Multi-parameter atmospheric environment data generation method based on stacked LSTM-GRU

Abstract

The invention discloses a multi-parameter atmospheric environment data generation method based on a stacked LSTM-GRU, and because an atmospheric environment data sample has the characteristics of multiple features and huge data volume, a large amount of time is needed for training by independently using an LSTM model. Compared with an LSTM, the GRU is simpler in structure, network parameters are reduced, and the convergence speed is higher. The LSTM and the GRU are combined, and on the premise that the LSTM effectively processes a large amount of atmospheric environment data, the network training speed is increased. Meanwhile, due to the fact that the single-layer network structure is too simple, the requirement for processing multiple features of atmospheric environment data cannot be met, and multi-parameter atmospheric environment data generation is achieved through a multi-layer network structure.

Description

technical field [0001] The invention relates to the technical field of meteorological monitoring, and more specifically relates to a method for generating multi-parameter atmospheric environment data based on stacked LSTM-GRU. Background technique [0002] Currently, accurate atmospheric environmental data play an important role in today's world, accompanied by the rapid development of computer technology. The operation and use of weapons and equipment must be completed in the natural environment, and the comprehensive natural environment in which it is located has a complex and profound impact on the combat performance of weapons and equipment. As an important part of the comprehensive natural environment, the atmospheric environment has the characteristics of multi-weather phenomena, multi-elements and time-varying compared with space, terrain and other environments. The wind speed, wind direction, precipitation, temperature, pressure, humidity, etc. in the atmospheric en...

AI Technical Summary

Problems solved by technology

However, at present, there are few methods of applying deep learning technology to atmospheric environment modeling at home and abroad. Most of the methods are based on a single parameter in the atmospheric environment, and do not consider simultaneously generating multiple atmospheric environment parameters at the same time; and because the atmospheric environment data samples have Due to the characteristics of many features and huge amount of data, the existing model training method needs a lot of time to complete the model training

Images