In the process of steel production and processing, the billet discharge temperature directly affects the quality of steel. In order to accurately predict the billet discharge temperature, an improved Pelican Optimization Algorithm (IPOA) and Long Short Memory Neural Network (LSTM) combined with a rolling temperature prediction model were proposed. Firstly, the data is processed using principal component analysis (PCA), and then the improved Pelican optimization algorithm is used to find the optimal parameters of LSTM. Finally, an IPOA-LSTM steel rolling temperature prediction model based on principal component analysis is established, and compared with LSTM model and IPOA-LSTM model. The root mean square error of the IPOA-LSTM model based on principal component analysis is 3.2763, the average absolute error is 2.1161, and the determination coefficient R2 is 0.9582, Compared with the other two models, it has higher prediction accuracy.