In order to solve problems such as drift, indication error, and jamming caused by factors such as working conditions and environment after long-term operation of oilfield automatic metering instruments. And the diagnosis of instrument failures in oil fields still relies on manual inspection or on-site experience. The real-time production data of the instrument is analyzed, and the characterization parameters of instrument faults in complex environments are studied in this thesis. And a Transformer based fault diagnosis model for automated instruments is built based on time series in a data driven manner. The comparative experimental results show that the accuracy, recall, and F1 scores of this model are 93.4%, 96.5%, and 95.3%, respectively, and its performance and prediction effect are better, and the diagnostic accuracy rate in field application can reach 92.8% when applied in the field, it is concluded that this model can be effectively applied to fault diagnosis of oilfield automation instruments.