Analysis Of Transformer Vibration Signal And Prediction Of Fundamental Frequency Amplitude

As an important hub for power transmission and distribution,transformers have a close influence on the safe operation and economic benefits of the power grid.Theoretical research and application practice show that the fundamental frequency amplitude of the vibration signal on the surface of the transformer is an important basis for judging the state of the transformer,and there is a non-linear mapping relationship with the data of the transformer operating condition.Predicting the fundamental frequency amplitude based on the transformer operating condition data and comparing and analyzing the deviation between the predicted value and the actual measured value is an effective means to judge whether the transformer is abnormal.Therefore,it is of great significance to predict the amplitude of the fundamental frequency accurately and quickly.Appropriate models and model parameters are the key to accurate prediction of fundamental frequency amplitude.This thesis proposes a fundamental frequency amplitude prediction model based on Hybrid Artificial Bee Colony algorithm(HABC)and Extreme Learning Machine(ELM).ELM is good at dealing with nonlinear problems,and has the advantages of fast running speed and low human interference factors.However,its input layer weights and hidden layer biases are randomly generated,which also makes the output results have certain volatility.Artificial Bee Colony algorithm(ABC)has the advantages of high search accuracy,easy operation,strong robustness,etc.It can optimize the parameters of ELM,but the algorithm also has the defects of slow convergence speed and weak local exploitation ability.This thesis improves the ABC algorithm and uses it to optimize the input layer weights and hidden layer biases of ELM to further improve the prediction accuracy of the fundamental frequency amplitude.The main research content is divided into three parts:(1)Analyze the correlation between the fundamental frequency amplitude of the vibration signal and operating condition data of the transformer.On the basis of analyzing the frequency spectrum characteristics of the vibration signal on the surface of the transformer,Gray Relation Analysis algorithm is used to analyze the correlation between the fundamental frequency amplitude of transformer vibration and the state variables in the operating condition data,and the three state variables of operating voltage,load current,and oil temperature are selected as the prediction related vectors of the fundamental frequency amplitude.(2)Modified the shortcomings of the ABC algorithm and propose the HABC algorithm.First of all,the Opposition-Based Learning is adopted in the food source initialization stage to make food source distribution more reasonable;Secondly,the search formula of each bee species is modified,allowing the bee colony to search for a new food source near the optimal food source and being guided by the optimal food source at all times during the search process,and embed the Cauchy mutation operator to enhance the ability of the algorithm to jump out of the local optimum;Finally,five standard test functions are used for testing,which verifies the effectiveness of the modified method in this thesis.(3)Establish the HABC-ELM prediction model and complete the prediction of the fundamental frequency amplitude.The HABC algorithm is used to iteratively optimize the input layer weights and hidden layer biases of the ELM model to establish the HABC-ELM prediction model;The three state variables of operating voltage,load current,and oil temperature are used as the input of the model to predict the amplitude of the fundamental frequency.Carry out experiments with measured data at different positions on the surface of transformers,verify the effectiveness and feasibility of the proposed model.