Multi-step temperature prediction model based on surrounding cities and long-term memory neural networks

In this paper, a multi-step temperature prediction model based on geographic location and long-term memory neural network is proposed. In order to verify the validity of the data of surrounding cities, two sets of comparative experiments are carried out which using 80-dimensional data of weather data added to surrounding cities and not joining the surrounding area. Urban weather data 8D data, 10D data added to surrounding city temperature data, and 1D data without temperature of surrounding cities. The test results show that the model of adding weather data of surrounding cities in most cases has better prediction effect than the model of weather data without surrounding cities. The results show that the model of weather data added to surrounding cities has better prediction effect.