关键词: 储粮温度预测, 改进粒子群算法, 粮食储藏, 通风控制

Abstract: In the current context of increasing concern about food security, accurate prediction of temperature fluctuations during grain storage and optimal management of grain storage environment through intelligent ventilation control system have become urgent problems. Based on this, a CNN-BiGRU-Attention network model is proposed, in which the potential relationship between the temporal data in the feature map is extracted by CNN and the processed feature vector is used as the input to the BiGRU network, and according to the temporal characteristics of the grain data, Attention is added to the BiGRU network to assign weights to the grain features, as well as the multimodel fusion algorithm of IPSO optimization model hyperparametric is used to predict the grain pile temperature. The prediction model was validated by using a dataset from a directly-affiliated grain depot in Yushu, Jilin Province, and the results showed that the root-mean-square error （RMSE） was 0.0469, the mean absolute error （MAE） was 0.031 5, and the coefficient of determination  R2 was 0.992 5, which effectively improved the prediction accuracy compared with other models. By applying the storage temperature prediction function to the grain condition measurement and control system, the intelligentization of mechanical ventilation is realized to guarantee the safe storage of grain.

Key words: grain storage temperature prediction, improved particle swarm algorithm, grain storage, ventilation control