Research On Fault Diagnosis And Pre-warning Of Transformer On Offshore Platforms

As an important power source for offshore oil production,it is vital to maintain long-term stable operation of the offshore platform power system.As the key equipment connecting different voltage levels,transformers play an important role in the stable operation of the offshore oil power system.Compared with the land-based power grid,the power system in the offshore oil platforms has a more severe working environment for each equipment and the insulation material of electrical devices is subjected to more significant stresses.These stresses can cause damage to the transformer,inducing various types of faults,and then affecting the reliable operation of the transformer.On the other hand,offshore platforms are generally far from land,equipment maintenance and replacement are difficult to implement.However,the traditional periodic maintenance does not adapt to the high reliability requirements of offshore platform transformers,which may lead to the false alarm or omission of transformer faults.For this reason,the transition to condition maintenance is needed to enhance online monitoring and maintenance of transformers,and transformer fault diagnosis and pre-warning methods need to be focused on as the key means to achieve online monitoring.Therefore,this paper focuses on improving the operational reliability of offshore power systems,and conducts research on fault diagnosis and pre-warning methods for oil-immersed and dry-type transformers.Transformer aging factors and fault mechanisms are deeply analyzed,and the performance of existing monitoring methods are evaluated.And then,various types of sensing monitoring data and artificial intelligence algorithms are used to establish fault diagnosis and pre-warning methods for transformers in offshore power systems.(1)Based on the offshore operating conditions,the aging factors,fault mechanisms and existing diagnostic methods of each type of transformer are analyzed in detail.The main factors affecting transformer insulation,including thermal aging,electrical aging,mechanical aging and salt spray are analyzed in the context of the offshore environment,and the mechanisms and types of major transformer faults(thermal and electrical faults)are described.On the basis of this,the performance of traditional Dissolved Gas Analysis(DGA)methods is studied with actual data,and the current situation of dry-type transformer fault monitoring is analyzed.(2)For oil-immersed transformers,a method of fault diagnosis using improved DS evidence theory fused with multiple probabilistic output algorithms is proposed to improve the accuracy of fault diagnosis of oil-immersed transformers.Firstly,the performance of soft classification algorithm and hard classification algorithm is compared and analyzed.After that,three models based on Multiclass Relevance Vector Machine(MRVM),Multiclass Support Vector Machine(MSVM)and Back Propagation Neural Network(BPNN)are established;in order to improve the accuracy of the model,Particle Swarm Optimization(PSO)are used to optimize the hyperparameters in the model.Finally,the fusion of the three probabilistic output models using improved DS evidence theory is proposed to obtain more accurate fault diagnosis results.The research result shows that this method achieves probabilistic output for fault diagnosis of oil-immersed transformers and overcomes the deficiency of the traditional DGA methods which are difficult to summarize the fault development law inductively.(3)For dry-type transformers,a temperature abnormality warning method based on Sparse Bayesian Learning(SBL)and sliding window analysis is proposed,which can effectively identify abnormal temperature rise of dry-type transformers.First,a dry-type transformer temperature prediction model is established by using the SBL,and the SBL's parameters are optimized by using PSO algorithm,which effectively improves the model accuracy.After that,the temperature warning range is constructed based on the temperature expectation and variance obtained from the output of SBL.Finally,the statistical analysis of the temperature residuals using sliding windows can effectively distinguish the fault temperature rise from the sensor measurement error and avoid the occurrence of misdiagnosis.The research result shows that the proposed method achieves a more objective and effective abnormal temperature warning of transformers.Through the above research,the transformer aging factors,fault mechanism and defects of existing diagnosis methods are clarified,and the problems of low accuracy of oil-immersed transformer fault diagnosis and dry-type transformer temperature threshold through human subjective setting are better solved to achieve transformer diagnosis and pre-warning,and then improve the operational reliability of transformers.