Research And Application Of Air Pollution Early Warning System

The air environment is the basic prerequisite for human survival and development.The deterioration of air quality may affect the living environment and living environment of cities,the impact of which can map to the material and spiritual aspects of individuals and multilevel society,and even threaten the national economy.With the development of era,high-quality environment and healthy air have gradually become a scarce resource.The improvement of air quality has become a governance requirement at the national strategic level.Under this situation,people's demand for air quality standards and environmental development is gradually increasing,and the importance of air pollution early warning prevention and control for the sustainable and healthy operation of society and economy would become increasingly significant.Currently,the role of air pollution early warning has become increasingly prominent.It is urgent to master modern management and governance methods.In order to improve the precise management level of the early-warning system,new explorations should be conducted in combination with large amounts of complex data and fusion of big data technologies.However,it is a requirement for air pollution early warning in the new era,such as that how to mine valuable information from complex and diverse data,make accurate judgments on the future trends and grades of air pollution,and then assist environmental protection departments to monitor and manage the air condition.Because there exist contain correlation among massive data and information in the air environment system,traditional methods and models can no longer effectively analyze and conduct complex related data.New research methods should be developed urgently to provide necessary technical support for air environmental issues,and then to solve the problems in the production and life of human society.Aiming at the above problems and research goals,this paper starts from the perspectives of data mining and processing,time series prediction and comprehensive evaluation analysis.Then the air pollution warning is built and applied based on data mining algorithms,machine learning theory,intelligent optimization algorithms and fuzzy mathematical theory.Aiming at the complex and multivariate air pollution data,the air pollution forecasting model and air quality evaluation model are constructed.Combined with the data pre-processing analysis method,it conducts an empirical study on the change of air pollution in typical cities,and further verifies the scientificity and operability of this system.The main content of this article includes the following three aspects:Firstly,this paper proposes a data pre-processing framework to provide a basis for modeling and data analysis for building air pollution early warning systems.The data pre-processing framework can provide a complete data source for the modeling,and realize the normalization of the data required for model training.It is beneficial to improve the model prediction performance and the efficiency of model operations with denoising the data after interpolation.Aiming at the missing data of air pollution information,this paper uses the piecewise cubic Hermite interpolation method to fill in the missing values existing in the original air pollution time series.This method can simulate the true change process of the data and supplement all the missing data.The improved complete ensemble empirical mode decomposition algorithm can decompose the original time series into high-frequency and low-frequency components,and then extract the trend and noise sequences,and filter the noise to obtain a stable data sequence.The empirical results show that the improved complete ensemble empirical mode decomposition algorithm can make the sample complexity of data sequence drop to the lowest compared after denoising with the singular spectrum analysis algorithm,the variational mode decomposition algorithm and the complete ensemble empirical mode decomposition algorithm.In addition,the denoising data can be standardized with the normalized processing method,through which the the problem of inconsistent chemotaxis properties and dimensions can be solved.Secondly,this paper proposes a new forecasting model based on the least squares support vector machine for air pollution prediction.The least squares support vector machine can transform the quadratic programming problem into solving linear equations,and then improve the computation speed and the efficiency of network convergence of the model.It plays an important role in improving the accuracy of time series forecasting.The forecasting model combines data pre-processing algorithm,and utilizes single-objective sine cosine and multi-objective sine cosine algorithm to optimize the regularization parameter and kernel width parameter of the least squares support vector machine.The purpose of optimization is to maintain the advantages of dealing with non-linear and multi-factor complexity problems,and improve the stability and accuracy of the algorithm.The empirical comparison shows that the forecasting model has high prediction accuracy.It also proves that the forecasting effectiveness and practicability of the least squares support vector machine through the comparison of multi-objective optimization models.Hence one can see that the establishment of the new forecasting model can monitor and predict the future real-time changes of air pollution,provide decision makers with reliable air pollution real conditions.Finally,this paper constructs an air quality evaluation based on fuzzy mathematical principles.The empirical results show that the evaluation model can obtain good and reasonable results.The basic idea of the model is to use fuzzy mathematical principles for describing and managing air pollution factors and evaluation level.Specifically,before performing comprehensive evaluation,the pollution factors affecting air quality should be determined firstly.Then,these pollution factors are expressed in the form of fuzzy sets,and the evaluation set can be determined based on the setting of ambient air quality standards and the needs of practical problems.Secondly,the fuzzy relationship matrix is calculated to determine the evaluation factor set by trapezoid membership function.The membership relationship of the evaluation factor set and the factor set would be evaluated.Then the sets can be performed fuzzy mathematical operations with the weight