Transformer Fault Diagnosis Based On Meta-heuristic Algorithm Optimized Probabilistic Neural Network

With the rapid development of the social economy,the demand for electricity is increasing,and the capacity of the grid is also increasing.However,the power transformer plays a pivotal role in the transmission and transformation process of the grid,so the excellent performance of the power transformer is directly Affect the stable and safe operation of the power grid.Any of these failures may cause power outages and other huge economic losses.Therefore,it is very important to detect potential faults in power transformers,especially for online monitoring and diagnosis without affecting the normal operation of the power network,which will avoid a lot of unnecessary losses.In recent years,with the continuous improvement of online monitoring technology and artificial intelligence technology,the advantages of dissolved gas analysis technology in transformer fault diagnosis have gradually expanded.This paper uses meta-heuristic algorithm to optimize the fault diagnosis method of probabilistic neural network.First of all,due to the complex internal structure of the transformer,the uncertainty of the status information and other factors,the accuracy of the traditional diagnosis method is low,so that the accuracy of the fault diagnosis cannot meet the predetermined expectations,so we introduce artificial intelligence technology for diagnosis.Secondly,neural networks have many advantages such as parallel distributed processing,self-adaptation,association,memory and clustering,and fault tolerance.They are suitable for transformers,which have complex internal structures and require a certain amount of state information to indicate fault characteristics.Among them,probabilistic neural networks are The radial basis function pre-feedback neural network based on Bayesian decision theory has strong fault tolerance and obvious advantages in pattern classification.However,the performance of the probabilistic neural network is greatly affected by the smoothing factor of the hidden layer unit.It affects its classification performance.Finally,the introduction of meta-heuristic algorithm can well overcome this defect,improve the performance of the probabilistic neural network and extract the most suitable parameters to improve the accuracy of fault diagnosis.Aiming at the characteristics of transformer characteristic gases and fault types,this paper establishes a transformer fault diagnosis simulation model in MATLAB.Two sets of different transformer characteristic gas data sets are used to analyze the probabilistic neural network,the three-ratio method proposed by the International Electrotechnical Commission,and multiple elements.The heuristic algorithm optimizes the probabilistic neural network model for training and testing,and records the fitness curve and mean square error of these model iterations,and the data of important evaluation indicators such as the accuracy of predicting failures.In the experiment using two different data sets,the results of the bat algorithm optimization probabilistic neural network diagnosis accuracy rate in data set 1 reached 98.46%,and the particle swarm algorithm optimization probabilistic neural network diagnosis accuracy rate reached 90.36%;in data set 2 The diagnostic accuracy rate of the probabilistic neural network optimized by the multiverse optimization algorithm reached 97.49%,and the diagnostic accuracy rate of the probabilistic neural network optimized by the particle swarm algorithm reached 95.18%.This is significantly higher than the current international current DL/T 722-2014 three-ratio method and probabilistic neural network model,and the safety and stability of the power network have been improved,and at the same time,the people have a better life in daily life and economic construction.Better protection to avoid many huge impacts and losses caused by power transformer failures and power failures.