Group Expert Intelligent Optimal Combination Diagnosis Of Latent Faults In Transformers

Power transformer is the key equipment of power system and its safe and stable operation is the guarantee of power transmission.The off-line monitoring of transformer based on dissolved gas analysis in oil is to sample and analyze the gas in transformer oil through regular maintenance.Because of the long test period,complex operation procedures and interference of human factors,it is difficult to grasp the information of transformer operation status,and it is often impossible to judge the latent faults in transformer in time.Conventional transformer fault diagnosis methods are based on the content and relative concentration of gases in oil.Because of the complex relationship between transformer fault mechanism and fault symptoms and the fuzziness of fault itself,this method can't accurately judge the fault type,so it is very important to diagnose the latent fault of transformer in time and accurately.Based on transformer dissolved gases analysis,BP neural network is used to train and learn fault samples.The characteristic gases and fault types are used as input and output nodes of the network,the intelligent diagnosis method of transformer fault diagnosis is realized.By optimizing the weights and thresholds of BP neural network through the improvement of ant colony algorithm,the problems of long training time and easy to fall into the minimum value of BP neural network are solved,the convergence speed of BP neural network is accelerated and the global optimization is realized.In this paper,genetic algorithm,fuzzy C-means clustering,particle swarm optimization and BP neural network are synthesized to establish a combined diagnosis model.Through the calculation of the optimal weights,the transformer fault type can be accurately judged.The validity and reliability of the method are proved by large example analysis.An on-line monitoring system of multi-component transformer based on absorption spectroscopy is also designed in this paper.The system adopts the principle of tunable diode laser absorption spectroscopy(TDLAS).The system does not need carrier gas and has the advantages of accurate measurement,high sensitivity and good stability.It can realize fast on-line monitoring.