A prediction model of grain storage temperature based on the improved feature enhanced broad learning

The prediction of grain storage temperature is a crucial measure for ensuring safe grain storage. Traditional machine learning models are limited by insufficient accuracy, while deep learning methods offer high accuracy but suffer from long training times. Based on this, an improved feature-enhanced broad learning prediction model is proposed. This model makes full use of the advantages of broad learning training speed and uses a long short-term memory network to enhance the learning ability of the model for the time series characteristics of grain storage data. Combined with the multi-head self-attention mechanism, the important correlation information focusing on multi-dimensional features is enhanced, and the prediction test of grain storage temperature is carried out. The results show that the grain storage temperature prediction accuracy of the improved feature-enhanced broad learning is higher than that of the unimproved broad learning based on some real grain storage data. At the same time, compared with the convolution-long short-term memory network and Transformer deep learning model, the proposed broad learning model has higher accuracy and less training time. This study provides an effective alternative scheme for safe grain storage in technology.