A temperature-based fault pre-warning method for the dry-type transformer in the offshore oil platform.

• Temperature is one of the most important indicator to reflect the operating condition of dry-type transformers. • The Sparse Bayesian Learning algorithm is used to establish the temperature model. • The temperature range for the transformer under normal operating conditions is derived. • The temperature residual statistical analysis can distinguish the operating abnormalities from the measurement errors. In the offshore oil platform power system, the replacement of electric equipment is very inconvenient due to the far distance between the platform and the land. Transformer is one of the most important electric equipment in the oil power system, whose reliable operation is of essential importance to the normal oil exploitation. Therefore, the condition of the transformer should be monitored continuously to find out the possible operating abnormality as soon as possible. The transformer temperature is a good indicator to reflect the transformer operating condition. Moreover, there is a large amount of historical operating data provided by the monitoring system in the oil platform power system, which can be used as the basis to distinguish the normal and abnormal operating condition of the transformer. In this paper, an abnormal temperature pre-warning method is proposed for the dry-type transformer in the oil platform power system. Based on the historical operating data, the dry-type transformer temperature model is established by the Sparse Bayesian Learning. The proposed model provides a temperature warning range. By entering the current transformer operating parameters into the model, the temperature range for a normal operating state can be obtained. If the temperature measured by the transformer sensors exceed the expected ranges, an abnormality may occur. The residual statistical analysis is adopted to distinguish the measurement errors and actual transformer abnormal operations. The effectiveness and validity of the proposed method are verified based on the real field data of an oil platform transformer. [ABSTRACT FROM AUTHOR]