The transformer oil temperature is an important indicator that directly reflects the operational lifespan and load capacity of a transformer. Analyzing and predicting transformer oil temperature data is of significant importance in understanding the operational status and ensuring the safety of transformers. However, transformer oil temperature data belong to long sequential time series, and traditional time prediction models often encounter issues such as slow speed and convergence difficulties. Therefore, this paper proposes a transformer oil temperature prediction method based on the informer model, which enables the model to predict long sequential time series of oil temperature data. The effectiveness and accuracy of the model’s predictions are demonstrated through real transformer oil temperature data.