Research On Transformer Fault Diagnosis Based On Learning Spiking Neural P System

With the rapid development of the power system and the improvement of voltage level in the Chinese state grid,the state of power equipment has become a key factor that will affect the continuous and stable power supply.As one of the most important power equipment,the transformer has been the most important part during the inspection.However,when the transformer shows the symptoms of failure,the transformer oil will decompose and release a large amount of hydrogen hydrocarbon gas quickly.The fault types of transformer are defined by the concentration of gas in general,but the accuracy is not high enough.Therefore,an accurate and efficient approach for transformer fault diagnosis is of great significance.With the application of artificial intelligence and big data technology,transformer fault diagnosis based on machine learning has become a research hotspot.However,as a branch of the third-generation neural network,the spiking neural P system is considered to be a combination of the spiking neural network and membrane computing,which have the potential to solve classification problems.Although there are some preliminary explorations on this issue in recent years,these methods still stay in the theoretical stage.In this paper,a novel transformer fault diagnosis approach based on learning spiking neural P system is discussed,the main research contents of this paper are as follows:1.The model and algorithm of learning spiking neural P system are proposed.Firstly,the structure of fuzzy weighted spiking neural P system is introduced.Secondly,the model and learning strategies of learning spiking neural P system is given based on the coding method of the prefrontal cortex neurons and the Widrow-Hoff learning law.Finally,the performance of learning spiking neural P system is verified by UCI datasets and the XOR problem,the comparison with some state-of-art algorithms is given as well.2.The transformer fault diagnosis method based on learning spiking neural P system is explored.First of all,the relationship between the transformer fault and the dissolved gas in oil is investigated.And then,the fault diagnosis method is given,including the ensemble approach of learning spiking neural P system and the framework of fault diagnosis.At last,the model is verified by the simulation experimental and makes a comprehensive comparison with some other fault diagnosis approaches.3.The transformer fault diagnosis based on learning spiking neural P system and Bayesian networks is given.Firstly,the Bayesian networks are established based on the operating condition of the transformer and the characteristics of the relay protection device.Secondly,the learning spiking neural P system is trained to acquire a prior probability.And then,the Monte Carlo simulation and maximum likelihood estimation are implemented to estimate the conditional probability of relay protection devices.At last,the fault cases have shown that the proposed method is an effective and superior method.