Study And Application Of Air Pollution Prediction Method Based On EEMD-GRU

In recent years,environmental problems have become increasingly serious.Haze pollution has a serious adverse impact on the environment,human health and social economy.In order to control haze pollution,the state and the government have taken many specific measures and achieved some results,but haze pollution is still serious,and the traditional treatment measures have failed to achieve good results.In the face of the haze problem,it is an important means to deal with the haze problem to detect the climate quality in advance,predict and understand the air pollution situation in time,and understand the protection of public health by providing early warning and guidance.It is more and more necessary to develop more agile and intelligent monitoring methods in environmental protection.In this paper,taking Beijing as an example,taking the meteorological data and air pollution data of Beijing for five years,in order to predict the PM2.5 content of air pollution,a hybrid model based on the combination of time series signal data processing and deep learning neural network algorithm is proposed,which is applied to the PM2.5prediction and application in this project.It mainly decomposes the haze big data into multi-modal time series data and then carries on the deep learning,integrates its results and carries on the forecast,improves the model forecast accuracy.The main research work of this paper is as follows:(1)This paper introduces the research background and current situation of this topic.According to the research status of equal time series data prediction at home and abroad,it summarizes the existing prediction methods,summarizes the shortcomings and progress space of the existing prediction methods,and puts forward the main research and innovation of this paper.(2)According to the characteristics of this research project,the time series model is classified and summarized.A hybrid algorithm model based on the combination of Ensemble Empirical Mode Decomposition(EEMD)and Gated Recurrent Unit(GRU)neural network(EEMD-GRU)is proposed.The algorithm divides the sequence into two parts,in which EEMD algorithm decomposes the sequence,from single mode decomposition to multi-mode,mining the inherent characteristics of the data to make the model carry out multi-mode decomposition feature learning,then using the training data set to establish GRU neural network,the PM2.5 sequence decomposition of sub models and meteorological data long-term dependence on features to model learning,and finally using the inverse EEMD operation to integrate the prediction value of each sub model to get the final output value.(3)The algorithm is applied to the research object of this paper,Beijing meteorological data and air pollution data.By setting up five comparative experiments and doing experimental analysis,we can compare and evaluate the learning situation of different models.(4)According to the prediction algorithm proposed in this paper,PM2.5 prediction system is designed and implemented,and the system needs analysis and database design are done,the system framework structure is described in detail,and the main function modules of the system are designed and implemented.