针对传统空气质量监测系统耗电量大、抗干扰能力差、稳定性不高以及空气质量指数（air quality index, AQI）预测精度不足等问题，设计了一种集低功耗广域物联网LPWAN、One-NET云平台和循环神经网络（gated recurrent unit, GRU）于一体的空气质量监测系统。系统采用光伏为主、市电为辅的混合模式供电；利用远距离无线电LoRa技术采集环境参数，集合云平台技术、麻雀搜索算法-变分模态分解-循环神经网络（sparrow search algorithm-variational mode decomposition-GRU, SSA-VMD-GRU）耦合模型实现远程监控和预测AQI指数。通过通信测试，结果表明通信距离1 000 m内，通信率在96%以上，丢包率不超过4%。将采集到的特征参数用传统的GRU模型、VMD-GRU模型和本文提出的SSA-VMD-GRU模型进行训练、测试仿真和对比，结果表明SSA-VMD-GRU模型相较于传统的GRU模型和VMD-GRU模型对AQI指数有更好的预测效果，均方根误差分别减小了22.434、0.833，平均绝对误差分别减小了16.849、0.623，预测误差率在3%以内。该系统能够实现对空气质量的实时监控和AQI指数的精准预测，为准确发布空气质量预警提供借鉴。

Aiming at the problems of large power consumption, poor anti-interference ability, low stability and insufficient prediction accuracy of AQI (air quality index) of traditional air quality monitoring system, an air quality monitoring system integrating low-power wide-area IoT LPWAN, One-NET cloud platform and gated recurrent unit was designed. The system adopted a hybrid mode power supply based on photovoltaics and supplemented by mains power. Long-distance radio LoRa technology was used to collect environmental parameters, and the cloud platform technology and SSA-VMD-GRU (sparrow search algorithm-variational mode decomposition-GRU) coupling model were integrated to realize remote monitoring and prediction of AQI index. Through the communication test, the results showed that within the communication distance of 1 000 meters, the communication rate was above 96%, and the packet loss rate was not more than 4%. The results showed that the SSA-VMD-GRU model had a better prediction effect on AQI index than the traditional GRU model and VMD-GRU model, the root mean square error decreased by 22.434 and 0.833, the mean absolute error decreased by 16.849 and 0.623, respectively, and the prediction error rate was within 3%. The system can realize real-time monitoring of air quality and accurate prediction of AQI index, providing reference for accurate issuance of air quality warnings.