基于Attention-BiLSTM混合模型的月尺度降水量预测OA

Precipitation is affected by various meteorological factors,leading to low prediction accuracy.To solve this problem,multiple meteorological factors affecting precipitation were considered,and the attention mechanism was used to assign different weights to various meteorological factors.Combined with the bidirectional long short-term memory neural network(BiLSTM),an improved attention-BiLSTM model was proposed to predict monthly precipitation.By taking the Nanchang Meteorological Station in Jiangxi Province as an example,the observation data of monthly precipitation and meteorological factors(temperature,evaporation,pressure,etc.)from 1989 to 2018 were used as input data for the model.The attention mechanism identified the weights of various meteorological factors to improve the prediction performance of the BiLSTM model for precipitation.The results show that the attention-BiLSTM model can effectively improve the accuracy of precipitation prediction.Through the correction of the attention mechanism,the low precipitation prediction values by the original BiLSTM model are significantly improved.