Research On Calibration Method Of Grid Air Quality Monitoring Data

The grid air pollutant monitoring system constructed by a large number of small,cheap,movable micro stations and a small number of expensive,large,high-precision national measurement stations,it provides a strong support for the control of air pollution.However,the monitoring data of a large number of micro stations have the challenges of cross interference,abnormal drift,time dependence and uncertainty,which are serious barriers to data analysis and mining.In response to the above-mentioned challenges,this paper conducts research on grid air quality monitoring data calibration methods from the perspective of air quality monitoring,signal processing and deep learning.The specific research content is as follows:(1)Aiming at the cross-interference between the monitored different gas attributes,the dependence on the time dimension,and the abnormal drift of sensor data,a grid pol-lutant data calibration model based on time convolutional networks and gated cycle units is studied.First,the time convolutional network is used to extract the cross-interference characteristics between various pollutants,and the time-dependent characteristics of the monitoring data and the data drift phenomenon are learned through the gated loop unit,and finally the calibration results are obtained through the training of the fusion module.The results of the calibration experiment and the ablation experiment proved the validity of the calibration model and the various modules in the model.(2)Aiming at the abnormal problem of calibration model error caused by uncertain data affected by natural phenomena,the algorithm optimization research based on gated recurrent unit and modal decomposition is carried out.First,the uncertainty measurement of air pollution data with uncertainty and randomness is carried out,and the instability of the original data is reduced through modal decomposition.The decomposed components with different features are extracted separately in the form of multiple channels,and finally the learned features are superimposed and input into the gated loop unit to obtain the calibration result after the uncertainty measurement.The calibration experiment proves the importance of modal decomposition in the preprocessing of unstable data.