State Assessment Technology For Distribution Transformer Based On Deep Fuzzy Neural Network

The transformer is a large dynamic system,and its internal performance is constantly changing.When a certain limit is exceeded,the transformer enters an abnormal state and malfunctions.However,for a long time,the power system has adopted regular maintenance for transformer status monitoring.Regardless of whether the transformer is in a good state or even in a state where it has never failed,the maintenance and testing are carried out in accordance with the " Power Equipment Maintenance Test Regulations".As for the test major,all transformers are tested every 3 years or even every year,and the status of the transformers is checked regularly.This often causes unnecessary waste.Therefore,starting from various test data,this paper proposes a transformer evaluation method based on a deep fuzzy neural network.This method can continuously adjust the weight of the evaluation value through the self-adaptability of the neural network.Based on the analysis of the state parameters that affect the working state of the transformer,in this paper,a simpler triangular membership function model is used to transfor m the evaluation score into the form of membership to evaluate the state of distribution trans formers.Aiming at the insulation performance of the transformer,the three-ratio method is c ompared based on reliability,safety,availabilitycombined with the existing research results,and further uses the fuzzy control algorithm to construct a comprehensive evaluation criterion for the state evaluation of the distribution transformer.Establish a reasonable fuzzy neural network evaluation plan in terms of performance,anti-interference and other aspects.Use the deep learning algorithm to establish a state scoring mathematical model,convert the evaluation parameter value into a function with a value range from 0 to 100,use the ascending half ladder model to score the status information of the transformer breakdown voltage index,and use the descending half ladder model to evaluate the transformer oil Dissolved gas content is used for state scoring,and deep neural network is used to optimize and fine-tune the weight threshold.