of the missing data imputation method and prediction method of meteorological element data proposed in this paper.