IOT DHT22 sensor data

The accuracy of temperature & humidity prediction directly affects indoor environmental control, and current predictions mainly focus on time modeling, lacking spatiotemporal modeling based prediction for distributed sensors installed in buildings. Therefore, this article proposes an indoor temperature and humidity prediction method based on spatiotemporal modeling and transfer learning of informer. A IOT platform is designed with 8 temperature & humidity integrated sensors in a public building. The node closest to the air conditioner is assigned as the two-dimensional spatial coordinate origin, and a traversal spatiotemporal dataset is constructed according to different combinations of time interval and distance. The transfered Informer model is adopted to predict the temperature & humidity of the other nodes based on the data of the coordinate origin. Two prediction models are performed including weather classification based spatiotemporal model and full parameter spatiotemporal model. The experimental results show that the overall prediction error of the spatiotemporal model based on weather classification is relatively small, but the universality of full parameter spatiotemporal model is better.