Research On Approaches Of Air Quality Monitoring Data Prediction Based On Fuzzy Cognitive Maps

The air quality monitoring network based on a large number of air quality micro-control stations and national monitoring and control stations has played an active role in scientifically and effectively controlling air pollution.At the same time,intensive grid monitoring also generates a lot of valuable data to be analyzed and mined.The large number of miniature measurement and control stations used in grid monitoring network deployment have the typical characteristics of low price,insufficient accuracy,and many sensor failures,which may lead to the problem of abnormal and unstable monitoring data in some micro measurement stations,as well as the lack of national measurement stations in some regions.In view of the above-mentioned difficulties in micro-stations,this article analyzes and mines historical monitoring data from a data-driven perspective,and conducts research on the prediction and calibration model of grid-based air quality micro-station monitoring data.The specific research contents are as follows:(1)For the problem of abnormal and unstable monitoring data of micro-stations,an air quality prediction model based on fuzzy cognitive map is proposed.First,from the historical data of the micro station,a fuzzy cognitive map with each air quality monitoring feature as a node is constructed.In this process,a fuzzy cognitive map ant colony optimization learning algorithm for air quality monitoring time series data is proposed.The obtained fuzzy cognitive map systematically reflects the interdependence of each monitoring feature.Then,the historical monitoring time series data is used as the initial value of the state of the fuzzy cognitive map,and the iterative output of each node is used as the predicted value.Finally,the effect of the proposed model is analyzed through the actual data of the specific environment.Experimental results show that the overall prediction effect of the algorithm proposed in this paper is close to the long-term and short-term memory network,which is superior to other typical regression algorithms and superior to all comparison algorithms in ozone prediction.In addition,compared with other algorithms,the proposed algorithm has the advantage of low computational complexity.(2)To solve the problem of data calibration in some regions without national survey stations,a fuzzy cognitive map prediction calibration model based on multi-objective ant colony optimization is proposed.First of all,the historical monitoring data of the micro-station and the national station are unified through the linear regression model.Then,through the standardized data,construct a fuzzy cognitive map for various monitoring features.In this process,this paper proposes a fuzzy cognitive graph learning algorithm based on multi-objective ant colony optimization.This method eliminates the abnormal problem caused by the mutation of the original data of the micro-station by adding the reliable data of the national station during the search.The deviation of the prediction model makes the prediction data more reliable,and at the same time retains the geographical location of the micro-station and the national station.The data characteristics of the grid distribution can provide true and reliable grid monitoring data.Finally,the effectiveness of the proposed model is analyzed through actual environmental monitoring actual data.Experimental results show that the algorithm proposed in this paper is superior to other algorithms in terms of CO、NO2 and SO2 prediction,and the computational complexity lower than other algorithms.