Research On Temperature Prediction And Early Warning Of Transformers In Substations

Substation is a crucial component of the power grid transmission system.With the progress of society,technology and economy,people use more and more electrical equipment,the operation of such equipment depends on high quality electricity,so the reliability of power supply is directly related to the production and life of people.The substation is essential for normal power supply and needs to ensure that transformers,circuit breakers and other major electrical equipment are in a stable working condition,and if a fault or abnormal problem occurs,it will definitely affect the operation of electrical equipment and even bring serious consequences.Among the many faults,thermal faults caused by rising temperature account for most of them,so attention should be paid to the monitoring and analysis of the temperature of power equipment,research and optimize the prediction and early warning of the temperature of key parts of electrical equipment to ensure that the temperature is always in the normal operating range.This paper first examines the structure of substation transformers,the causes of transformer temperature rise,fault types and data characteristics.Then the transformer temperature prediction and fault warning methods are analyzed.Based on the existing temperature data,the data signal is preprocessed by the variational modal decomposition(VMD)and local mean decomposition(LMD),and then the processed data is predicted by BP neural network and whale optimization algorithm(WOA)optimized extreme learning machine(ELM)for the original data,and the prediction performance of these two decomposition methods and two prediction models in the context of this paper is studied to find the the most optimal prediction model.Finally,three early warning methods are used in this paper,which are surface temperature method,similar comparison method and relative temperature difference method,to determine whether to issue an early warning message or not.In this paper,a double decomposition limit learning machine hybrid optimization(LMDVMD-WOA-ELM)model based on whale optimization algorithm is proposed for transformer temperature prediction and research,which includes LMD and VMD double decomposition data processing algorithm,whale optimization algorithm,limit learning machine and evaluation method.In the MATLAB simulation experiments,the original data are predicted by the dual decomposition of local mean decomposition method and variable modal decomposition method(LMD-VMD),and then each component of the decomposition is substituted into the extreme learning machine(WOA-ELM)model of whale optimization algorithm,and the comparison model and comprehensive evaluation index are introduced.After experiments,the hybrid model proposed in this paper has higher accuracy and stability than other common models,and can be effectively used for transformer temperature prediction.