Design And Implementation Of Power Transformer Fault Diagnosis System Based On Neural Network

The transformer is one of the core components of the power system and itresponsible for the important task of converting the voltage, it is necessary for thenormal operation of the power supply system. If the transformer fails, then the entirenormal power supply will cause great damage to the power system. Therefore, rapidand accurate diagnosis of transformer fault type has important implications fortransformer fault timely, prompt service restoration. Especially BP neural networkneural network has a simple structure, high parallelism, strong non-linearrelationship modeling ability. It is ideally suited to solve the fault diagnosis of suchmultivariable, complex internal relations issues. Therefore, it is reasonable todiagnosis the transformer fault by neural network theory.XJ Group as China’s power equipment R&D and production of electricitylarge-scale backbone enterprises has always attached great importance to thedevelopment and research of fault diagnosis technology of transformer. In recentyears, the XJ Group has been cooperating with my school teachers to develop asystem for transformer fault diagnosis. Base on the requirements of this project, Imake two respects in this paper. First, the neural network is studied and the BPmodel is applied to transformer fault diagnosis. Second, design the powertransformer fault diagnosis system based on Neural Network and achieve the systemin VC++6.0platform. The system includes three modules: network training,faultdiagnosis and data management. Data management module mainly realizes thepreprocessing of fault samples. Network training module mainly realizes the threealgorithms for network training. The fault diagnosis module mainly realizes thetypes of fault diagnosis samples.In this paper, we have60groups of transformer fault data sourcing from theactual production environment as an example to show the effect of the system. Theaccuracy rate can reach91%using the L-M algorithm in this system. This systemhas been applied in the relevant product of XJ Group, and it can satisfy the userrequirement.