Research On Air Quality Prediction Based On Deep Fusion Of Multi-dimensional Features

In recent years,due to the increase in energy consumption,the problem of air pollution has become increasingly serious.Air pollution has an important impact on agriculture,water resources,transportation and many other aspects.The national environmental protection department has been committed to solving air quality problems.The current atmospheric environment is still very serious.The prediction of pollutant concentration is of great significance and value for the early warning of serious pollution events,and the prediction of pollutant concentration is an important modeling task.In the existing methods,the ordinary convolutional neural networks fail to conduct a multi-dimensional and fine-grained modelling for the multi-dimensional gas flow correlation problem.In addition,traditional spatio-temporal data modeling only separately considers the internal correlation within either time or space,and does not consider the correlation between time and space.In response to this,a deep fusion air quality prediction based on multi-dimensional features is proposed.method.The main research contents and innovations of this article are as follows:This paper proposes a deep fusion air quality prediction method based on multi-dimensional features.First of all,compared with the ordinary convolutional neural network,when using the proposed multi-dimensional feature extraction convolutional neural network to model the flow relationship of the space gas,not only more knowledge but also more fine-grained information are learned,which improves the prediction accuracy.Secondly,the proposed method considers the information related to the pollutant value from three perspectives: space,time and space.Inspired by the idea of multi-tasking,it learns the interaction between multiple time and space tasks,so as to reduce the prediction bias and obtain a more accurate prediction for both temporal and spatial models.Finally,we conduct experimental exploration on the deep fusion air quality prediction method based on multi-dimensional features proposed in this paper,and make relevant comparative experiments.The data collected by the Global Data Assimilation System of the National Center for Environmental Prediction Global Forecast System is used as the experimental data set to prove the advantages and effectiveness of this method from different evaluation standards.The experimental results show that the prediction effect of the deep fusion air quality prediction method based on multi-dimensional features is generally better than that of the baseline method.It is worth noting that in the long-term prediction with stricter experimental conditions,such as the 19-24 hours,the prediction of the deep fusion air quality prediction method based on multi-dimensional features,compared with the baseline method,the MAE value decreases by 4.74,and the RMSE value decreases by 11.3,indicating a better performance.The results show that the method in this paper can predict the concentration of pollutants well.The experimental results show that the prediction effect of the deep fusion air quality prediction method based on multi-dimensional features is generally improved compared with that of the baseline method...