Methods and systems for predicting air quality in smart cities based on an internet of things

A system for predicting air quality in a smart city based on an Internet of Things, comprising a user platform, a service platform, a management platform, a sensor network platform, and an object platform, wherein

• the sensor network platform is configured to obtain environmental monitoring data obtained by the object platform and an actual air pollution-situation of a target area, wherein the environmental monitoring data includes at least one of air quality data, weather data, and satellite image data;
• the management platform is configured to predict an air pollution situation of the target area through a regional prediction model based on the environmental monitoring data, wherein the regional prediction model is a machine learning model;
• the management platform is further configured to determine deviation data between the air pollution situation of the target area predicted by the regional prediction model and the actual air pollution situation, and determine, based on the deviation data, a suspicious area, wherein the suspicious area is an area likely to include a pollution source;
• the management platform is further configured to determine the pollution source, including:
• obtaining image data of the suspicious area, and processing, through a pollution source recognition model, the image data of the suspicious area to determine the pollution source, wherein the pollution source recognition model includes an image feature extraction layer and a determination layer, the image feature extraction layer is used to extract an image feature vector of the suspicious area based on the image data of the suspicious area, the determination layer is used to determine the pollution source based on the image feature vector of the suspicious area and an air quality feature;
• the service platform is configured to determine, based on the air pollution-situation of the target area or the pollution source of the suspicious area, prompt information and send the prompt information to the user platform, wherein
• the target area is a hexagonal area, and an input of the regional prediction model includes features of the hexagonal area and other hexagonal areas adjacent to the hexagonal area;
• the features include time-series features of the hexagonal area and the other hexagonal areas, and the time-series features include a time feature, a traffic feature, and a meteorological feature;
• wherein the traffic feature includes a traffic flow and congestion situation;
• the regional prediction model includes seven regional models and an air quality determination model, the seven regional models respectively correspond to the hexagonal area and the other hexagonal areas adjacent to the hexagonal area, each of the seven regional models processes the time-series feature of the corresponding hexagonal area to output a corresponding time-series feature vector, and the air quality determination model processes the time-series feature vectors output by the seven regional models to determine the pollution situation of the target area;
• wherein
• the regional prediction model is obtained by joint training based on training data and a label, wherein the training data includes sample time-series features of a sample hexagonal area and other sample hexagonal areas adjacent to the sample hexagonal area and sample time-series feature vectors, and the label is an actual air quality in a sample target area, including:
• inputting the sample time-series features of the sample hexagonal area and the other sample hexagonal areas adjacent to the sample hexagonal area into seven preliminary regional models to output the sample time-series feature vectors;
• inputting the sample time-series feature vectors into a preliminary air quality determination model to output an air pollution situation in the sample target area;
• obtaining the regional prediction model by performing verification based on the air pollution situation in the sample target area and the actual air quality in the sample target area;
• a prediction mode of the regional prediction model is an iterative prediction, and the iterative prediction includes:
• iteratively predicting air pollution situation in the hexagonal area and the other hexagonal areas adjacent to the hexagonal area at a second time point through the regional prediction model based on air pollution situation in the hexagonal area and the other hexagonal areas adjacent to the hexagonal area at a first time point.