The Design And Implementation Of Air Quality Monitoring System Based On Federated Learning

In recent years,environmental pollution has developed into a major problem affecting people’s health.Accurate fine-grained detection of air quality is a prerequisite for air pollution prevention.However,sensory data is sparse due to inadequate monitoring locations and incomplete records,which is a major challenge for Air pollution prevention.To this end,this paper designs and implements an air quality monitoring system based on Federated Learning.Different from the traditional data concentration training,the design and implementation of system adopts the distributed model training method,and does not need to upload the training data to the server.Instead,paper proposes a new regional joint learning framework(FRL)for each regional sub-training area model.)to improve the quality of distributed training models.Based on this framework,the paper describes the design process of the entire FRL system,and builds and implements the whole system based on the Python language and TensorFlow platform.Finally,the paper also applies the system to Beijing PM2.5 monitoring for experimental evaluation.The experimental results show that compared with the conventional distributed model training,FRL has improved the accuracy of nearly 5%and 8%on RNN and CNN respectively,which is nearly 3 times and 5 times higher than that of the centralized training mode on RNN and CNN respectively.