Study On The System In Fault Intelligent Diagnosis Of Power Transformer

The power transformer takes great role in electrical network, its normal operation has a great effect on the electrical power system security, reliable, high quality and the economical movement. But along with transformer rated voltage enhancement, single stage presence quantity enlargement as well as movement age limit growth, the power transformer carries on the preventive test price to be getting higher and higher, therefore, it arises at the historic moment take the online overhaul as fundamental mode's power transformer malfunction examination technology. It takes the transformer current practical work condition as the basis, the recognition fault early time indication, and to the fault location, the breakdown order of severity and the trend of development makes the judgement.This paper in the establishment transformer fault running status appraisal's level indicator system and to the quota, in the stationary index quantification's foundation, has conducted the thorough research to transformer condition appraisal's several key questions.As a result of the power transformer fault's fuzziness and the multiplicity, have the rate of low accuracy at present using the IEC three ratio law in the transformer failure diagnosis. This paper is taken the power transformer's characteristic as the basis.It is taken into account the overall evaluation various factors' influence.Then it combined to the fuzzy logic and the pi-sigma neural network,so we constructed a transformer failure diagnosis model of mixing the pi-sigma neural network. In the research study of the speed's choice, the degree of membership function parameter's renewal and so on many place has made the improvement, it further reduced the prediction error. According trained the mix pi-sigma neural network model to the transformer fault carries on the confirmation and the diagnosis simulation result indicated that this algorithm has the quick convergence rate and the high computational accuracy. The result confirmed this algorithm applies in the power transformer failure diagnosis validity.Studies the classified knowledge from the mass data, when particularly obtains massively has the category label sample price to be high, the increase study solves this question efficient path. The dissolved gas analysis is analyzed in the paper to unify other electrical test result primarily on the basis of the increase type Bayesian sorter the transformer combination failure diagnosis method.This model might accumulate unceasingly consummates the training sample, the automatic corrective net design parameters and the probability distribution parameters, then the diagnosis error is minimized. The experimental results indicate that this algorithms are feasible and effective.