Multi-region precipitation prediction model construction method based on multi-graph convolution and memory network

Abstract

The invention discloses a multi-region precipitation prediction model construction method based on multi-graph convolution and a memory network. The method comprises the following steps: establishingat least two adjacent matrixes; constructing a plurality of regional precipitation input data sets and calculating dependence attention scores; inputting the dependence attention score into a pre-constructed LSTM memory network to obtain an output value, and respectively inputting the output value into a pre-constructed graph convolutional neural network; summing the outputs of all the graph convolutional neural networks, inputting the summed outputs to a full connection layer for feature regression prediction to obtain precipitation prediction output of the whole network, training the whole model according to the error until the error meets a preset value, and obtaining a final multi-region precipitation prediction model. According to the method, the time sequence characteristics existingin the precipitation process are fully utilized, the situation that in the prior art, only various variable values at a single moment are considered to predict the precipitation, and the change trendinformation of the variable values along with time is lost is avoided, and the precipitation prediction accuracy is effectively improved.

Description

technical field [0001] The invention belongs to the field of multi-regional precipitation prediction methods, in particular to a method for constructing multi-regional precipitation prediction models based on multi-image convolution and memory networks. Background technique [0002] Precipitation, as an important part of the cycle process of the hydrological system, plays an extremely important role in the entire water cycle, and the precipitation has the characteristics of rapid changes in a short period of time, and it is precisely because so many areas are prone to natural disasters such as drought and flood. Accurate prediction of precipitation in multiple regions plays a decisive role in early warning of natural disasters such as flood control and drought prevention, and water resource scheduling decisions. The region where the region is located has a great influence, which leads to the low prediction accuracy of a large number of precipitation prediction models propose...

AI Technical Summary

Problems solved by technology

The most important thing is that the precipitation process has time-series characteristics. Only the environmental factors at the current moment are input into the deep belief network to predict the precipitation, ignoring the change trend of various environmental factors before this moment. The characteristics of drastic changes make it difficult for the model to have strong generalization and robustness, resulting in low precipitation prediction accuracy

Images