coefficients.Among them,the model uses the entropy weight method to weight each factor,and it does not need to calculate the weight of daily monitoring data.The process not only simplifies the fuzzy evaluation process,but also can greatly improve the information utilization rate and the reliability of the evaluation results,thereby the applicability of fuzzy mathematics in air quality evaluation.Finally,based on the principle of maximum membership,the air quality level can be determined.The proposed air quality evaluation method can make up for the shortcomings of the traditional air quality evaluation methods for evaluating air quality in terms of systematicness,objectivity,and comparability.Thus the comprehensive evaluation of air quality is achieved.The output of the constructed new evaluation model can not only provide environmental decision makers with early warning information on air pollution,but also remind relevant people to take effective conservatory measures in a timely manner according to the corresponding air quality level.Through the above studies,the main conclusions are as follows:(1)The construction of the air pollution early warning system requires effective data analysis tools as the research basis.Data is the basis of problem analysis and model building.This paper uses a data pre-processing analysis framework to analyze the air pollution data.On the one hand,the data pre-processing analysis framework can effectively supplement the missing data and solve the problem of inconsistent data dimensions in the modeling process;on the other hand,for the noise term of non-stationary time series data,the framework can effectively extract valuable information from the data by introducing denoising algorithm.It proves that the improved complete ensemble empirical mode decomposition algorithm has the best effect of denoising,which can significantly improve the performance of the forecasting model.Therefore,it plays a very important role in establishing the early warning system.(2)The successful application of the early warning system benefits from the effective improvement and reasonable optimization of the model.The least squares support vector machine has a good effect on non-linear data fitting,but the good effect should be established on the basis of stationary time series and reasonable setting parameters of the algorithm.The new constructed model in this paper makes effective treatment of the non-stationarity of historical time series by combining denoising algorithm;the intelligent optimization algorithm is a very effective method for parameter optimization.In this paper,the convergence efficiency of the algorithm can be significantly enhanced by improving the sine and cosine optimization algorithm.Through the empirical analysis,it proved that the single-objective optimization can reduce the error of the forecasting model,and multi-objective optimization can take into account the coverage probability and interval average width of interval prediction.(3)The accurate and objective early warning result mainly depends on the fusion and complementation of fuzzy mathematical methods.The constructed newly evaluation model in this paper effectively incorporates fuzzy mathematical theory,which can convert the evaluation factors and evaluation levels of air quality into the form of fuzzy sets,so that the fuzzy membership function can be combined with air pollution concentration limits to effectively calculate the air quality level.In addition,the model utilizes information entropy to set fuzzy weight sets and the relative importance of various pollution factors to air quality is considered comprehensively.The analysis of the early warning result demonstrates the rationality of the air pollution early warning system.The main innovations of this research include:(1)the single time-varying regression model in the existing research does not consider the instability of input data and the adaptability of model prediction parameter optimization.Based on the analysis of the non-linear and non-stationary characteristics of the data,this paper combines the least square support vector machine and the swarm intelligent optimization algorithm is proposed to tune the parameters.Through the empirical comparison,it is found that the single objective optimization of swarm intelligent optimization can effectively improve the forecasting performance,and the multi-objective optimization can guarantee the coverage probability and average width of the interval prediction;(2)for the air quality evaluation,a comprehensive evaluation model is proposed using the pollution factors which affect the air environment.The evaluation model based on the principle of fuzzy mathematics can transform the pollution factors and evaluation levels of the air environment into the form of fuzzy sets.Membership function and air pollution concentration limit are assisted to evaluate air quality,and the evaluation model achieves reasonable model effect.The model can effectively avoid the subjective preference of different environmental decision makers,and make up the incomparable gap of air quality level caused by the difference of air quality standards in the existing research;(3)in the research of data pre-processing,this paper puts forward a preprocessing analysis framework through comparing and selecting the conventional data processing methods and noise reduction processing methods.All the missing information of the original time series is interpolated by the framework.By removing the noise information from the data,the prediction performance of the model can be graranteed.And the results of empirical analysis can prove that the improved complete ensemble empirical mode decomposition algorithm has the best noise reduction effect.The shortcomings of this study include:This paper uses several single methods and auxiliary tools to construct the model.It is undeniable that there may be other alternative algorithms or tools.Due to the limited space,this paper does not carry out a detailed study of all the algorithms,so the theoretical research still needs to be further explored.Due to the objective unavailability of relevant data,this paper does not consider the impact of environmental investment decision-making on the change of air pollution.In the future,it will carry out relevant research from the perspective of the impact of environmental investment changes on air quality,so as to provide a more solid research basis for the sustainable development of environmental air environment.