Short-time Prediction Of Urban Livability Climate Factors Based On Deep Learning

With the rapid development of cities and the significant improvement in the standard of living of their inhabitants,the discussion of livability occupies an important place in the process of urban development reaching an advanced stage.Climate is a prerequisite for the livability of a city and has a great impact on the lives of its inhabitants,including both the microclimate of indoor spaces and the macroclimate of outdoor spaces.However,current research on climate issues is a hot topic of social concern.In microclimate research,room temperature is an important indicator of human comfort.Accurately grasping the changing patterns and trends of indoor temperature and establishing accurate room temperature prediction models are important means of improving the indoor microclimate environment and are also key to achieving efficient and intelligent heating.In macro-climate research,air quality prediction is a hot spot and a difficult area of research in the field of sustainable urban development,especially PM2.5 concentration prediction has far-reaching significance for improving the macro-climate environment of cities,enhancing the quality of urban space and benefiting residents’health and public travel.Therefore,exploring methods for predicting climate factors is the key to studying climate problems.Deep learning,as a frontier direction in the field of artificial intelligence,is an effective method to address climate factor prediction,but its prediction accuracy is also a difficult problem currently faced.To this end,this paper conducts research on urban habitability climate factor prediction,mainly including room temperature prediction and PM2.5 prediction,based on deep learning methods,which include the following.First,build a room temperature data collection system based on Io T technology.The working principle of the system and the data transfer-storage path are described separately.The hardware equipment of the lower computer acquisition system and the design of the upper computer management system are described and the system is demonstrated respectively.The data obtained by the system will provide data support for room temperature prediction.Secondly,indoor microclimate studies are carried out.An LSTM-based room temperature prediction method is proposed.The correlation analysis of room temperature and its influencing factors is carried out,and the algorithm principles and network structures of LSTM,RNN and BP are described.Prediction experiments based on the three models are carried out and the LSTM is verified to have the highest prediction accuracy and the lowest error with an R~2 of 0.9839 by index evaluation.Thirdly,outdoor macro-climate studies are carried out.A short time PM2.5prediction model based on WVPBCM is proposed.The correlation analysis of PM2.5and its influencing factors is carried out,the data processing process using wavelet denoising,variational modal decomposition and principal component analysis methods is introduced,the model theory is elaborated and prediction experiments are carried out.The experimental results show that the WVPBCM model has the highest accuracy,with R~2 of 0.9438,and the error of the results does not exceed 10%,and the peak error is 1.8247.The accuracy remains the highest after increasing the prediction time.