Air quality prediction model and method based on improved LSTM

Abstract

The invention discloses an air quality prediction model and method based on an improved LSTM, and belongs to the technical field of environmental protection. According to the model, the internal circulation structure of the Vanilla LSTM is simplified and improved, certain parameters are reduced, the inclusiveness of the structure to an abnormal cell state is enhanced, and a final IV-LSTM structure is generated. Meanwhile, a data input and output model is improved, and related sites are selected through an LS-DTW algorithm to increase data input channels. Compared with a pure DTW algorithm, the similarity between the site selected by the LS-DTW and the target site is higher, and the data input correlation is higher, so that the prediction result of the whole model is more accurate.

Description

technical field [0001] The invention relates to an air quality prediction model and method based on an improved LSTM, and belongs to the technical field of environmental protection. Background technique [0002] In recent years, with the growth of the economic level, people have higher and higher requirements for health. Clean air is the basic need to maintain human health, and the Air Quality Index (AQI) is an intuitive measure of air quality. Therefore, the air quality index AQI can be predicted to know whether the air quality is good or bad. [0003] At present, people have tried various methods in the field of air prediction research, from the earliest established air pollution diffusion model through empirical knowledge, to traditional machine learning models such as linear regression and random forest, and various breakthroughs have also been made. . Later, with the continuous development of machine learning, some people linked deep learning networks and air quality ...

AI Technical Summary

Problems solved by technology

However, this form generates more parameters than S-LSTM. If the parameters increase, the weight update effect during each round of training will become weaker, which makes it more difficult to converge during the training process.

At the same time, the way VLSTM uses historical information makes the recursive training process more sensitive to new input data, which means that a small amount of abnormal data may cause a large delay in the convergence speed of the entire deep learning network, and the prediction accuracy under the same number of iterations will decrease. That is to say, although VLSTM has a better effect on accurate time interval prediction to a certain extent, it will make it more difficult for the entire deep learning network to converge during the training process, which will eventually lead to a decline in the prediction accuracy of the model.

Images